CN1398118A - Method based on slide window for estimating and equalizing channels of block signals containing pilot - Google Patents

Abstract

The channel estimation and equalization method for block signals containing pilots based on sliding windows belongs to the field of digital television signal transmission, and is characterized in that: it includes the side paths before and after the main path component into a movable window so as to use the window's The interval between the beginning and the end of the window is used to determine the correct channel estimation interval; then the channel impulse response estimate with length N, length M or length M+2×N is obtained; then the positions of the beginning and end of the window are used as The signal and channel impulse response are constructed as the positioning information required for circular convolution, and the data blocks transmitted through the channel are processed into data blocks after frequency domain equalization to cancel channel distortion; in order to improve the time domain resolution of channel estimation, also The time-domain oversampling channel estimation can be performed in the selected sliding window interval. The data block can be an OFDM IDFT block or a single carrier modulated data block, or any combination of the two. It provides a transmission method against channel multipath.

Description

Channel Estimation and Equalization Method for Block Signals Containing Pilots Based on Sliding Window

technical field

The invention belongs to the field of information transmission, especially the information transmission technology in the application of Internet, digital TV, data broadcasting, data communication and the like.

Background technique

After more than ten years of unremitting research and development, Digital Television Terrestrial Broadcasting (DTTB) has achieved many results and reached the stage of realization. Since November 1998, DTTB programs have been broadcast in North America and Europe, and many countries have announced their DTTB system selection and implementation plans. Currently, there are three main DTTB transmission standards in the world:

1) Trellis-Coded 8-Level Vestigial Side-Band (8-VSB) modulation system developed by the Advanced Television Systems Committee (ATSC).

The ATSC digital television standard was developed by the Advanced Television Systems Committee ATSC.

In May 1993, several groups researching digital HDTV in the United States formed the Grand Alliance (GA). After the Advanced Television Test Center (ATTC) conducted on-site testing of the major league system, in September 1995, the American Advanced Television Standards Committee (ATSC) submitted a digital television standard report to the Federal Communications Commission (FCC), which was debated in Congressional hearings. On December 26, FCC officially announced the "Digital Television Standard" ATSC. ATSC not only includes high-definition television (HDTV), but also adds the standard definition television (SDTV) standard. The system transmits high-quality video, audio and ancillary data in 6MHz channels, capable of sending about 19Mbps of aggregate capacity in a 6MHz terrestrial broadcast channel, and about 38Mbps in a 6MHz cable television channel. The compression ratio is 50:1 or higher. The system consists of three subsystems. That is: source coding and compression subsystem; business multiplexing and transmission subsystem and RF transmission subsystem.

2) Coded Orthogonal Frequency Division Multiplexing (COFDM) modulation adopted by Digital Video Terrestrial Broadcasting-Terrestrial (DVB-T) standard.

The DVB-T system was developed by Digital Video Broadcasting (DVB), a European association of public and private organisations.

In 1993, Europe stopped the original digital-analog hybrid HD-MAC system and started the research on digital television broadcasting DVB. television broadcasting) and DVB-T (terrestrial broadcasting) standards. This series of standards takes into account the transmission of digital video and audio, as well as the upcoming multimedia programs. In terms of source coding, the DVB standard stipulates that the digital TV system uses a unified moving picture coding group-2 (MPEG-2) compression method and MPEG-2 transport stream and multiplexing method; COFDM (Coded Orthogonal Frequency Division Multiplexing) modulation technology, which has obvious advantages for anti-multipath interference and mobile reception.

3) Bandwidth Segmented Transmission (BST) Orthogonal Frequency Division Multiplexing OFDM adopted by Integrated Service Digital Broadcasting-Terrestrial (ISDB-T).

The ISDB-T system was developed by the Association of Radio Industries and Businesses (ARIB) in Japan.

Japan is a powerful country in the production of radio and television equipment, has mastered many high-tech radio and television, and is in a leading position in the development of HDTV camera, video, display and other equipment. High-definition TV satellite broadcasting in analog format The Hivision system was developed in Japan and officially broadcast to users. It is the earliest high-definition TV broadcast in the world. However, because it is in the form of an analog signal and uses satellite channels as the transmission medium (bandwidth 24MHz), it is not an all-digital TV broadcast with a high compression ratio. Japan did not show its trend in the early stage of the development boom of the full digitalization of the world's television broadcasting system, and it seems to have disappeared a bit. However, in 1996, Japan suddenly proposed its DTTB (Digital Television Terrestrial Broadcasting) system-ISDB-T (Terrestrial Integrated Services Digital Broadcasting). This scheme is suggested by Japan's DiBEG (Digital Broadcasting Expert Group), so it is also called DiBEG system. The modulation method used by the system is called Band Segment Transmission (BST) OFDM, which consists of a common set of basic frequency blocks called BST segments.

Since there are multiple DTTB systems, many countries and regions are choosing their own DTTB systems. For political and economic considerations, our country should formulate its own digital TV standards according to its national conditions. The terrestrial digital multimedia broadcasting (DMB-T) protocol proposed by Tsinghua University is against this background. Aiming at the above-mentioned problems existing in the three terrestrial digital television systems in the world, a novel terrestrial digital television system suitable for my country's national conditions is proposed. . In the terrestrial digital multimedia broadcasting (DMB-T) protocol proposed by Tsinghua University, its core physical layer technology is Time Domain Synchronous Orthogonal Frequency Division Multiplexing (TDS-OFDM) technology.

We first introduce the general model of channel transmission. After the information sequence Info(n) is transmitted through a channel with an impulse response of h(n), the received signal is

Rec(n)=Info(n)*h(n)+w(n)

Where w(n) is additive noise, and Info(n)*h(n) represents the linear convolution operation of Info(n) and h(n). Due to the impulse response h(n) of the transmission channel, the received signal after linear convolution will produce time spread and intersymbol interference (ISI).

Currently, there are two technologies for effectively eliminating ISI: time domain equalization and orthogonal frequency division multiplexing (OFDM). Time-domain equalization generally inserts a transversal filter (also called a transversal filter) after the matched filter, which consists of a delay line with taps, and the tap interval is equal to the symbol period. The delayed signal of each tap is weighted and then sent to an adding circuit for output, which is in the same form as the finite impulse response filter (FIR), and the added signal is sent to the decision circuit after sampling. The weighting coefficient of each tap is adjustable, and ISI can be eliminated by adjusting the weighting coefficient. The equalization effect of the equalizer is mainly determined by the number of taps and the equalization algorithm. Commonly used equalization algorithms include the zero-forcing algorithm and the least mean square distortion algorithm. There are two types of equalizers: preset and adaptive. There is also noise interference in the actual channel, which will affect the convergence of the equalizer. In order to further improve the performance, a decision feedback equalizer is often used in practical applications. The tap coefficients of the feedback equalizer are determined by the channel impulse response tailing caused by the forward equalizer. The final effect of equalization is to eliminate the channel multiplicative effect (*h(n)) in the received signal y(n)=x(n)*h(n)+w(n), and obtain x(n) +w'(n), where w'(n) is the additive noise after channel and equalization cascade processing, and channel coding and decoding are generally used to eliminate w'(n).

Digital TV in the United States uses a decision feedback equalizer, and the modulation technology uses a digital 8-VSB method.

The equalizer technology is relatively mature and is widely used in various communication fields, but it has two disadvantages: one is that the structure is complex and the cost is higher; The ISI effect is relatively poor. At this time, it is better to adopt Orthogonal Frequency Division Multiplexing (OFDM) technology.

When the ISI delay is in the same order of magnitude as the period of the transmitted symbol, the impact of the ISI becomes severe. Therefore, extending the period of the transmission symbol can effectively overcome the influence of ISI, which is the principle of OFDM to eliminate ISI. OFDM consists of a large number of subcarriers equally spaced in frequency (assuming there are N carriers in total). The serially transmitted symbol sequence is also divided into groups with a length of N, and the N symbols in each group are respectively modulated by N subcarriers, and then sent together. So OFDM is essentially a parallel modulation technique. Extends the symbol period by N times, thus improving the resistance to ISI.

However, when there is ISI in the channel, the orthogonality between OFDM subcarriers will be destroyed, so that the receiver cannot correctly extract the modulation symbols on each subcarrier. Therefore, in practical application, a guard interval Δ needs to be inserted before each OFDM signal period, and the actual transmission period of OFDM becomes Ts=T+Δ. The signal in the guard interval is generated by periodic extension of the OFDM signal, which is equivalent to turning the tail of the OFDM signal back to the front. When the ISI delay does not exceed Δ, since the OFDM signal passes through the channel, it is equivalent to circular convolution with the channel impulse response h(n), which is equivalent to the influence of the frequency response H(k) of the channel, and the OFDM signal Each subcarrier Y(k) of Y experiences different fading. However, the orthogonality between OFDM subcarriers can still be maintained. The receiver only extracts the time-domain signal within the effective OFDM period T to perform discrete Fourier transform to obtain Y(k), and then estimates the channel to obtain the channel impulse response After h(n), do discrete Fourier transform to get H(k) or directly get the channel frequency response H(k), and finally use Y(k)÷H(k) to eliminate the channel frequency response H( k) or in other words the influence of the inter-symbol interference (ISI) generated by the impulse response h(n) of the channel to obtain the demodulated signal.

For the principle of the receiver as described above, we found that the channel estimation H(k) and the maintenance of the orthogonality of each subcarrier of the OFDM signal or use some method to recover (keep or restore the received signal as the OFDM at the transmitting end The circular convolution of the signal and the channel impulse response h(n)) are two important steps to realize the correct demodulation of OFDM. Among the three existing DTTB transmission standards mentioned above, the first ATSC digital television standard is single-carrier technology, while the second digital video terrestrial broadcasting (DVB-T) standard and the third terrestrial integrated service digital (ISDB-T) have adopted OFDM technology. And the main difference between ISDB-T and DVB-T is that it uses a long interleaving and channel coding technology, and there is not much difference. So we mainly discuss DVB-T technology.

Coded Orthogonal Frequency Division Multiplexing (COFDM) is used in the European DVB-T system. One of the meanings of "coding" in COFDM is that some "pilot" signals are randomly inserted in the OFDM spectrum. The so-called "pilot" here refers to such OFDM carriers, which are composed of The data known to the receiver is modulated, and what they transmit is not the modulated data itself, because these data receivers are known to the system. The purpose of setting the pilot frequency is that the system transmits some transmitter parameters or parameters through the data on the pilot frequency. Test the characteristics of the channel.

The role of the pilot in COFDM is very important, and its usefulness includes: synchronization, channel estimation, transmission mode recognition, and tracking phase noise. The data for the modulated pilot is a pseudo-random sequence generated from a predetermined pseudo-random sequence generator.

Regardless of how the position of the pilot frequency changes, the number of carriers used to transmit effective program information in each COFDM symbol is constant, which is 1512 in the 2k mode and 6048 in the 8k mode. Due to the important role of the pilot in the system, in order to ensure the reliability of the data on the pilot and prevent noise interference, the average power of the pilot signal is 16/9 times greater than the average power of other carrier signals, that is, the pilot signal is Transmitted at "elevated" power levels.

Just because OFDM has the above characteristics, it has the following main advantages: (1) resist multipath interference; (2) support mobile reception; (3) can form single frequency network SFN and so on.

However, because FFT and pilot are mutually required in COFDM, in the receiver, the received pilot is obtained after FFT processing, and the FFT calculation needs to be synchronized first (assisted by the pilot), and then can be calculated FFT. Therefore, COFDM uses an iterative approximation algorithm, so there is a problem of convergence error and convergence time. Therefore, in COFDM, synchronization can only be obtained after iterative calculations, and when using pilots for channel estimation, numerical interpolation in the frequency domain is required, and the estimation of the channel frequency response obtained by interpolation is consistent with the actual channel frequency response There is an error in comparison, and the greater the time length of the channel impulse response h(n), that is, the higher the frequency domain resolution of the channel frequency response, the greater the error will be.

In Time Domain Synchronous Orthogonal Frequency Division Multiplexing Modulation (TDS-OFDM), the frequency domain pilot in the OFDM signal is canceled, and the time domain pilot before the OFDM signal frame is used as synchronization and channel estimation. Using TDS-OFDM technology can realize fast synchronization without iterative processing through time-domain pilot. Time Domain Synchronous Orthogonal Frequency Division Multiplexing Modulation (TDS-OFDM) is a published patent application, its name is "Time Domain Synchronous Orthogonal Frequency Division Multiplexing Modulation Method", the application number is 01115520.5, and the publication number is CN1317903A . Moreover, the TDS-OFDM technology can replace the guard interval in the traditional OFDM through the time-domain pilot. Using a pseudo-random PN sequence to replace the guard interval in OFDM and to use it for time synchronization, frequency synchronization and channel estimation is also a published patent application, and its name is "Method for Filling Guard Interval in Orthogonal Frequency Division Multiplexing Modulation System" Application No. It is 01124144.6, and the publication number is CN 1334655A.

Contents of the invention

The present invention is applicable to the signal processing method of the receiver of time domain synchronous orthogonal frequency division multiplexing modulation (TDS-OFDM) technology, the time domain pilot frequency of TDS-OFDM is made up of two or more than two pseudo-random PN sequence periods In the case of composition, a sliding window-based channel estimation and equalization method for block signals containing pilots is proposed.

According to the characteristics of block data processing in the OFDM receiver, we found that if the time-domain pilot of time-domain synchronous orthogonal frequency division multiplexing modulation (TDS-OFDM) consists of two or more periods of pseudo-random PN sequences, And when the time length of each PN sequence is greater than the time length of the channel impulse response, within a TDS-OFDM signal frame, it has already included enough synchronization and channel estimation information, and the original data can be linearly rolled after passing through the channel. After the characteristics of the product are processed, the result of circular convolution between the data and the channel is obtained, because only when the data and the channel are circularly convolved can the channel distortion be offset by simple frequency domain equalization.

