CN107070603B - Space-time block code system signal method of sending and receiving - Google Patents

Abstract

The invention belongs to fields of communication technology, more particularly to space-time block code system signal method of sending and receiving.The present invention is in order to solve the problems, such as that traditional STBC system based on Channel Detection needs long pilot tone to carry out channel estimation, the solution of the present invention directly carries out cluster to reception signal and carries out classification mark (Labeling) using label symbol (Labelled Symbols), and then recovers transmission symbol.The beneficial effects of the present invention are relative to traditional technology, method of the invention does not need channel estimation, and realizes MLD with lower complexity.

Description

Space-time block coding system signal transmission and reception method

technical field

The invention belongs to the technical field of communication, and in particular relates to a signal sending and receiving method of a space-time block coding system.

Background technique

STBC (Space-time Block Coding, space-time block coding) is a multi-antenna system that uses multiple transmitting antennas to improve data transmission reliability. STBC and its variants have been widely used in wireless communication systems, such as WiFi, LTE, etc. The traditional STBC system needs to send pilot signals (Pilot Symbols) through the transmitter to enable the receiver to perform channel estimation, so that the receiver can use channel information to perform decoupling maximum likelihood detection (MLD) to recover the transmitted symbols. Due to the existence of noise, channel information can be estimated more accurately only by using a longer pilot signal. However, even if the channel information is fully known, the complexity of ML detection is extremely high.

Contents of the invention

The object of the present invention is to solve the above problems and propose a method for sending and receiving signals in a space-time block coding system that does not require channel estimation at the receiving end.

Technical scheme of the present invention is as follows:

A space-time block coding system signal sending and receiving method, characterized in that:

At the transmitting end: the transmitter inserts M marker symbols before sending data information, M is the number of possible modulation symbol combinations, that is, the total number of possible types of symbol vectors sent, M is a natural number, and the content of the marker symbols is Known by the receiving end;

At the receiving end: cluster all the signals received within a period of time to form M categories through a clustering algorithm, label each category with the received signals corresponding to the obtained marker symbols, and perform data signal processing according to the clustering results and marker symbols The decision resumes sending symbols.

The general technical solution of the present invention, in order to solve the problem that the traditional STBC system based on channel detection requires long pilots to perform channel estimation, the solution of the present invention directly clusters the received signals and uses LabeledSymbols to carry out category labeling (Labeling ), and then recover the transmitted symbols.

The present invention also proposes another STBC system signal sending and receiving method, characterized in that,

At the transmitting end: before sending the data information, the transmitter inserts N _t <M mutually unrelated marker symbols, the contents of which are known to the receiving end;

At the receiving end: use the received signals corresponding to the N _t mutually uncorrelated marker symbol vectors sent by the transmitter to reconstruct a marker matrix composed of M marker symbols by linear combination, for the data signals received in a period and The marker matrix is clustered by a clustering algorithm to form M categories, the marker symbol matrix obtained by reconstruction is used to mark the transmission symbols corresponding to each category, and the data signal is judged according to the clustering result and the marker symbol matrix to recover the transmission symbols;

The specific method of reconstructing the marker matrix with the N _t mutually uncorrelated marker symbol vectors sent by the transmitter is:

Let the mark matrix formed by M mark symbols be:

Its rank is N _t , so only N _t orthogonal marker symbols need to be sent That is, the complete marker matrix can be reconstructed, that is, L=L _s ·A, where A _ij represents the weight coefficient of l _sj when reconstructing the marker symbol l _i .

The beneficial effect of the present invention is that, compared with the traditional technology, the method of the present invention does not require channel estimation, and realizes MLD with relatively low complexity.

Description of drawings

Fig. 1 shows the symbol detection method one that the present invention proposes;

Fig. 2 shows the second symbol detection method proposed by the present invention;

Figure 3 shows a traditional pilot design method;

Fig. 4 shows the marking symbol design method 1 of the present invention;

FIG. 5 shows the first labeling method after receiving signal clustering;

Fig. 6 shows the second method of marking symbol design of the present invention;

Fig. 7 shows the marking symbol design method three of the present invention;

Fig. 8 shows the clustering algorithm process of the present invention;

Fig. 9 shows the flow of the multi-view clustering algorithm proposed by the present invention;

Fig. 10 shows the performance comparison between the method proposed by the present invention and the traditional method.