The present invention is characterized in that reliable channel estimation can be obtained within a received OFDM signal frame, and correct data demodulation can be realized. This is a fast and reliable channel estimation, and the channel estimation can be obtained within one frame, so that one frame of data can be demodulated after one frame is received. Therefore, the time-domain synchronous orthogonal frequency division multiplexing modulation (TDS-OFDM) receiver can still realize reliable reception under the time-varying channel. In the case of static reception and time-invariant channel, better performance can be obtained by further smoothing and filtering between multiple OFDM signal frames based on the method proposed by the present invention.

In the study of Time Domain Synchronous Orthogonal Frequency Division Multiplexing (TDS-OFDM) technology, we found that the channel estimation and equalization method based on the sliding window for the block signal containing pilot is actually suitable for the whole data block A general method of equalizing the data in the data block to offset the influence of channel multipath on them. The data in this data block can be one OFDM IDFT block, or multiple OFDM IDFT blocks, or multiple OFDM IDFT blocks. A combination of an IDFT block and multiple single carrier modulated data blocks. The purpose of the present invention is to equalize the whole data block once, and the small data blocks can be separately demodulated and processed. And for the case where the entire data block is composed of multiple OFDM IDFT blocks, the original method is to add a cyclic prefix to each OFDM IDFT block. Now it is only necessary to add a time-domain pilot to the entire data block. , this receiver processing method greatly improves the efficiency of information transmission. The key point of this method is that because it supports any combination of multiple OFDM IDFT blocks and multiple single-carrier modulated data blocks, it supports a flexible time-frequency two-dimensional signal design capability. There is more room for signal design and processing to adapt to complex channel environments.

Below we introduce the algorithm flow. There are two very similar calculation methods. The difference is that in the process of constructing data and channel circular convolution, the first method uses some addition and subtraction to compensate the data block, while the second Method When the data block and its previous periodic PN sequence and its subsequent periodic PN sequence are combined as a large signal block, this large signal block is equivalent to a circular convolution with the channel after passing through the channel, so this large signal block can be The signal block is transformed into the frequency domain by FFT and equalized in the frequency domain, and then the obtained frequency domain signal is transformed into the time domain by IFFT. At this time, the large signal block has already compensated for channel distortion, and then the front part of the large signal block is and the two-period PN sequence at the rear are removed, and the remaining data blocks are useful information;

The present invention proposes a channel estimation and equalization method for block signals containing pilots based on a sliding window, which includes a data frame containing time-domain pilots transmitted by the transmitter, and the time-domain pilots are composed of two consecutive or Multiple periods and composed of pseudo-random PN sequences agreed by the transmitter and receiver, characterized in that: during channel estimation, the method includes side path components before and after the main path component into a movable sliding window to Determine the interval of the PN sequence for correct channel estimation, so that the beginning n _b (i) and end n _e (i) of the sliding window determine the interval for correct channel estimation; and then obtain the channel impulse response of length N h _N (n), and then use the beginning of the window n _b (i) and the end of the window n _e (i) as the positioning information for the zero-padding operation on the above h _N (n), to obtain the channel impulse response of length M The estimated h _M (n′) or the estimated channel impulse response h M+2 _{×N (n′) of length M+2×} N; then the beginning of the window n _b (i) and the end of the window n _e (i ) position as the positioning information required to construct the signal and channel impulse response as a circular convolution, and process the data block DATA _r (n) transmitted through the channel into a data block DATA _c (n); when a period of the PN sequence The length is N, and the length of the transmitted time-domain pilot SYN(n) is L (L=S×N), where n represents discrete time, and S is the number of PN cycles in the known time-domain pilot SYN(n), The transmitted data block is DATA(n), and when its length M is variable, it contains the following steps in turn:

(a) Obtain the start time n ₁ (i) of the i-th frame time domain pilot SYN _r (n) and the start time n ₂ (i) of the i-th frame data block DATA _r (n) in the received data stream : The received data stream can be regarded as the superposition of the time-domain pilot SYN _r (n) and the data block DATA _r (n), and after synchronization processing, the i-th frame time-domain pilot SYN _r (n) in the received data stream can be obtained ) of the start time n ₁ (i) and the start time n ₂ (i) of the i-th frame data block DATA _r (n);

(b) Sliding window initialization: Use the sliding window to determine the interval of the PN sequence that can obtain correct channel estimation. The length of the sliding window is equal to the length N of a cycle of the PN sequence. The initialized window interval is any jth in the time domain pilot PN sequence periods, where 1<j<=S, the beginning of the i-th frame sliding window is n _b (i)=n ₁ (i)+L-(S-j+1)*N, and the end is n _e ( i)=n ₁ (i)+L-(Sj)*N, the sliding window can slide within the entire time-domain pilot;

(c) Determine the positions of the beginning n _b (i) and the end n _e (i) of the sliding window: perform circular correlation on the first PN period of the received time-domain pilot SYN _r (n) to obtain R1(τ), Perform circular correlation on the S-th PN cycle in the time-domain pilot SYN _r (n) to get R2(τ), after filtering and smoothing R2(τ) and R1(τ), compare R2(τ) and R1( The amplitude of the effective multipath component with the same delay in τ) is compared from the multipath component with the longest delay. If R2(τ) is smaller than the amplitude of the effective multipath component with the same delay in R1(τ), Then the initial position of the sliding window defined in (b) is incorrect, and the end of the new sliding window n _e (i) is moved forward until the delay is less than (n _e (i)-n ₁ (i))mod When the magnitude of the multipath component in R2(τ) of N is greater than or approximately equal to the magnitude of the multipath component in R1(τ) with the same delay, the sliding stops. Since the end of the window n _e (i) moves, the beginning of the window n _b (i) also move accordingly, keeping the window length unchanged;

(d) Use the positioning information of the window start position n _b (i) and end position n _e (i) to obtain the channel impulse response estimate h _N (n), and then perform zero padding on the above h _N (n) to obtain the length is the estimate h _M (n′) of the channel impulse response of M or the estimate h M+2×N (n′) of the channel impulse response of length _M+2×N :

(d.1) Use any of the following two methods to obtain the estimate h _N (n) of the channel impulse response of length N:

(d.1.1) Define a period of time-domain pilot received by the receiver in the selected sliding window interval (n∈[n _b (i), n _e (i)]) as pilot(n), take the known In the time-domain pilot SYN(n) transmitted by the transmitter, the pseudo-random PN sequence with a period length determined by the sliding window interval (n∈[n _b (i), _ne (i)]) is pn _c (n ), using pn _c (n) to perform circular correlation on pilot (n), the estimate h _N (n) of the channel impulse response with length N can be obtained. (or use a version of pn _c (n) with a circular shift of shift bits pn _N ′(n) to perform circular correlation on pilot(n) to obtain h _N ″(n), h _N ″(n) is equal to h _N (n) is shifted by shift bits in the circle, and h _N ″(n) is shifted by shift bits in the opposite direction to get h _N (n)); this is the method of channel estimation in the time domain, and it is also mathematically equivalent The method for channel estimation in the frequency domain, the process is: do FFT to the above-mentioned pilot (n) to obtain PILOT (k), do FFT to the above-mentioned pn _c (n) to obtain PN _c (k), calculate PILOT (k) ÷ PN _c (k) = H _N (k), and then perform N-point IFFT on H _N (k) of length N to obtain h _N (n).

(d.1.2) The estimated channel impulse response h _N (n) can also be obtained from the obtained R1(τ) and R2(τ) according to the following formula, and the following formula is used for the moving operation:

(1). h _N (n) = R1 (τ),

where τ∈[(n _e (i)-n ₁ (i)) mod N+1, N], n ∈ [(n _e (i)-n ₁ (i)) mod N+1, N];

(2). h _N (n) = R2 (τ),

where τ∈[1, (n _e (i)-n ₁ (i)) mod N], n ∈ [1, (n _e (i)-n ₁ (i)) mod N];

(d.2) The h _N (n) of length N obtained by using the method of time domain or frequency domain is zero-filled according to the following formula to obtain h _M (n′) of length M, n is from 1 to N, n' from 1 to M:

(1). h _M (n') = h _N (n),

where n′∈[1, (n _e (i)-n ₁ (i)) mod N], n ∈ [1, (n _e (i)-n ₁ (i)) mod N];

(2). h _M (n') = h _N (n),

where n′∈[M-(N-(n _e (i)-n ₁ (i)) mod N)+1, M], n ∈ [(n _e (i)-n ₁ (i)) mod N +1,N];

(3). h _M (n') = 0,

where n′∈[(n _e (i)-n ₁ (i)) mod N+1, M-(N-(n _e (i)-n ₁ (i)) mod N)];

Then perform FFT on h _M (n') to get H _M (k), and H _M (k) will be used for the final frequency domain equalization;

The h _N (n) of length N obtained by using the method of time domain or frequency domain is zero-filled according to the following formula, and the length of h M+2× _{N (n′) of M+2×} N is obtained, and n starts from 1 to N, n' from 1 to M+2×N:

(1).h _M+2×N (n')=h _N (n),

where n′∈[1, (n _e (i)-n ₁ (i)) mod N], n ∈ [1, (n _e (i)-n ₁ (i)) mod N];

(2).h _M+2×N (n')=h _N (n),

where n′∈[M+2×N-(N-(n _e (i)-n ₁ (i))mod N)+1, M+2×N]n∈[(n _e (i)-n ₁ (i))mod N+1,N];

(3).h _M+2×N (n')=0,

where n′∈[(n _e (i)-n ₁ (i)) mod N+1, M+2×N-(N-(n _e (i)-n ₁ (i)) mod N)];

Then perform FFT on h _M+2×N (n′) to obtain H _M+2×N (k), and H _M+2×N (k) will be used for the final frequency domain equalization.

(e) Process the received data blocks according to the above times n ₁ (i), n ₂ (i) and window positions n _b (i), n _e (i), and construct the signal and channel impulse responses as cyclic volumes Product relationship, so that the frequency domain equalization can be performed in the next step to offset the channel distortion, so that the receiver can correctly restore the signal transmitted by the transmitter: after the transmitted data block DATA(n) is transmitted through the channel, it is actually linear with the impulse response of the channel The relationship between convolution, in order to facilitate the frequency domain equalization to offset the distortion of the channel, the following processing is required to make the impulse response of the data and the channel form a circular convolution relationship; after obtaining n ₁ (i), n ₂ (i) and After the window positions n _b (i) and n _e (i), the data block DATA _r (n) transmitted through the channel is processed through the following steps to obtain DATA _c (n), whose length is M:

(1). DATA _c (nn ₂ (i)) = DATA _r (n) + SYN _r (n+M) - SYN _r (nN),

where n∈[ _n2 (i)+1， _n2 (i)+( _ne (i) _-n1 (i)) mod N-1];

(2). DATA _c (nn ₂ (i)) = DATA _r (n) + SYN _r (nM) - SYN _r (nMN),

where n∈[n ₂ (i)+M-(N-(n _e (i)-n ₁ (i)) mod N), n ₂ (i)+M];

(3). DATA _c (nn ₂ (i)) = DATA _r (n),

where n∈[n ₂ (i)+(n _e (i)-n ₁ (i)) mod N, n ₂ (i)+M-(N-(n _e (i)-n ₁ (i)) mod N)-1];

After the transmitted data block DATA(n) is transmitted through the channel, it has a linear convolution relationship with the impulse response of the channel, but if the data block DATA(n) is considered together with the PN sequence of the previous cycle and the next cycle , they have formed a circular convolution relationship with the impulse response of the channel after passing through the channel; the data block DATA _r (n) after channel transmission and the PN sequence of the previous cycle and the next cycle are defined as DATA _{M+ 2×N} (n), whose length is M+2×N, for the next step of processing;

(f) Find the frequency domain signal X(k) after frequency domain equalization: first perform fast Fourier transform (FFT) on the DATA _c (n) obtained through the above step (e) to obtain Y(k), and then use Divide Y(k) by the estimated channel frequency response H _M (k), that is, Y(k)/H _M (k)=X(k), to obtain the frequency domain signal X(k) after frequency domain equalization; or The DATA _M+2×N (n) obtained through the above step (e) is subjected to fast Fourier transform (FFT) to obtain Y _M+2×N (k), and then divided by Y _M+2×N (k) The estimated channel frequency response H _M+2×N (k) obtained through the above step (d), that is, Y _M+2×N (k)/H _M+2×N (k)=X _{M+2 ×N} (k), to obtain the frequency domain signal X _M+2×N (k) after frequency domain equalization, and then perform an inverse fast Fourier transform (IFFT) on X _M+2×N (k) to obtain x _{M+ 2×N} (n), remove the PN sequence of the first N points and the PN sequence of the last N points of x _M+2×N (n) to obtain the time domain signal x _M (n), and x _M (n) is the frequency domain signal The time-domain form of X(k).

According to the above-mentioned channel estimation and equalization method to the block signal containing pilot based on the sliding window, it is characterized in that: the data block DATA(n) sent by the transmitter is an OFDM inverse discrete Fourier transform (IDFT) data block, then output the obtained X(k) as the result after equalization, or output the obtained x _M (n) after M-point Fast Discrete Fourier Transform (FFT).

According to the channel estimation and equalization method for the block signal containing pilot based on the sliding window as described above, it is characterized in that: the data block DATA(n) sent by the transmitter is a single carrier modulated data block, then Perform M-point IFFT on the obtained X(k) again, and output the obtained result as the equalized result; or output the obtained x _M (n) as the result.

According to the channel estimation and equalization method for block signals containing pilots based on the sliding window as described above, it is characterized in that: the data block DATA(n) sent by the transmitter is several OFDM data blocks and several single For any combination of carrier-modulated data blocks, the obtained frequency-domain signal X(k) is first subjected to an M-point inverse fast Fourier transform (IFFT) to obtain the data block DATA _block (n)=IFFT(X(k) ), where the DATA _block (n) and x _M (n) are mathematically equivalent, and then according to these OFDM and single-carrier block sub-data blocks agreed in some way by the transmitter and receiver in the data block DATA _block The positions and sizes in (n) are respectively positioned and processed for these data blocks. For the OFDM data blocks, FFT needs to be performed again to obtain the equalized result signal, and the single-carrier block signal is directly output.

A channel estimation and equalization method for block signals containing pilots based on a sliding window, which contains a data frame containing time-domain pilots transmitted by the transmitter, whose time-domain pilots consist of two or more consecutive periods and are composed of The pseudo-random PN sequence agreed by the transmitter and the receiver is characterized in that: during channel estimation, the method includes the side path components before and after the main path component into a movable sliding window to determine the correct progress The interval of the PN sequence of channel estimation, so that the beginning n _b (i) and end n _e (i) of the sliding window determine the interval for obtaining correct channel estimation; in order to improve the time domain resolution of channel estimation, the selected Perform time-domain oversampling in the sliding window interval and then perform oversampled channel estimation to obtain an estimate h _{N_oversample} (n) of the channel impulse response with a length of N×Fs, and then use the window start n _b (i) and window end n _e (i) As the location information of the above h _{N_oversample} (n) with zero-padding operation, the estimated channel impulse response h _{M_oversample} (n′) with a length of M×Fs or a length of (M+2×N)×Fs is obtained The estimate of the channel impulse response h _{M+2×N_oversample} (n′); then the position of the beginning n _b (i) of the window and the position of the end n _e (i) of the window is used as the result of constructing the signal and channel impulse response as circular convolution The required positioning information processes the data block DATA _{r_oversample} (n) after channel transmission and time-domain oversampling by the receiver into a data block DATA _{c_oversample} (n); when the length of a cycle of the PN sequence is N, when transmitting The length of the domain pilot SYN(n) is L (L=S×N), where n represents discrete time, S is the number of PN periods in the known time domain pilot SYN(n), and the transmitted data block is DATA( n), when its length M is variable, then it contains the following steps in turn:

(a) Obtain the start time n ₁ (i) of the i-th frame time domain pilot SYN _r (n) and the start time n ₂ (i) of the i-th frame data block DATA _r (n) in the received data stream : The received data stream can be regarded as the superposition of the time-domain pilot SYN _r (n) and the data block DATA _r (n), and after synchronization processing, the i-th frame time-domain pilot SYN _r (n) in the received data stream can be obtained ) of the start time n ₁ (i) and the start time n ₂ (i) of the i-th frame data block DATA _r (n);

(b) Sliding window initialization: Use the sliding window to determine the interval of the PN sequence that can obtain correct channel estimation. The length of the sliding window is equal to the length N of a cycle of the PN sequence. The initialized window interval is any jth in the time domain pilot PN sequence periods, where 1<j<=S, the beginning of the i-th frame sliding window is n _b (i)=n ₁ (i)+L-(S-j+1)*N, and the end is n _e ( i)=n ₁ (i)+L-(Sj)*N, the sliding window can slide within the entire time-domain pilot;

(c) Determine the positions of the beginning n _b (i) and the end n _e (i) of the sliding window: perform circular correlation on the first PN period of the received time-domain pilot SYN _r (n) to obtain R1(τ), Perform circular correlation on the S-th PN cycle in the time-domain pilot SYN _r (n) to get R2(τ), after filtering and smoothing R2(τ) and R1(τ), compare R2(τ) and R1( The amplitude of the effective multipath component with the same delay in τ) is compared from the multipath component with the longest delay. If R2(τ) is smaller than the amplitude of the effective multipath component with the same delay in R1(τ), Then the initial position of the sliding window defined in (b) is incorrect, and the end of the new sliding window n _e (i) is moved forward until the delay is less than (n _e (i)-n ₁ (i))mod When the magnitude of the multipath component in R2(τ) of N is greater than or approximately equal to the magnitude of the multipath component in R1(τ) with the same delay, the sliding stops. Since the end of the window n _e (i) moves, the beginning of the window n _b (i) also move accordingly, keeping the window length unchanged;

(d) Obtain the estimated h _{N_oversample} (n) of the oversampled channel impulse response by using the positioning information of the window start position n _b (i) and the end position n _e (i), and then perform zero padding on the above h _{N_oversample} (n) The estimation h _{M_oversample} (n′) of the channel impulse response with a length of M×Fs or the estimation h M+2 _{×N_oversample (} n′) of a channel impulse response with a length of (M+2×N)×Fs is obtained by processing:

(d.1) In order to improve the time-domain resolution of channel estimation, over-sampling in the time-domain can be performed within the selected sliding window interval and then over-sampled channel estimation: set the over-sampling coefficient as Fs, set the transmitter and receiver The delay of the machine-side band-pass filter is SRRC_Delay, and the oversampled period of time-domain guide received by the receiver within the selected sliding window interval (n∈[n _b (i), n _e (i)]) The frequency is pilot _oversample (n), a cycle length determined by the sliding window interval (n ∈ [n _b (i), n _e (i)]) in the time-domain pilot SYN (n) transmitted by the known transmitter The pseudo-random PN sequence of is pn _c (n), which is interpolated with the sampling coefficient Fs (that is, Fs-1 zeros are inserted after each element of pn _c (n)) to obtain pn _{c_oversample} (n), and then can be obtained from Choose one of the following methods:

The method in the time domain is: use pn _{c_oversample} (n) to perform circular correlation on pilot _oversample (n) to obtain the estimate h _{N_oversample} (n) of the channel impulse response with a length of N×Fs (a circular shift of pn _{c_oversample} (n) can also be used The version pn′ _{c_oversample} (n) of the bit shift bit is used to perform circular correlation on pilot _oversample (n) to obtain h″ _{N_oversample} (n), and h″ _{N_oversample} (n) is equal to shifting h _{N_oversample} (n) circle by shift bit, and the h″ _{N_oversample} (n) shifts the shift position in the opposite direction to get h _{N_oversample} (n));

The frequency domain method is: do FFT to pilot _oversample (n) to get PILOT _oversample (k), do FFT to above-mentioned pn _{c_oversample} (n) to get PN _{c_oversample} (k), calculate PILOT _oversample (k)÷PN _{c_oversample} (k )=H _{N_oversample} (k), and then doing N×Fs point IFFT to H _{N_oversample} (k) whose length is N×Fs can also obtain h _{N_oversample} (n);

(d.2) After obtaining h _{N_oversample} (n), you need to perform zero padding operation. Before zero padding, you first need to adjust n _e (i) to n _e ′(i) according to the following formula for zero padding operation: n _e ′ (i)=min((n _e (i)-n ₁ (i))mod N+SRRC_Delay, N-SRRC_Delay), and then perform zero padding on h _{N_oversample} (n) of length N×Fs to obtain a length of M The h _{M_oversample} (n′) of ×Fs, n from 1 to N×Fs, n′ from 1 to M×Fs, the zero padding operation is:

(1). h _{M_oversample} (n') = h _{N_oversample} (n),

where n′∈[1,ne _′ (i)×Fs], n∈[1, _ne ′(i)×Fs];

(2). h _{M_oversample} (n') = h _{N_oversample} (n),

where n′∈[M×Fs-(Nn _e ′(i))×Fs+1, M×Fs], n∈[n _e ′(i)×Fs+1, N×Fs];

(3). h _{M_oversample} (n') = 0,

where n′∈[n _e ′(i)×Fs+1, M×Fs-(Nn _e ′(i))×Fs];

Then perform FFT on h _{M_oversample} (n′) to obtain H _{M_oversample} (k), and H _{M_oversample} (k) can be used for the final frequency domain equalization; or:

After h _{N_oversample} (n) is obtained, zero padding operation is required. Before zero padding, ne (i) needs to be adjusted to n _e ′(i) according to the following formula for zero padding operation: _{n e ′} (i)=min ((n _e (i)-n ₁ (i))mod N+SRRC_Delay, N-SRRC_Delay), and then perform zero padding on h _{N_oversample} (n) of length N×Fs to obtain a length of (M+2×N )×Fs h _{M+2×N_oversample} (n′), n is from 1 to N×Fs, n′ is from 1 to (M+2×N)×Fs, the zero padding operation is:

(1).h _{M+2×N_oversample} (n′)=h _{N_oversample} (n),

where n′∈[1, _ne ′(i)×Fs], n∈[1, _ne ′(i)×Fs];

(2).h _{M+2×N_oversample} (n′)=h _{N_oversample} (n),

where n′∈[(M+2×N)×Fs-(Nn _e ′(i))×Fs+1, (M+2×N)×Fs]

n∈[n _e '(i)×Fs+1, N×Fs];

(3).h _{M+2×N_oversample} (n')=0,

where n′∈[n _e ′(i)×Fs+1, (M+2×N)×Fs-(Nn _e ′(i))×Fs];

Then perform FFT on h _{M+2×N_oversample} (n′) to obtain H _{M+2×N_oversample} (k), and H _{M+2×N_oversample} (k) can be used for final frequency domain equalization.

(e) Process the received data blocks according to the above times n ₁ (i), n ₂ (i) and window positions n _b (i), n _e (i), and construct the signal and channel impulse responses as cyclic volumes Product relationship, so that the frequency domain equalization can be done in the next step to offset the channel distortion, so that the receiver can correctly recover the signal transmitted by the transmitter: For the case of oversampling, the receiver will transmit the data block DATA _r (n) through the channel Perform oversampling to obtain DATA _{r_oversample} (n), perform oversampling on the time-domain pilot SYN _r (n) after channel transmission to obtain SYN _{r_oversample} (n), process DATA _{r_oversample} (n) through the following steps to obtain DATA _{c_oversample} (n ), whose length is M×Fs:

(1) DATA _{c_oversample} (nn ₂ (i)×Fs)＝DATA _{r_oversample} (n)+SYN _{r_oversample} (n+M×Fs)-SYN _{r_oversample} (nN×Fs)

where n∈[n ₂ (i)×Fs+1, n ₂ (i)×Fs+ _ne ′(i)×Fs-1];

(2). DATA _{c_oversample} (nn ₂ (i)×Fs)=DATA _{r_oversample} (n)+SYN _{r_oversample} (nM×Fs)-SYN _{r_oversample} (nM×Fs-N×Fs),

where n∈[n ₂ (i)×Fs+M×Fs-(Nn _e ′(i))×Fs-Fs+1, n ₂ (i)×Fs+M×Fs];

(3). DATA _{c_oversample} (nn ₂ (i) × Fs) = DATA _{r_oversample} (n),

where n∈[n ₂ (i)×Fs+n _e ′(i)×Fs, n ₂ (i)×Fs+M×Fs-(Nn _e ′(i))×Fs-Fs];

where n _e '(i)=min((n _e (i)-n ₁ (i)) mod N+SRRC_Delay, N-SRRC_Delay); or:

For the case of oversampling, the data block DATA _r (n) after channel transmission and the PN sequence of the previous cycle and the next cycle are defined as DATA _M+2×N (n), for DATA _{M+2× N} (n) is oversampled to obtain DATA _{M+2×N_oversample} (n), and its length is (M+2×N)×Fs, which is used for the next step;

(f) Find the frequency domain signal X(k) after frequency domain equalization: first use the oversampled version DATA _{c_oversample} (n) of DATA _c (n) to perform fast Fourier transform (FFT) to obtain Y _oversample (k), and then Divide Y _oversample (k) by the estimate H _{M_oversample} (K) of the channel frequency response after oversampling, that is, Y _oversample (k)/H _{M_oversample} (K)=X _oversample (k), and obtain the frequency domain equalized Frequency domain signal X(k):

(1), X(k)=X _oversample (k')

Among them, k∈[1, M÷2], k′∈[1, M÷2]

(2), X(k)=X _oversample (k')

Among them, k∈[M÷2+1, M], k′∈[(Fs-1)×M+M÷2+1, Fs×M]

or:

Use the oversampled version of DATA _M+2×N (n) DATA _{M+2×N_oversample} (n) for fast Fourier transform (FFT) to get Y _{M+2×N_oversample} (k), and then use Y _{M+2× N_oversample} (k) is divided by the estimated channel frequency response after oversampling H _{M+2×N_oversample} (K), that is, Y _{M+2×N_oversample} (k)/H _{M+2×N_oversample} (K)=X _{M+2 ×N_oversample} (k), the frequency domain signal X _M+2×N (k) after frequency domain equalization can be obtained as follows:

(1), X _M+2×N (k)=X _{M+2×N_oversample} (k′)

Among them, k∈[1, M÷2], k′∈[1, M÷2]

(2), X _M+2×N (k)=X _{M+2×N_oversample} (k′)

Among them, k∈[M÷2+1, M], k′∈[(Fs-1)×M+M÷2+1, Fs×M]

Perform an M _{+2×N point IFFT on X M+2×N} (k), and obtain x _M+2×N (n)=IFFT(X _M+2×N (k)). Remove the PN sequence of the first N points and the PN sequence of the last N points of x _M+2×N (n) to obtain x _M (n), and x _M (n) is the time domain form of the frequency domain signal X(k).

According to the above-mentioned channel estimation and equalization method to the block signal containing pilot based on the sliding window, it is characterized in that: the data block DATA(n) sent by the transmitter is an OFDM inverse discrete Fourier transform (IDFT) data block, then output the obtained X(k) as the result after equalization, or output the obtained x _M (n) after M-point Fast Discrete Fourier Transform (FFT).

According to the channel estimation and equalization method for the block signal containing pilot based on the sliding window as described above, it is characterized in that: the data block DATA(n) sent by the transmitter is a single carrier modulated data block, then Perform M-point IFFT on the obtained X(k) again, and output the obtained result as the equalized result; or output the obtained x _M (n) as the result.

According to the channel estimation and equalization method for block signals containing pilots based on the sliding window as described above, it is characterized in that: the data block DATA(n) sent by the transmitter is several OFDM data blocks and several single For any combination of carrier-modulated data blocks, the obtained frequency-domain signal X(k) is first subjected to an M-point inverse fast Fourier transform (IFFT) to obtain the data block DATA _block (n)=IFFT(X(k) ), where the DATA _block (n) and x _M (n) are mathematically equivalent, and then according to these OFDM and single-carrier block sub-data blocks agreed in some way by the transmitter and receiver in the data block DATA _block The positions and sizes in (n) are respectively positioned and processed for these data blocks. For the OFDM data blocks, FFT needs to be performed again to obtain the equalized result signal, and the single-carrier block signal is directly output.

Features and effects of the present invention:

The present invention is characterized in that reliable channel estimation is obtained within a received data frame, and signals transmitted through multipath can be recovered without iteration between multiple frames. So that the receiver can still achieve reliable reception under the time-varying channel. Description of drawings:

Figure 1A depicts the general channel transmission model.

Figure 1B depicts the linear convolution of two signals.

Figure 1C depicts the period-extended linear convolution of two signals.

Figure 1D depicts the circular convolution of two signals and the relationship between circular and linear convolution.

Fig. 2 depicts a schematic diagram of the structure of a general OFDM receiver.

FIG. 3 depicts an optional OFDM frame structure method for filling guard intervals with PN sequences described in the published patent application "Method for Filling Guard Intervals in Orthogonal Frequency Division Multiplexing Modulation Systems".

Figure 4A depicts the total effect of typical multipath interference (dotted line) on the time-domain pilot and data components.

Figure 4B depicts the impact of typical multipath interference (dotted line) on the time-domain pilot portion.

Figure 4C depicts the effect of typical multipath interference (dotted line) on the data portion.

FIG. 5 is a schematic structural diagram of a receiver implemented in the present invention.

Fig. 6A is an algorithm flow chart of Method 1 in the implementation process of the present invention.

Fig. 6B is an algorithm flow chart of Method 2 in the implementation process of the present invention.

FIG. 6C is an algorithm flow chart of the oversampling method of Method 1 in the implementation process of the present invention.

FIG. 6D is an algorithm flow chart of the oversampling method of Method 2 in the implementation process of the present invention.

Figure 7. Schematic diagram of the initial sliding window.

Fig. 8A is a schematic diagram of the positioning sliding window method of the present invention, h(n) of an actual channel.

Fig. 8B is a schematic diagram of positioning the sliding window from the obtained R1(τ) and R2(τ) according to the present invention.

Fig. 9A is a schematic diagram of obtaining h _N (n) from R1(τ) and R2(τ) using sliding window positioning information, and the obtained R1(τ) and R2(τ).

Fig. 9B is a schematic diagram of obtaining h _N (n) from R1(τ) and R2(τ) using sliding window positioning information.

Figure 9C is a schematic diagram of obtaining _hN (n) from R1(τ) and R2(τ) with sliding window positioning information, period expansion of _hN (n).

FIG. 10A is a schematic diagram of zero padding the estimated channel impulse response with the positioning information of the sliding window, two parts of h _N (n).

FIG. 10B is a schematic diagram of padding the estimated channel impulse response with zeros using the positioning information of the sliding window. h _M (n′) is obtained by moving and padding the two parts of h _N (n) respectively.

Fig. 11A1, Fig. 11A2, Fig. 11B1, Fig. 11B2, Fig. 11C1 and Fig. 11C2 are schematic diagrams of methods for constructing data and channel circular convolution results with some addition, subtraction and data moving operations.

FIG. 12A and FIG. 12B are schematic diagrams of taking a part of the result after frequency domain equalization in the oversampling method as output frequency domain data.

Detailed ways:

We first introduce the general model of channel transmission. As shown in Figure 1A, the signal received by the information sequence Info(n) after being transmitted through a channel with an impulse response of h(n) is:

Rec(n)=Info(n)*h(n)+w(n)

Where w(n) is additive noise, and Info(n)*h(n) represents the linear convolution operation of Info(n) and h(n). Since there is an impulse response h(n) of the transmission channel, h(n) is composed of some multipath components with different delays and different amplitudes and phases, which means that during the transmission process, the information passes through multiple channel reflections and refractions. After each path reaches the receiver with different attenuation and delay, the signal Rec(n) received by linear convolution will produce time dispersion and intersymbol interference (ISI). Here Info(n), h(n), w(n) and Rec(n) are complex-valued functions of n, where n represents a discrete time variable.

，则y(n)的FFT运算结果Y(k)与Info(n)的FFT运算结果INFO(k)和h(n)的FFT运算结果H(k)有这样的关系Y(k)＝INFO(k)×H(K)，即Y(k)等于INFO(k)和H(k)的乘积。这是进行频域均衡的基本思想，接收机估计出H(k)，用Y(k)除以H(k)就补偿了信道的失真得到INFO(k)。一般的OFDM接收机的结构如图2所示，假设要传送的信号是In(n)，一般的OFDM发射机发射的信号是In(n)的作IFFT后的信号，即IFFT(In(n))，并且将IFFT(In(n))的后部的一段数据搬移到IFFT(In(n))之前作为OFDM信号的保护间隔，这样通过信道后使得IFFT(In(n))与信道冲激响应h(n)相当于进行了一个循环卷积，(由于保护间隔信号的扩散叠加到IFFT(In(n))的前部)，得到 ，在接收机中将

作FFT得到

，再用估计出H(k)去除上式，就恢复出信息In(n)。我们看出要使用频域均衡的关键之处就是要首先估计出H(k)，而且要构造出信号与信道的循环卷积的结果。Let's introduce the difference between linear convolution and circular convolution, as shown in Figure 1B, Figure 1C and Figure 1D, where each multipath component of h(n) is not drawn, but is represented by an envelope. Info(n) has a finite length, the length is N, after the linear convolution of Info(n) and h(n), Info(n) undergoes backward signal diffusion and forward signal diffusion, where the backward and Forward is relative to the main path signal. Due to the complex channel environment, some small paths with greater attenuation may reach the receiver first, and then a main path signal with the least attenuation reaches the receiver, and then some attenuation is greater. The small path of the signal reaches the receiver, resulting in backward and forward spread of the signal relative to the main path. Accompanying drawing 1B represents the result that Info (n) and h (n) carry out linear convolution, and Fig. 1 C represents that Info (n) and h (n) are carried out the periodic signal after carrying out periodic expansion with N as period to carry out linear convolution respectively As a result, Fig. 1D shows that for Info(n) (one period of the periodic signal after the period extension of Info(n) is Info(n) itself) and h(n) after the period extension The result of circular convolution of one period of the periodic signal is equivalent to the result of linear convolution of the periodic signal after Info(n) and h(n) are periodically extended with N as the period, and the length of one period is taken. From Figure 1D, it can be seen that the backward diffusion part of the signal obtained by the linear convolution of Info(n) and h(n) is superimposed on the head of the signal, and the forward diffusion part of the signal is superimposed on the tail of the signal , keeping the signal length unchanged, it constitutes the result of circular convolution of one cycle of the periodic signal after the cycle extension of Info(n) and h(n). According to the theory of digital signal processing, if the result of circular convolution of two signals Info(n) and h(n) is y(n), that is

, then the FFT operation result Y(k) of y(n) has such a relationship with the FFT operation result INFO(k) of Info(n) and the FFT operation result H(k) of h(n) Y(k)=INFO (k)×H(K), that is, Y(k) is equal to the product of INFO(k) and H(k). This is the basic idea of frequency domain equalization. The receiver estimates H(k), and divides Y(k) by H(k) to compensate channel distortion to obtain INFO(k). The structure of a general OFDM receiver is shown in Figure 2. Assuming that the signal to be transmitted is In(n), the signal transmitted by a general OFDM transmitter is the IFFT signal of In(n), that is, IFFT(In(n )), and move a piece of data at the back of IFFT(In(n)) to before IFFT(In(n)) as the guard interval of the OFDM signal, so that after passing through the channel, IFFT(In(n)) and the channel conflict The excitation response h(n) is equivalent to performing a circular convolution (due to the diffusion of the guard interval signal superimposed on the front of the IFFT(In(n))), we get , in the receiver will

Do FFT to get

, and then use the estimated H(k) to remove the above formula, and then recover the information In(n). We see that the key to using frequency domain equalization is to first estimate H(k), and to construct the result of the circular convolution of the signal and the channel.

In the scheme proposed by Tsinghua University as shown in Figure 3, the guard interval of ordinary OFDM is gone, and time-domain pilots are used instead. The advantages of time-domain pilots have been mentioned above. At this time, in order to use frequency domain equalization We must devise methods that can construct the result of the circular convolution of the signal and the channel.

As shown in FIG. 4A , the influence of the channel impulse response received by the data frame including the time-domain pilot after being transmitted through the channel is represented by a dotted line. The data blocks can be OFDM signals. The frame structure at this time is the TDS-OFDM scheme proposed by Tsinghua University. The data blocks can also be single-carrier signals, or a mixed signal of several OFDM blocks and several single-carrier blocks. In the invention, regardless of the structure of these signals in the data block, the method of the present invention is used for channel estimation, and the distortion produced by the channel to the data block is offset through equalization. This data block can then be further demodulated for output. Since the data stream is composed of two parts, the time-domain pilot stream and the data block stream transmitted through the channel, as shown in Figure 4B and Figure 4C, we denote the time-domain pilot stream as SYN _r (n), and the data block The stream is denoted as DATA _r (n), and the actual received data stream is denoted as recv(n), as shown in Figure 4A, which is the superposition of the time-domain pilot stream SYN _r (n) and the data block stream DATA _r (n) .

A schematic diagram of a receiver for such a data stream is shown in FIG. 5 . The synchronization module obtains the time-domain pilot SYN _r (n) start time n ₁ (i) of the i-th frame in the received TDS-OFDM data through synchronous processing, and the IDFT block OFDM of the OFDM of the i-th frame in the TDS-OFDM data _r (n) starts at time n ₂ (i). The specific processing method is the same as the synchronous processing method of Direct Sequence Spread Spectrum Code Division Multiple Access (DS-CDMA), see "Spread Spectrum Communication" edited by Zha Guangming et al. (Spread Spectrum Communication, Xidian University Press, 1990, pp. 97-108). The purpose of the sliding window positioning module is to obtain the correct channel estimation interval, the beginning of which is n _b (i) and the end is n _e (i). This information will also be used for the construction of the result of the circular convolution of the data and the channel , and are used to zero pad the estimated channel impulse response. The channel estimation module is used to obtain the estimation of channel impulse response. Constructing a circular convolution characteristic module. The DATA _r (n) block after the channel transmission is obtained through some moving and addition and subtraction operations to obtain the result of the circular convolution between the data and the channel. This work can also be replaced by another method, the method That is, when the received data block and its previous periodic PN sequence and its subsequent periodic PN sequence are combined as a large signal block, this large signal block is equivalent to a circular convolution with the channel after passing through the channel, so it can be Transform the large signal block into the frequency domain by FFT, and perform frequency domain equalization. After that, the obtained frequency domain signal is transformed into the time domain by IFFT. At this time, the large signal block has already compensated for channel distortion, and then the large signal block The PN sequences of the two periods in the front and rear are removed, and the remaining data blocks are useful information. The FFT module performs fast Fourier transform (FFT) on the data block DATA _c (n) constructed by circular convolution to obtain Y(k), or combines the received data block with its previous period PN sequence and the subsequent one A large signal block formed by combining periodic PN sequences is subjected to Fast Fourier Transform (FFT). The two results are equalized in the frequency domain and then demodulated for the content of the data block to obtain the results.

Figure 6A and Figure 6B are the algorithm flow charts of the present invention, the calculation process of the two is similar, the difference is that in the process of constructing data and channel circular convolution, Figure 6A describes the calculation process of using some addition and subtraction methods to compensate the data block , Figure 6B describes that when the data block and its previous periodic PN sequence and its subsequent periodic PN sequence are combined as a large signal block, this large signal block is equivalent to performing circular convolution with the channel after passing through the channel, Therefore, the large signal block can be transformed into the frequency domain by FFT and equalized in the frequency domain, and then the obtained frequency domain signal can be transformed into the time domain by IFFT. At this time, the large signal block has compensated for channel distortion, and then the large The PN sequences of the two periods before and after the signal block are removed, and the remaining data blocks are useful information. Both methods have oversampling methods corresponding to them respectively. The oversampling method in Fig. 6A is shown in Fig. 6C, and the oversampling method in Fig. 6B The oversampling method is shown in Figure 6D; now we first describe the variables and symbols used therein, and then explain the principles of some main steps in detail with reference to the accompanying drawings.

corr(): Indicates correlation operation.

abs(): Indicates the modulo value operation.

SYNC_PROC(): Indicates synchronous processing.

oversample(): Indicates time oversampling processing.

Interpolation(): Indicates interpolation processing, inserting Fs-1 zeros after each element of the input vector.

DATA(n): Indicates the data block in the transmitted data frame.

DATA _r (n): Indicates the data blocks separated according to the positioning information n ₁ (i) and n ₂ (i) in the received data frame.

DATA _{r_oversample} (n): obtained by oversampling DATA _r (n).

DATA _c (n): The data block obtained by constructing circular convolution on DATA _r (n).

DATA _{c_oversample} (n): obtained by oversampling DATA _c (n).

DATA _M+2×N (n): The data block DATA _r (n) after channel transmission and the PN sequence of the previous cycle and the next cycle are defined as DATA _M+2×N (n), where The length is M+2×N.

DATA _{M+2×N_oversample} (n): obtained by oversampling DATA _M+2×N (n).

SYN(n): time-domain pilot transmitted by a known transmitter.

SYN _r (n): the time-domain pilot separated from the signal received by the receiver according to the positioning information n ₁ (i) and n ₂ (i).

SYN_PN ₁ (n): the first PN period in the time-domain pilot SYN _r (n) received by the receiver.

SYN_PN _s (n): the S-th PN period in the time-domain pilot SYN _r (n) received by the receiver.

SYN _{r_oversample} (n): Obtained by oversampling SYN _r (n).

PN(n): a periodic PN sequence constituting the time-domain pilot SYN(n) known at the receiver.

pilot(n): a period of time-domain pilot determined by the sliding window interval (n∈[n _b (i), _ne (i)]) in the received time-domain pilot SYN _r (n).

pil _otoversample (n): obtained by oversampling pilot(n).

pn _c (n) _{: Pseudo} Random PN sequence. pn _c (n) is the result of the cyclic shift of PN(n).

pn _{c_oversample} (n): obtained by interpolating pn _c (n).

PILOT(k): is the frequency-domain counterpart of pilot(n).

PILOT _oversample (k): is the frequency-domain counterpart of pil _otoversample (n).

PN _c (k): is the frequency domain corresponding quantity of pn _c (n).

PN _{c_oversample} (k): is the frequency-domain counterpart of pn _{c_oversample} (n).

h _N (n): The estimate of the channel impulse response, the length is N.

h _{N_oversample} (n): The estimate of the channel impulse response under oversampling, the length is N×Fs.

h _M (n′): The estimate of the channel impulse response, the length is M, obtained from h _N (n) with zero padding.

h _{M_oversample} (n′): The estimate of the channel impulse response under oversampling, the length is M×Fs, obtained from h _{N_oversample} (n) with zero padding.

h _M+2×N (n′): The estimate of the channel impulse response, the length is M+2×N, obtained from h _N (n) with zero padding.

h _{M+2×N_oversample} (n′): The estimate of the channel impulse response under oversampling, the length is (M+2×N)×Fs, obtained from h _{N_oversample} (n) with zero padding.

H _N (k): The estimate of the channel frequency response, the length is N, and it is the frequency domain corresponding quantity of h _N (n).

H _M (k): The estimation of the channel frequency response, the length is M, which is the frequency domain corresponding quantity of h _M (n′).

H _M+2×N (k): the estimation of the channel frequency response, the length is M+2×N, which is the corresponding quantity in the frequency domain of h _M+2×N (n′).

H _{N_oversample} (k): The estimation of channel frequency response, the length is N×Fs, which is the frequency domain corresponding quantity of h _{N_oversample} (n).

H _{M_oversample} (k): The estimation of channel frequency response, the length is M×Fs, which is the frequency domain corresponding quantity of h _{M_oversample} (n′).

H _{M+2×N_oversample} (k): The estimation of channel frequency response, the length is (M+2×N)×Fs, which is the frequency domain corresponding quantity of h _{M+2×N_oversample} (n′).

recv(n): The signal received by the receiver, including time domain pilot and data block signal.

n ₁ (i): the start time of the time-domain pilot SYN _r (n) of the i-th frame in the data frame obtained from the signal recv(n) received by the receiver through synchronization processing.

n ₂ (i): the end time of the time-domain pilot SYN _r (n) of the i-th frame in the data frame obtained by synchronizing the signal recv(n) received by the receiver, that is, the start time of the data block.

N: The symbol length of one period of the PN sequence constituting the time-domain pilot.

M: The length of the data block.

L: the length of the time-domain pilot.

R1(τ): The result obtained by performing circular correlation on the first PN cycle in the time-domain pilot SYN _r (n).

R2(τ): The result obtained by performing circular correlation on the Sth PN cycle in the time-domain pilot SYN _r (n).

n _{b_initialization} (i): the beginning of the initial i-th frame sliding window.

n _{e_initialization} (i): the end of the initial i-th frame sliding window.

n _e (i): the end of the sliding window in the current i-th frame.

n _b (i): the beginning of the sliding window of the current i frame.

n _e ′(i): is an adjustment to n _e (i), defined as

n _e '(i)=min((n _e (i)-n ₁ (i))mod N+SRRC_Delay, N-SRRC_Delay)

Fs: system oversampling rate.

SRRC_Delay: Delay of the time response of the string pass filter at the transmitter and receiver.

S: Number of PN cycles in the time-domain pilot SYN _r (n) received by the receiver.

OFDM _M (K): an OFDM signal of M symbols obtained through demodulation.

x(n): the demodulated single-carrier signal of M symbols.

DATA _block (n): The data block mixed with single-carrier data and multi-carrier data obtained in method 2.

The operation steps in the algorithm flow chart of Fig. 6A are now described:

The first step is to obtain the time-domain pilot SYN _r (n) of the i-th frame in the received TDS-OFDM data frame by performing synchronous processing on the signal recv(n) received by the receiver, represented by SYNC_PROC() The start time n ₁ (i) and the end time n ₂ (i), n ₂ (i) are also the start times of the data block DATA _r (n).

The second step is to initialize the position of the sliding window. The length of the sliding window is equal to a cycle length N of the PN sequence, and the window interval of the sliding window of the i-th frame is initialized to be any j-th PN sequence cycle in the time-domain pilot in the i-th frame. , where 1<j<=S, S is the number of PN periods in the time-domain pilot SYN(n). The window starts at n _{b_init} (i) and ends at n _{e_init} (i). As shown in Figure 7.

The third step is to do correlation processing. Use PN(n) to perform circular correlation on the first PN period SYN_PN ₁ (n) in the time-domain pilot SYN _r (n) of the i-th frame to obtain R1(τ). The S-th PN period SYN_PN _s (n) in the time-domain pilot SYN _r (n) of the i frame is used for circular correlation to obtain R2(τ), where S is the number of PN periods in the time-domain pilot SYN _r (n), R1(τ) and R2(τ) are shown in Figure 8B, the lengths of R1(τ) and R2(τ) are less than N, where τ is a discrete time variable, which is used for the correlation function R1(τ) and R2(τ) , to distinguish it from n in order not to cause confusion.

The fourth step, a practical example shown in Figure 8A, h(n) is the impulse response of the channel, and the multipath component of h(n) is reflected in the form of some peaks in R1(τ) and R2(τ) out, and some noise is superimposed on R1(τ) and R2(τ), because there is always noise in the channel. The method to determine the multipath component is to compare the amplitude of R2(τ) and R1(τ) with a certain threshold after smoothing and filtering R2(τ) and R1(τ), if it is greater than the threshold, it is judged to be multipath Components smaller than the threshold are noise. The choice of the threshold can be determined by the different anti-noise and sensitivity to resolve multipath required by the application. After filtering and smoothing R2(τ) and R1(τ) respectively, set the time offsets of the detected multipath components as τ=τ _i , i=1, 2,..., Count, Count<N , k is the number of multipath components. Compare the amplitudes of the multipath components whose delays are τ _i in R2(τ) and R1(τ), i=1, 2,..., Count, and compare from the multipath component with the longest delay of τ _count , if R2(τ _count )＞R1(τ _count ), the initial position of the sliding window is correct; if R2(τ _count )＜R1(τ _count ), then it is judged that there is a side path before the main path, because the main path is The side path of the cause the pre-diffusion of the signal relative to the main path, the pre-diffusion of the second PN cycle in the time domain pilot is superimposed on the first PN cycle, and for the S-th PN cycle, since it is the last One PN period, there is no such forward diffusion superimposed on it, so the amplitude of R2(τ _count ) obtained by correlation is smaller than R1(τ _count ), so the initial position of the sliding window is incorrect, and it should slide forward , including the side paths preceding this main path. When moving to a certain τ _i , the sliding stops when R2(τ _i )>R1(τ _i ), the reason is that the side path behind the main path causes the signal to diffuse backward relative to the main path, and the first in the time domain pilot The post-diffusion of S-1 PN periods is superimposed on the S-th PN period, and for the first PN period, since it is the first PN period, there is no such back-diffusion superimposed on it , so the magnitude of R2(τ _i ) obtained through correlation is larger than that of R1(τ _i ). From the flow chart in Figure 6A, we can see that when τ _i =Multipath_Set(i), R2(τ _i )>R1 (τ _i ), when τ _i ′=Multipath_Set(i+1), there is R2(τ _i )<R1(τ _i ), from Figure 8B we can see that the position of the end of the sliding window can be determined by the difference between the two comparisons, Calculate n _{e_min} = n _{e_initialization} -N+Multipath_Set(i) and n _{e_max} = n _{e_initialization} -N+Multipath_Set(i+1), where n _{e_min} (i) and n _{e_max} (i) determine the correct positioning of n The interval of _e (i) [n _{e_min} , n _{e_max} ], the sliding window positioned like this includes the side path before the main path and the side path behind the main path; All other side path components are R2(τ _i )<R1(τ _i ), it is judged that there is no side path after the main path at this time, calculate n _{e_min} = n _{e_initialization} -N+Multipath_Set(1) and n _{e_max} = n _{e_initialization} -N+Multipath_Set(2), where n _{e_min} (i) and n _{e_max} (i) determine the interval [n _{e_min} , n _{e_max} ] for correctly positioning n _e (i), and the sliding window so positioned Including the side diameters before the main diameter, finally n _e (i) makes a relative displacement from the initial position, the window length remains unchanged, and the beginning of the positioning window is n _b (i)= _ne (i)-N. R1(τ) and R2(τ) are drawn together in FIG. 8B to illustrate the function of the sliding window vividly. The actual position of the sliding window is shown in FIG. 7 .

In the fifth step, as shown in Figure 6A, the channel estimation module uses the original signal known by the receiver to use the original signal pilot(n) within the interval determined by the sliding window in the time-domain pilot received through channel convolution. A section of signal pn _c (n) within the interval determined by the sliding window in the time-domain pilot after channel convolution is subjected to time-domain circular correlation or mathematically equivalent frequency-domain processing to obtain an estimate of the channel impulse response h _N (n ), or the estimated channel impulse response h _N (n) can be obtained from the obtained R1(τ) and R2(τ), as follows:

This process is shown in Figure 9A and Figure 9B. The signal of R1(τ) in the interval τ∈[(n _e (i)-n ₁ (i))mod N+1, N] is moved to h _N ( n) of n∈[(n _e (i)-n ₁ (i))mod N+1, N] in this interval, and then τ∈[1, (n _e (i)-n ₁ (i) )mod N], the signal of R2(τ) in this interval is moved to the interval of n∈[1, (n _e (i)-n ₁ (i))mod N] of h _N (n), and h _N (n). As shown in FIG. 9C , after h(n) is period-extended with N as the period, taking one period of this period signal is h _N (n).

Fill h _N (n) with zeros to obtain h _M (n′), and copy the formula in Flowchart 6A to the following:

This process is shown in Figure 10A and Figure 10B. First, h _M (n′) can be regarded as a function of length M before operation, and its values are all zero. In order to avoid confusion with n, define n′ as its own The variable represents the discrete time, and the main operation here is to move the signal of n∈[1, (n _e (i)-n ₁ (i))mod N] in the interval h _N (n) to h _M ( n′) of n′∈[1, (n _e (i)-n ₁ (i))mod N] in this interval, and then n∈[(n _e (i)-n ₁ (i))mod N+1, N] The signal of h _N (n) in this interval is moved to h _M (n′) n′∈[M-(N-(n _e (i)-n ₁ (i))mod N) +1, _M ] in this interval, and then in the remaining interval n′∈[(n _e (i)-n ₁ (i))mod N+1, M-(N-( n _e (i)-n ₁ (i))mod N)] to get the zero-filled result of h _N (n), and then do fast Fourier transform (FFT) to h _M (n′) to get the channel The frequency response estimate H _M (k); the purpose of zero padding is to obtain the frequency response estimate H _M (k) of point M from h _M (n′) through FFT, and the frequency domain data is also of point M, and the two are divided Achieve frequency domain equalization.

The sixth step is to construct the circular convolution feature module to construct the data block DATA transmitted by the source by using some addition and subtraction operations between the time-domain pilot signal SYN _r (n) transmitted through the channel and the data block DATA _r (n) (n) and the channel impulse response are circularly convoluted to obtain DATA _c (n), whose length is M. This process is shown in Figure 11. We copy the corresponding formula in Figure 6A to the following:

As shown in Figure 11A2, n∈[n ₂ (i)+1, n ₂ (i)+(n _e (i)-n ₁ (i))mod N-1] in the above formula means that n is n ₂ (i)+1, the end is n ₂ (i)+(n _e (i)-n ₁ (i))mod N-1 interval change, where n is a discrete time variable, DATA _c (n ) can be regarded as a function with a length of M before the operation, and can be realized with a storage space with a length of M in the implementation; in the first formula of the above formula, the argument of DATA _r (n) is n, and it When changing from n ₂ (i)+1 to n ₂ (i)+(n _e (i)-n ₁ (i))mod N-1, as shown in Figure 11A2, it shows that DATA _r (n) in this A segment of the signal in the interval; DATA _c (nn ₂ (i)) and DATA _c (n) represent the same function, only DATA _c (nn ₂ (i)) has made a time shift, when n changes from When n ₂ (i)+1 changes to n ₂ (i)+(n _e (i)-n ₁ (i))mod N-1, the argument of DATA _c (nn ₂ (i)) is nn ₂ ( i), it changes from 1 to (n _e (i)-n ₁ (i)) mod N-1, which represents a section of signal of DATA _c (n) in this interval, SYN _r (nN) and SYN _r (n +M) and SYN _r (n) represent the same signal, that is, the received time-domain pilot signal, SYN _r (n+M) and SYN _r (nN) just make a comparison with SYN _r (n) Time translation, when n changes from n ₂ (i)+1 to n ₂ (i)+(n _e (i)-n ₁ (i))mod N-1, SYN _r (n+M) The independent variable is n+M, which varies from n ₂ (i)+M+1 to n ₂ (i)+M+(n _e (i)-n ₁ (i))mod N-1, as shown in Figure 11A2 , which represents a section of signal of SYN _r (n) in this interval, the independent variable of SYN _r (nN) is nN, which changes from n ₂ (i)-N+1 to n ₂ (i)-N+(n _e ( i)-n ₁ (i))mod N-1, as shown in Figure 11A2, represents a section of signal of SYN _r (n) in this interval; according to the first operation relationship of the above three formulas, after addition and subtraction operation, add the back-diffused signal of the data block DATA _r (n) in the i-th frame relative to the main path back to its head, and subtract a part of the time-domain pilot signal superimposed due to the addition operation, only Data remaining. In the second formula of the above formula, the independent variable of DATA _r (n) is n, which is from n ₂ (i)+M-(N-(n _e (i)-n ₁ (i))mod N) When it changes to n ₂ (i)+M, as shown in Figure 11B2, it represents a section of signal of DATA _r (n) in this interval; DATA _c (nn ₂ (i)) and DATA _c (n) represent the same A function of DATA _c (nn ₂ (i)) is just a time translation, when n changes from n ₂ (i)+M-(N-(n _e (i)-n ₁ (i)) mod N) changes to n ₂ (i)+M, the independent variable of DATA _c (nn ₂ (i)) is nn ₂ (i), it changes from M-(N-(n _e (i)-n ₁ ( i))mod N) changes to M, indicating a section of signal of DATA _c (n) in this interval; SYN _r (nM) and SYN _r (nMN) and SYN _r (n) represent the same signal, that is, receiving The received time-domain pilot signals, SYN _r (nM) and SYN _r (nMN) are only shifted in time relative to SYN _r (n), when n changes from n ₂ (i)+M-(N-( When n _e (i)-n ₁ (i))mod N) changes to n ₂ (i)+M, the argument of SYN _r (nM) is nM, which changes from n ₂ (i)-(N-(n _e (i)-n ₁ (i)) mod N) changes to n ₂ (i), as shown in Figure 11B2, it represents a section of signal of SYN _r (n) in this interval, the independent variable of SYN _r (nMN) is nMN, which varies from n ₂ (i)-N-(N-(n _e (i)-n ₁ (i))mod N) to n ₂ (i)-N, as shown in Figure 11B2, which represents SYN _r (n) is a segment of the signal in this interval; according to the second operation relationship of the above three formulas, after addition and subtraction operations, the forward direction of the data block DATA _r (n) in the i-th frame relative to the main diameter The diffused signal is added back to its rear part, and at the same time, a part of the time-domain pilot signal superimposed due to the addition operation is subtracted, leaving only the data. In the third formula of the above formula, the independent variable of DATA _r (n) is n, which changes from n ₂ (i)+(n _e (i)-n ₁ (i))mod N to n ₂ (i )+M-(N-(n _e (i)-n ₁ (i))mod N)-1, as shown in Figure 11C2, represents a section of signal of DATA _r (n) in this interval; DATA _c ( nn ₂ (i)) and DATA _c (n) represent the same function, only DATA _c (nn ₂ (i)) has made a time translation, when n changes from n ₂ (i)+(n When _e (i)-n ₁ (i))mod N changes to n ₂ (i)+M-(N-(n _e (i)-n ₁ (i))mod N)-1, DATA _c (nn ₂ (i))'s argument is nn ₂ (i), which varies from (n _e (i)-n ₁ (i)) mod N to M-(N-(n _e (i)-n ₁ (i ))mod N)-1 represents a section of signal of DATA _c (n) in this interval, a signal transfer operation is performed in the third formula, and DATA _c (n) is finally formed.

In the seventh step, the FFT module performs fast Fourier transform (FFT) on DATA _c (n) to obtain frequency domain data Y(k) before frequency domain equalization.

In the eighth step, the frequency domain equalization module divides Y(k) by the estimated channel frequency response H _M (k) to obtain frequency domain data X(k) after frequency domain equalization.

In the ninth step, if the data block DATA(n) in the known transmitted signal is an OFDM signal, X(k) is output as the equalized data; if the known transmitted data block DATA(n) is a single carrier block signal, then do another M-point IFFT on X(k), and output the result as equalized data; if it is known that the transmitted data block DATA(n) is a combination of several OFDM block signals and several single-carrier block signals Combination, first perform an M-point IFFT on X(k), and position and process the results according to the positions and sizes of these OFDM and single-carrier block signals agreed by the transmitter and receiver. For OFDM data blocks, further processing is required. Perform an FFT to obtain the equalized data output, and the single carrier block signal is the equalized data that can be directly output.

Now the operation steps in the algorithm flow chart of Fig. 6B are described:

The first to fourth steps are exactly the same as the first to fourth steps in Fig. 6A.

The fifth step is to add (2×N) zeros in the middle of h _M (n′) obtained by filling zeros in the fifth step in the algorithm flow chart of Fig. 6A to obtain h _M+2×N (n′), and calculate H _M+2×N (k)=FFT(h _M+2×N (n′)), H _M+2×N (k) will be used for frequency domain equalization.

The sixth step, since the transmitted data block DATA(n) is actually linearly convolved with the impulse response of the channel after it is transmitted through the channel, but if the data block DATA(n) is combined with its previous cycle and the next cycle Periodic PN sequences are considered together, and they form a circular convolution relationship with the impulse response of the channel after passing through the channel; the data block DATA _r (n) after channel transmission and the PN sequence of the previous cycle and the next cycle are defined It is DATA _M+2×N (n), and its length is M+2×N. It can be processed directly. This method requires more calculation than the method in the algorithm flow chart in Figure 6A, but the method in the algorithm flow chart in Figure 6A uses some addition and subtraction operations, which will lead to a certain degree of amplification of the additive noise on the signal, which is harmful to the signal. The performance has a certain impact, and these two methods can be used according to the actual situation.

In the seventh step, the FFT module performs fast Fourier transform (FFT) on the DATA _M+2×N (n) to obtain frequency domain data Y _M+2×N (k) before frequency domain equalization.

In the eighth step, the frequency domain equalization module divides Y _M+2×N (k) by the estimate H _M+2×N (k) of the channel frequency response to obtain the frequency domain equalized data X _M+2×N (k ). Then perform an inverse fast Fourier transform (IFFT) on X _M+2×N (k) to obtain x _M+2×N (n), remove the PN sequence of the first N points of x _M+2×N (n) and The PN sequence of the last N points obtains the data block x _M (n).

In the ninth step, if the data block DATA(n) sent is known to be an OFDM signal, do M-point FFT with x _M (n) as the equalized data output; if the data block DATA(n) sent is known to be a single-carrier block signal, the obtained x _M (n) is output as equalized data; if it is known that the transmitted data block DATA(n) is a combination of several OFDM block signals and several single-carrier block signals, then the The obtained x _M (n), according to the position and size of these OFDM and single-carrier block signals agreed by the transmitter and receiver, respectively locate and process them, and perform FFT again for the OFDM data block to obtain the equalized data output, The single-carrier block signal means that the data after equalization can be output directly.

Now the operation steps in the algorithm flow chart of Fig. 6C are described:

The first to fourth steps are exactly the same as the first to fourth steps in Fig. 6A.

The fifth step, as shown in Figure 6C, using the oversampling method, the channel estimation module oversamples a section of the signal pilot(n) within the interval determined by the sliding window in the received channel convolution time domain pilot Obtain pilot _oversample (n), interpolate a segment of the signal pn _c (n) within the interval determined by the sliding window in the original time-domain pilot known to the receiver without channel convolution, and in each element Fs-1 zeros are then inserted, where Fs is the oversampling rate, resulting in pn _{c_oversample} (n). Perform time-domain cyclic correlation on pilot _oversample (n) and pn _{c_oversample} (n), or mathematically equivalent frequency-domain processing, to obtain an estimate of the channel impulse response h _{N_oversample} (n), and then zero-fill h _{N_oversample} (n), The method of zero padding is also a process similar to the signal shifting process of obtaining h _M (n′), only a little correction is made to the window end n _e (i), and the new window end is calculated as n _e ′(i)=min( (n _e (i)-n ₁ (i))mod N+SRRC_Delay, N-SRRC_Delay), this is because the channel estimate h _{N_oversample} (n) obtained after oversampling is the actual channel impulse response and the transmitter and receiver The result of the time response convolution of the band-pass filter at the end, the function of the band-pass filter at the transmitting end is to limit the frequency band of the transmitted signal, so as not to interfere with the signal of the adjacent frequency band, and the function of the band-pass filter at the receiving end is to suppress the adjacent frequency band Input to the receiver to generate noise, the time delay of the time response of the transmitter and receiver bandpass string filter is SRRC_Delay, and the zero padding operation for h _{N_oversample} (n) should ensure that the time response waveform of the bandpass string filter does not It is corrupted by inserting zeros, so _ne (i) at the end of the window needs to be corrected a little to get ne _′ (i). For the case where no oversampling is used, the obtained channel estimate h _N (n) is not affected by the time response of the band-pass filters at the transmitting end and the receiving end, so there is no need to consider modifying the end of the window. Finally, h _{M_oversample} (n') is obtained by padding with zeros, and fast Fourier transform (FFT) is performed on h _{M_oversample} (n') to obtain the channel frequency response estimate H _{M_oversample} (k). Here n' and n both represent discrete time variables, and n' is used only to prevent confusion with n in h _{N_oversample} (n).

The sixth step is to construct a circular convolution feature module using the oversampled result SYN _{r_oversample} (n) of the time-domain pilot signal SYN _r (n) after channel transmission and the oversampled result DATA _{r_oversample} of the data block DATA _r (n) (n) to construct the result of circular convolution of the data block DATA(n) transmitted by the source and the channel impulse response to obtain DATA _{c_oversample} (n), whose length is M×Fs, the process The principle is the same as the algorithm flow chart of the sixth step in Fig. 6A, except that due to oversampling, the time scale is enlarged by Fs times.

In the seventh step, the FFT module performs fast Fourier transform (FFT) on DATA _{c_oversample} (n) to obtain frequency domain data Y _oversample (k) before frequency domain equalization.

In the eighth step, the frequency domain equalization module divides Y _oversample (k) by the estimated channel frequency response H _{M_oversample} (k) to obtain frequency domain data X _oversample (k) after frequency domain equalization. As shown in Figure 12A, the X _oversample (k) obtained by the oversampling method has an expansion in the frequency domain compared with the frequency domain data obtained by the frequency domain equalization without the oversampling method, so the effective data at this time is It is the combination of the two pieces of data at the head and tail of X _oversample (k), as shown in Figure 12A and Figure 12B, after a signal transfer process, the two pieces of data at the head and tail of X _oversample (k′) are moved to X(k ), the frequency domain data X(k) after frequency domain equalization is obtained. Here k represents a discrete frequency variable, (in the above text, k generally represents a discrete frequency variable), k' also represents a discrete frequency variable like k, using k' is just to prevent the k from X(k) Confusion occurs.

In the ninth step, if the data block DATA(n) in the known transmitted signal is an OFDM signal, X(k) is output as the equalized data; if the known transmitted data block DATA(n) is a single carrier block signal, then do another M-point IFFT on X(k), and output the result as equalized data; if it is known that the transmitted data block DATA(n) is a combination of several OFDM block signals and several single-carrier block signals Combination, first perform an M-point IFFT on X(k), and position and process the results according to the positions and sizes of these OFDM and single-carrier block signals agreed by the transmitter and receiver. For OFDM data blocks, further processing is required. Do an FFT to get the equalized data output, and the single-carrier block signal is the equalized data that can be output directly.

Now the operation steps in the algorithm flow chart of Fig. 6D are described:

The first to fourth steps are exactly the same as the first to fourth steps in Fig. 6A.

In the fifth step, compared with the algorithm flow chart in Fig. 6C, only the method of zero padding is different. To obtain _{hM_oversample} (n') in the zero filling step of Fig. 6C, add (2×N)×Fs zeros to obtain hM _{+2×N_oversample} (n′), calculate H _{M+2×N_oversample} (k )=FFT(h _{M+2×N_oversample} (n′)), used for frequency domain equalization.

The sixth step is to oversample the DATA _M+2×N (n) obtained in the sixth step of FIG. 6C to obtain DATA _{M+2×N_oversample} (n), whose length is (M+2×N)×Fs.

In the seventh step, the FFT module performs fast Fourier transform (FFT) on DATA _{M+2×N_oversample} (n) to obtain frequency domain data Y _{M+2×N_oversample} (k) before frequency domain equalization.

In the eighth step, the frequency domain equalization module divides Y _{M+2×N_oversample} (k) by the estimated channel frequency response H _{M+2×N_oversample} (k) to obtain frequency domain data X _{M+2×N_oversample} after frequency domain equalization (k). The principle is the same as that in the algorithm flow chart in Fig. 6C. Compared with the frequency domain data after frequency domain equalization obtained by the oversampling method, the X _{M+2×N_oversample} (k) obtained by the oversampling method is compared with the frequency domain data obtained by the frequency domain equalization method. expansion, so the effective data at this time is the combination of the two pieces of data at the head and tail of X _{M+2×N_oversample} (k), and after a signal transfer process, the frequency domain data after frequency domain equalization X _{M+2× N} (k). Do IFFT to it to obtain x _M+2×N (n)=IFFT(X _M+2×N (k)), remove the PN sequence of the first N points and the rear N points of x _M+2×N (n) The PN sequence results in data block x _M (n). Here k represents a discrete frequency variable, (in the above text, k generally represents a discrete frequency variable), k' in X _{M+2×N_oversample} (k') also represents a discrete frequency variable like k, use k' Just to prevent confusion with k in X _M+2×N (k).

In the ninth step, if the data block DATA(n) sent is known to be an OFDM signal, do M-point FFT with x _M (n) as the equalized data output; if the data block DATA(n) sent is known to be a single-carrier block signal, the obtained x _M (n) is output as equalized data; if it is known that the transmitted data block DATA(n) is a combination of several OFDM block signals and several single-carrier block signals, then the The obtained x _M (n), according to the position and size of these OFDM and single-carrier block signals agreed by the transmitter and receiver, respectively locate and process them, and perform FFT again for the OFDM data block to obtain the equalized data output, The single-carrier block signal means that the data after equalization can be output directly.

Claims (8)

1. based on the channel estimating and the equalization methods to the block signal that contains pilot tone of sliding window, a kind of Frame that contains time domain pilot that contains the transmitter emission, its time domain pilot constitutes by continuous two or more cycles and by the pseudorandom PN sequence of transmitter and receiver agreement, it is characterized in that: when channel estimating, this method is included in one to the other footpath component before and after the component of main footpath and movably decides acquisition correctly to carry out the interval of the PN sequence of channel estimating with this in the sliding window, thereby makes the top n of sliding window _b(i) and terminal n _e(i) determined to obtain the interval of correct channel estimating; From then on obtaining length again is the estimation h of the channel impulse response of N _N(n), and then with window top n _b(i) and the terminal n of window _e(i) conduct is to above-mentioned h _N(n) carry out the locating information of zero padding computing, obtaining length is the estimation h of the channel impulse response of M _M(n ') or length are the estimation h of the channel impulse response of M+2 * N _{M+2 * N}(n '); Then the top n of window _b(i) and the terminal n of window _e(i) position is treated to data block DATA as signal and channel impulse response being configured to the data block DATAr (n) of the required locating information of circular convolution after the channel transmission _c(n); When the length of the one-period of PN sequence is N, time domain pilot SYN (n) length of emission is L (L=S * N), wherein n represents discrete time, S is the number in PN cycle among the known time domain pilot SYN (n), the data block of emission is DATA (n), when its length M was variable, then it contained successively and has the following steps:

(a) i frame time domain pilot SYN in the data flow that obtains receiving _r(n) time started n ₁(i) and i frame data piece DATA _rThe time n of beginning (n) ₂(i): the data flow that receives can be regarded time domain pilot SYN as _r(n) and data block DATA _r(n) stack, i frame time domain pilot SYN in the data flow that the process Synchronous Processing obtains receiving _r(n) time started n ₁(i) and i frame data piece DATA _rThe time n of beginning (n) ₂(i);

(b) sliding window initialization: use sliding window to decide the interval of the PN sequence that can obtain correct channel estimating, the length of sliding window equals the one-period length N of PN sequence, between initialized window region any j PN sequence period in the time domain pilot, 1＜j＜=S wherein, the top of i frame slip window is n _b(i)=n ₁(i)+L-(S-j+1) ^*N, end is n _e(i)=n ₁(i)+L-(S-j) ^*N, sliding window can slide in whole time domain pilot;

(c) determine sliding window top n _b(i), terminal n _e(i) position: to the time domain pilot SYN that receives _r(n) first PN cycle obtains R1 (τ) do circular correlation in, to time domain pilot SYN _r(n) S PN cycle obtains R2 (τ) do circular correlation in, to R2 (τ) and R1 (τ) do respectively filtering and level and smooth after, the amplitude that effective multipath component of identical time-delay is relatively arranged among R2 (τ) and the R1 (τ), begin comparison from the longest multipath component of time-delay, if R2 (τ) is less than the amplitude that effective multipath component of identical time-delay is arranged among the R1 (τ), then the initial position of the sliding window of definition is incorrect in (b), moves forward the terminal n of new sliding window _e(i), move to time-delay less than (n always _e(i)-n ₁(i)) amplitude of the multipath component among the R2 of mod N (τ) greater than or slide when approximating the amplitude of the multipath component that identical time-delay is arranged among the R1 (τ) and stop because the terminal n of window _e(i) move window top n _b(i) also do corresponding moving, keep length of window constant;

(d) use window top position n _b(i) and terminal position n _e(i) locating information is tried to achieve the estimation h of channel impulse response _N(n), again to above-mentioned h _N(n) carry out the zero padding processing and obtain the estimation h that length is the channel impulse response of M _M(n ') or length are the estimation h of the channel impulse response of M+2 * N _{M+2 * N}(n ');

(d.1) try to achieve the estimation h that length is the channel impulse response of N with any in following two kinds of methods _N(n);

(d.1.1) be defined in selected sliding window interval (n ∈ [n _b(i), n _e(i)]) one section time domain pilot receiving of inner receiver is pilot (n), gets among the time domain pilot SYN (n) of known transmitter emission by sliding window interval (n ∈ [n _b(i), n _e(i)]) the pseudorandom PN sequence of Jue Ding one-period length is pn _c(n), use pn _c(n) pilot (n) is done circular correlation and just can obtain the estimation h that length is the channel impulse response of N _N(n), (or adopt pn _cThe version pn of a circular shifting shift position (n) _N' (n) come pilot (n) is obtained h do circular correlation _N" (n), h _N" (n) just equal h _N(n) circular shifting shift position is with h _N" (n) just obtain h by opposite direction circular shifting shift position _N(n)); This is the time domain channel estimation approach, also has frequency domain channel estimation approach of equal value on the mathematics, and its process is: aforesaid pilot (n) is obtained PILOT (k) as FFT, to aforesaid pn _c(n) obtain PN as FFT _c(k), calculate PILOT (k) ÷ PN _c(k)=H _N(k), to length be the H of N again _N(k) make N point IFFT and also can obtain h _N(n);

(d.1.2) the estimation h that also can obtain channel impulse response by following formula from the R1 (τ) that obtained and R2 (τ) _N(n), as shown in the formula moving operation:

(1).h _N(n)＝R1(τ)，

τ ∈ [n wherein _e(i)-n ₁(i)) mod N+1, N], n ∈ [(n _e(i)-n ₁(i)) mod N+1, N];

(2).h _N(n)＝R2(τ)，

Wherein τ ∈ [1, (n _e(i)-n ₁(i)) mod N], and n ∈ [1, (n _e(i)-n ₁(i)) mod N];

(d.2) length that the method for using time domain or frequency domain is obtained is the h of N _N(n) carry out zero padding by following formula, obtain the h that length is M _M(n '), n are from 1 to N, and n ' is from 1 to M:

(1).h _M(n′)＝h _N(n)，

Wherein n ' ∈ [1, (n _e(i)-n ₁(i)) mod N], and n ∈ [1, (n _e(i)-n ₁(i)) mod N];

(2).h _M(n′)＝h _N(n)，

N ' ∈ [M-(N-(n wherein _e(i)-n ₁(i)) mod N)+1, M], n ∈ [(n _e(i)-n ₁(i)) mod N+1, N];

(3).h _M(n′)＝0，

N ' ∈ [(n wherein _e(i)-n ₁(i)) mod N+1, M-(N-(n _e(i)-n ₁(i)) mod N)];

Then to h _M(n ') obtains H as FFT _M(k), H _M(k) will be used for last frequency domain equalization;

The length that the method for using time domain or frequency domain is obtained is the h of N _N(n) carry out zero padding by following formula, obtain the h that length is M+2 * N _{M+2 * N}(n '), n are from 1 to N, and n ' is from 1 to M+2 * N:

(1).h _M+2×N(n′)＝h _N(n)，

Wherein n ' ∈ [1, (n _e(i)-n ₁(i)) mod N], and n ∈ [1, (n _e(i)-n ₁(i)) mod N];

(2).h _M+2×N(n′)＝h _N(n)，

N ' ∈ [M+2 * N-(N-(n wherein _e(i)-n ₁(i)) M+2 * N mod N)+1 ,] n ∈ [(n _e(i)-n ₁(i)) mod N+1, N];

(3).h _M+2×N(n′)＝0，

N ' ∈ [(n wherein _e(i)-n ₁(i)) mod N+1, M+2 * N-(N-(n _e(i)-n ₁(i)) mod N)];

Then to h _{M+2 * N}(n ') obtains H as FFT _{M+2 * N}(k), H _{M+2 * N}(k) will be used for last frequency domain equalization;

(e) according to above-mentioned time n ₁(i), n ₂(i) and the window's position n _b(i), n _e(i) data block that receives is handled, signal and channel impulse response are configured to the relation of circular convolution, so that making frequency domain equalization, next step offsets channel distortion, make the signal that receives the correct recovery transmitter emission of function: after the transmission of the data block DATA of transmission (n) channel, relation with the actual linear convolution of the impulse response of channel, offset the distortion of channel for ease of making frequency domain equalization, need to do following the processing, make the impulse response of data and channel constitute the relation of circular convolution; Obtaining n ₁(i), n ₂(i) and the window's position n _b(i) and n _e(i) after, with the data block DATA after the channel transmission _r(n) obtain DATA by steps of processing _c(n), its length is M:

(1).DATA _c(n-n ₂(i))＝DATA _r(n)+SYN _r(n+M)-SYN _r(n-N)，

N ∈ [n wherein ₂(i)+1, n ₂(i)+(n _e(i)-n ₁(i)) mod N-1];

(2).DATA _c(n-n ₂(i))＝DATA _r(n)+SYN _r(n-M)-SYN _r(n-M-N)，

N ∈ [n wherein ₂(i)+M-(N-(n _e(i)-n ₁(i)) n mod N), ₂(i)+M];

(3).DATA _c(n-n ₂(i))＝DATA _r(n)，

N ∈ [n wherein ₂(i)+(n _e(i)-n ₁(i)) mod N, n ₂(i)+M-(N-(n _e(i)-n ₁(i)) mod N)-1];

After data block DATA (n) the channel transmission that sends, relation with the actual linear convolution of the impulse response of channel, but the PN sequence of superimpose data piece DATA (n) and its previous cycle and back one-period considers that together they are through having constituted the relation of circular convolution with the impulse response of channel behind the channel; With the data block DATA after the channel transmission _r(n) and its previous cycle and the back one-period the PN sequence definition be DATA _{M+2 * N}(n), its length is M+2 * N, is used for next step processing;

(f) ask frequency domain signal X (k) behind the frequency domain equalization: the DATA to obtaining earlier by above-mentioned (e) step _c(n) make fast fourier transform (FFT) and obtain Y (k), use the estimation H of Y (k) again divided by channel frequency response _M(k), i.e. Y (k)/H _M(k)=and X (k), obtain the frequency domain signal X (k) behind the frequency domain equalization; The perhaps DATA that will obtain by above-mentioned (e) step _{M+2 * N}(n) make fast fourier transform (FFT) and obtain Y _{M+2 * N}(k), use Y again _{M+2 * N}(k) divided by the estimation H of the channel frequency response that obtains by above-mentioned (d) step _{M+2 * N}(k), i.e. Y _{M+2 * N}(k)/H _{M+2 * N}(k)=X _{M+2 * N}(k), obtain frequency domain signal X behind the frequency domain equalization _{M+2 * N}(k), again to X _{M+2 * N}(k) make anti-fast fourier transform (IFFT) and obtain x _{M+2 * N}(n), remove x _{M+2 * N}(n) the PN sequence that PN sequence that preceding N is ordered and back N are ordered obtains time-domain signal x _M(n), x _M(n) be the time domain form of frequency domain signal X (k).

2. channel estimating and equalization methods based on sliding window according to claim 1 to the block signal that contains pilot tone, it is characterized in that: the data block DATA (n) that described transmitter sends is anti-discrete fourier transform (IDFT) data block of an OFDM, then the X that obtains (k) is exported as the result after the equilibrium, perhaps the x that obtains _M(n) do output as a result of behind the M point fast discrete fourier transform (FFT).

3. channel estimating and equalization methods based on sliding window according to claim 1 to the block signal that contains pilot tone, it is characterized in that: the data block DATA (n) that described transmitter sends is the data block of a single-carrier modulated, then the X that obtains (k) is remake M point IFFT one time, the result output of the result who obtains after as equilibrium; Perhaps the x that obtains _M(n) do as the result and export.

4. channel estimating and equalization methods based on sliding window according to claim 1 to the block signal that contains pilot tone, it is characterized in that: the data block DATA (n) that described transmitter sends is the combination in any of the data block of several OFDM data blocks and several single-carrier modulated, then earlier the frequency domain signal X (k) that obtains is made an anti-fast fourier transform of M point (IFFT), obtain data block DATA _Block(n)=and IFFT (X (k)), the DATA here _Block(n) and x _M(n) be of equal value on mathematics, these OFDM that arrange in some way according to transmitter and receiver and single carrier block sub-block are at data block DATA again _Block(n) position in and its size to these data blocks location, are handled respectively, need remake consequential signal after a FFT obtains equilibrium for the OFDM data block, directly export for the single carrier block signal.

5. based on the channel estimating and the equalization methods to the block signal that contains pilot tone of sliding window, a kind of Frame that contains time domain pilot that contains the transmitter emission, its time domain pilot constitutes by continuous two or more cycles and by the pseudorandom PN sequence of transmitter and receiver agreement, it is characterized in that: when channel estimating, this method is included in one to the other footpath component before and after the component of main footpath and movably decides acquisition correctly to carry out the interval of the PN sequence of channel estimating with this in the sliding window, thereby makes the top n of sliding window _b(i) and terminal n _e(i) determined to obtain the interval of correct channel estimating; In order to improve the time domain resolution of channel estimating, remake the channel estimating of over-sampling after can in selected sliding window interval, making time domain oversampling, obtaining length is the estimation h of the channel impulse response of N * Fs _{N_oversample}(n), and then with window top n _b(i) and the terminal n of window _e(i) conduct is to above-mentioned h _{N_oversample}(n) carry out the locating information of zero padding computing, obtaining length is the estimation h of the channel impulse response of M * Fs _{M_oversample}(n ') or length are (the estimation h of the channel impulse response of M+2 * N) * Fs _{M+2 * N_oversample}(n '); Then the top n of window _b(i) and the terminal n of window _e(i) position is as signal and channel impulse response are configured to the required locating information of circular convolution channel transmission and the data block DATA after receiver is made time domain oversampling _{R_oversample}(n) be treated to data block DATA _{C_oversample}(n); When the length of the one-period of PN sequence is N, time domain pilot SYN (n) length of emission is L (L=S * N), wherein n represents discrete time, S is the number in PN cycle among the known time domain pilot SYN (n), the data block of emission is DATA (n), when its length M was variable, then it contained successively and has the following steps:

(a) the time started n of i frame time domain pilot SYNr (n) in the data flow that obtains receiving ₁(i) and i frame data piece DATA _rThe time n of beginning (n) ₂(i): the data flow that receives can be regarded time domain pilot SYN as _r(n) and data block DATA _r(n) stack, i frame time domain pilot SYN in the data flow that the process Synchronous Processing obtains receiving _r(n) time started n ₁(i) and i frame data piece DATA _rThe time n of beginning (n) ₂(i);

(b) sliding window initialization: use sliding window to decide the interval of the PN sequence that can obtain correct channel estimating, the length of sliding window equals the one-period length N of PN sequence, between initialized window region any j PN sequence period in the time domain pilot, 1＜j＜=S wherein, the top of i frame slip window is n _b(i)=n ₁(i)+L-(S-j+1) ^*N, end is n _e(i)=n ₁(i)+L-(S-j) ^*N, sliding window can slide in whole time domain pilot;

(c) determine sliding window top n _b(i), terminal n _e(i) position: to the time domain pilot SYN that receives _r(n) first PN cycle obtains R1 (τ) do circular correlation in, S PN cycle among the time domain pilot SYNr (n) obtained R2 (τ) do circular correlation, to R2 (τ) and R1 (τ) do respectively filtering and level and smooth after, the amplitude that effective multipath component of identical time-delay is relatively arranged among R2 (τ) and the R1 (τ), begin comparison from the longest multipath component of time-delay, if R2 (τ) is less than the amplitude that effective multipath component of identical time-delay is arranged among the R1 (τ), then the initial position of the sliding window of definition is incorrect in (b), moves forward the terminal n of new sliding window _e(i), move to time-delay less than (n always _e(i)-n ₁(i)) amplitude of the multipath component among the R2 of mod N (τ) greater than or slide when approximating the amplitude of the multipath component that identical time-delay is arranged among the R1 (τ) and stop because the terminal n of window _e(i) move window top n _b(i) also do corresponding moving, keep length of window constant;

(d) with window top position n _b(i) and terminal position n _e(i) locating information is tried to achieve the estimation h of the channel impulse response of over-sampling _{N_oversample}(n), again to above-mentioned h _{N_oversample}(n) carry out the zero padding processing and obtain the estimation h that length is the channel impulse response of M * Fs _{M_oversample}(n ') or length are (the estimation h of the channel impulse response of M+2 * N) * Fs _{M+2 * N_oversample}(n ');

(d.1) in order to improve the time domain resolution of channel estimating, remake the channel estimating of over-sampling after can making time domain oversampling in selected sliding window interval: establishing the over-sampling coefficient is Fs, if the time-delay of transmitter and receiver end band pass filter is SRRC_Delay, at selected sliding window interval (n ∈ [n _b(i), n _eWhat (i)]) inner receiver received is pilot through one section time domain pilot of over-sampling _Oversample(n), among the time domain pilot SYN (n) of known transmitter emission by sliding window interval (n ∈ [n _b(i), n _e(i)]) the pseudorandom PN sequence of Jue Ding one-period length is pn _c(n), it is made interpolation (promptly at pn with sampling coefficient Fs _c(n) insert Fs-1 zero after each element) obtain pn _{C_oversample}(n), then can from following method, choose any one kind of them:

Time domain approach is: use pn _{C_oversample}(n) to pilot _Oversample(n) obtain the estimation h that length is the channel impulse response of N * Fs do circular correlation _{N_oversample}(n) (also can adopt pn _{C_oversample}The version pn of a circular shifting shift position (n) _{C_oversample}(n) come pilot _Oversample(n) obtain h do circular correlation " _{N_oversample (n)}(n), h " _{N_oversample}(n) just equal h _{N_oversample}(n) circular shifting shift position is with h " _{N_oversample}(n) just obtain h by opposite direction circular shifting shift position _{N_oversample}(n));

Frequency domain method is: to pilot _Oversample(n) obtain PILOT as FFT _Oversample(k), to aforesaid pn _{C_oversample}(n) obtain PN as FFT _{C_oversample}(k), calculate PILOT _Oversample(k) ÷ PN _{C_oversample}(k)=H _{N_oversample}(k), to length be the H of N * Fs again _{N_oversample}(k) make N * Fs point IFFT and also can obtain h _{N_oversample}(n);

(d.2) obtain h _{N_oversample}(n) to do the zero padding operation after, at first need before the zero padding n _e(i) be adjusted into n by following formula _e' (i), be used for zero padding operation: n _e' (i)=min ((n _e(i)-n ₁(i)) mod N+SRRC_Delay N-SRRC_Delay), is the h of N * Fs to length afterwards _{N_oversample}(n) carry out zero padding, obtain the h that length is M * Fs _{M_oversample}(n '), n be from 1 to N * Fs, and n ' is from 1 to M * Fs, and zero padding is operating as:

(1).h _{M_oversample}(n′)＝h _{N_oversample}(n)，

Wherein n ' ∈ [1, n _e' (i) * Fs], n ∈ [1, n _e' (i) * Fs];

(2).h _{M_oversample}(n′)＝h _{N_oversample}(n)，

N ' ∈ [M * Fs-(N-n wherein _e' (i)) * Fs+1, M * Fs], n ∈ [n _e' (i) * and Fs+1, N * Fs];

(3).h _{M_oversample}(n′)＝0，

N ' ∈ [n wherein _e' (i) * and Fs+1, M * Fs-(N-n _e' (i)) * Fs];

Then to h _{M_oversample}(n ') obtains H as FFT _{M_oversample}(k), H _{M_oversample}(k) can be used for last frequency domain equalization; Perhaps:

Obtain h _{N_oversample}(n) to do the zero padding operation after, at first need before the zero padding n _e(i) be adjusted into n by following formula _e' (i), be used for zero padding operation: n _e' (i)=min ((n _e(i)-n ₁(i)) mod N+SRRC_Delay N-SRRC_Delay), is the h of N * Fs to length afterwards _{N_oversample}(n) carry out zero padding, obtain length and be (the h of M+2 * N) * Fs _{M+2 * N_oversample}(n '), n be from 1 to N * Fs, n ' from 1 to (M+2 * N) * Fs, zero padding is operating as:

(1).h _{M+2×N_oversample}(n′)＝h _{N_oversample}(n)，

Wherein n ' ∈ [1, n _e' (i) * Fs], n ∈ [1, n _e' (i) * Fs];

(2).h _{M+2×N_oversample}(n)＝h _{N_oversample}(n)，

N ' ∈ [(M+2 * N) * Fs-(N-n wherein _e' (i)) * Fs+1, (M+2 * N) * Fs], n ∈ [n _e' (i) * and Fs+1, N * Fs];

(3).h _{M+2×N_oversample}(n′)＝0，

N ' ∈ [n wherein _e' (i) * Fs+1, (M+2 * N) * Fs-(N-n _e' (i)) * Fs];

Then to h _{M+2 * N_oversample}(n ') obtains H as FFT _{M+2 * N_oversample}(k), H _{M+2 * N_oversample}(k) can be used for last frequency domain equalization;

(e) according to above-mentioned time n ₁(i), n ₂(i) and the window's position n _b(i), n _e(i) data block that receives is handled, signal and channel impulse response are configured to the relation of circular convolution, so that making frequency domain equalization, next step offsets channel distortion, make the signal that receives the correct recovery transmitter emission of function: for the situation that adopts over-sampling, receiver will be through the data block DATA after the Channel Transmission _r(n) obtain DATA as over-sampling _{R_oversample}(n), will be through the time domain pilot SYN after the Channel Transmission _r(n) obtain SYN as over-sampling _{R_oversample}(n), with DATA _{R_oversample}(n) obtain DATA by steps of processing _{C_oversample}(n), its length is M * Fs:

(1)DATA _{c_oversample}(n-n ₂(i)×Fs)＝DATA _{r_oversample}(n)+SYN _{r_oversample}(n+M×Fs)-SYN _{r_oversample}(n-N×Fs)

N ∈ [n wherein ₂(i) * and Fs+1, n ₂(i) * Fs+n _e' (i) * Fs-1];

(2).DATA _{c_oversample}(n-n ₂(i)×Fs)＝DATA _{r_oversample}(n)+SYN _{r_oversample}(n-M×Fs)-SYN _{r_oversample}(n-M×Fs-N×Fs)，

N ∈ [n wherein ₂(i) * Fs+M * Fs-(N-n _e' (i)) * Fs-Fs+1, n ₂(i) * Fs+M * Fs];

(3).DATA _{c_oversample}(n-n ₂(i)×Fs)＝DATA _{r_oversample}(n)，

N ∈ [n wherein ₂(i) * Fs+n _e' (i) * and Fs, n ₂(i) * Fs+M * Fs-(N-n _e' (i)) * Fs-Fs];

N wherein _e' (i)=min ((n _e(i)-n ₁(i)) mod N+SRRC_Delay, N-SRRC_Delay); Perhaps:

For the situation that adopts over-sampling, with the data block DATA after the channel transmission _r(n) and its previous cycle and the back one-period the PN sequence definition be DATA _{M+2 * N}(n), to DATA _{M+2 * N}(n) obtain DATA as over-sampling _{M+2 * N_oversample}(n), its length is that (M+2 * N) * Fs is used for next step processing;

(f) ask frequency domain signal X (k) behind the frequency domain equalization: use DATA earlier _c(n) over-sampling version d ATA _{C_oversample}(n) make fast fourier transform (FFT) and obtain Y _Oversample(k), use Y again _Oversample(k) divided by the estimation H of channel frequency response behind the over-sampling _{M_oversample}(K), i.e. Y _Oversample(k)/H _{M_oversample}(K)=X _Oversample(k), obtain frequency domain signal X (k) behind the frequency domain equalization by following formula:

(1)、X(k)＝X _oversample(k′)

Wherein, k ∈ [1, M ÷ 2], k ' ∈ [1, M ÷ 2]

(2)、X(k)＝X _oversample(k′)

Wherein, k ∈ [M ÷ 2+1, M], k ' ∈ [(Fs-1) * and M+M ÷ 2+1, Fs * M]

Perhaps:

Use DATA _{M+2 * N}(n) over-sampling version d ATA _{M+2 * N_oversample}(n) make fast fourier transform (FFT) and obtain Y _{M+2 * N_oversample}(k), use Y again _{M+2 * N_oversample}(k) divided by the estimation H of the channel frequency response behind the over-sampling _{M+2 * N_oversample}(K), i.e. Y _{M+2 * N_oversample}(k)/H _{M+2 * N_oversample}(K)=X _{M+2 * N_oversample}(k), obtain frequency domain signal X behind the frequency domain equalization by following formula _{M+2 * N}(k):

(1)、X _M+2×N(k)＝X _{M+2×N_oversample}(k′)

Wherein, k ∈ [1, M ÷ 2], k ' ∈ [1, M ÷ 2]

(2)、X _M+2×N(k)＝X _{M+2×N_oversample}(k′)

Wherein, k ∈ [M ÷ 2+1, M], k ' ∈ [(Fs-1) * and M+M ÷ 2+1, Fs * M]

To X _{M+2 * N}(k) make a M+2 * N point IFFT, obtain x _{M+2 * N}(n)=IFFT (X _{M+2 * N}(k)), remove x _{M+2 * N}(n) the PN sequence that PN sequence that preceding N is ordered and back N are ordered obtains x _M(n), x _M(n) be the time domain form of frequency domain signal X (k).

6. channel estimating and equalization methods based on sliding window according to claim 5 to the block signal that contains pilot tone, it is characterized in that: the data block DATA (n) that described transmitter sends is anti-discrete fourier transform (IDFT) data block of an OFDM, then the X that obtains (k) is exported as the result after the equilibrium, perhaps the x that obtains _M(n) do output as a result of behind the M point fast discrete fourier transform (FFT).

7. channel estimating and equalization methods based on sliding window according to claim 5 to the block signal that contains pilot tone, it is characterized in that: the data block DATA (n) that described transmitter sends is the data block of a single-carrier modulated, then the X that obtains (k) is remake M point IFFT one time, the result output of the result who obtains after as equilibrium; Perhaps the x that obtains _M(n) do as the result and export.

8. channel estimating and equalization methods based on sliding window according to claim 5 to the block signal that contains pilot tone, it is characterized in that: the data block DATA (n) that described transmitter sends is the combination in any of the data block of several OFDM data blocks and several single-carrier modulated, then earlier the frequency domain signal X (k) that obtains is made an anti-fast fourier transform of M point (IFFT), obtain data block DATA _Block(n)=and IFFT (X (k)), the DATA here _Block(n) and x _M(n) be of equal value on mathematics, these OFDM that arrange in some way according to transmitter and receiver and single carrier block sub-block are at data block DATA again _Block(n) position in and its size to these data blocks location, are handled respectively, need remake consequential signal after a FFT obtains equilibrium for the OFDM data block, directly export for the single carrier block signal.

Images