Detailed ways

The technical solutions of the present invention will be described in detail below in conjunction with the accompanying drawings.

Assuming that the channel remains constant over a period of time, the signal transmitted during that period traverses the same channel on each receive antenna. Since the noise obeys the CSCG distribution, the signal received on the receiving antenna obeys the multi-dimensional CSCG distribution under the condition that the transmitted symbols are known. Considering all the signals received in this short period of time, which obey the Gaussian mixture model, the received signals corresponding to the same transmitted symbol can be clustered into the same category by clustering the Gaussian mixture model formed by the received signals. The correspondence between the label information and the clustering categories is used to indicate the corresponding transmission symbols of each category, and then the received signal is judged to recover the corresponding transmission symbols.

Taking the Alamouti coding system as an example, two transmitting antennas transmit a symbol group consisting of two symbols every two time slots. To send the nth symbol group, two antennas send [s ₁ (n), s ₂ (n)] ^T in the first time slot, where s _i (n), i=1, 2 are modulated A symbol on a constellation diagram. Both antennas transmit in the second slot Let the channel be h=[h ₁₁ ,h ₂₁ ], then the received signal is expressed as

where u _i (n), i=1, 2 is the noise that obeys the cyclic symmetric complex Gaussian (CSCG) distribution, namely u _i (n) and s(n) are independent of each other. make

Then (1) can be expressed as:

y(n)=H·s(n)+u(n). (3)

The MLD of the Alamouti coding system can be expressed as:

which is

Assuming that the channel remains constant over a period of time, the signal transmitted during that period traverses the same channel on each receive antenna. Let C _k indicate that the transmitted symbol is the kth modulation state combination, and there are M state combinations in total, that is, there are M types of transmitted signals. Since the noise obeys the CSCG distribution, so that in a given Under the condition of , the signal y ₁ (n) received by the receiving antenna in the first time slot obeys the CSCG distribution with mean h·C _k and variance σ ² , namely Similarly, it can be seen that the signal y ₂ (n) received by the receiving antenna in the second time slot obeys the CSCG distribution with mean h·C′ _k and variance σ ² , namely

Therefore, the distribution of all symbol groups received by the receiver in this period satisfies

Where π _k is the probability of occurrence of each C _k . Taking QPSK modulation as an example, C _k =[a+jb,p+jq] ^T , a={±1}, b={±1}, p={±1}, q={±1}, so M =4 ² . From (6), it can be observed that whether the received signal y _i (n) of each time slot or the received signal y (n) of each symbol group is a Gaussian mixture model composed of M Gaussian distribution components (Gaussian Mixture Model, GMM).

Therefore, the method of the present invention can cluster the received signals corresponding to the same transmitted symbol into the same category by clustering the received signals, and can recover the transmitted symbols corresponding to the received signals by using the correspondence between the marker symbols and the clustering categories. .

As shown in Figure 1, for the symbol detection method one that the present invention proposes, concrete steps are as follows:

1. Clustering all signals y(n) received within a period of time to form M categories;

2. Mark the sending symbols corresponding to each category through the receiving signal corresponding to the marking symbols sent by the sending end;

3. According to the clustering result and the corresponding relationship between the category and the transmitted symbol, the data signal is judged to recover the corresponding transmitted symbol.

Taking the Alamouti coding system and QPSK modulation as an example, the transmitter inserts M=4 ² marker symbols before sending data information, and each marker symbol is a possible combination of transmitted symbols. For example, if two antennas transmit 1+j and 1+j, after clustering, the received signal corresponding to this symbol will be in the same category as the marked symbol [1+j, 1+j] ^T , and then the received signal can be judged.

As shown in Figure 2, for the symbol detection method two that the present invention proposes of the present invention, concrete steps are as follows:

1. Use the N _t uncorrelated label symbol vectors sent by the transmitter to reconstruct the label matrix (Label Matrix) composed of M label symbols by linear combination

2. Clustering the data signals and marker matrices received within a period of time to form M categories,

3. Label the sent symbols corresponding to each category by reconstructing the marked symbol matrix,

4. According to the clustering result and the corresponding relationship between the category and the transmitted symbol, the data signal is judged to recover the corresponding transmitted symbol.

Taking the Alamouti coding system and QPSK modulation as an example, M marker symbols form a marker matrix L=[l ₁ ,l ₂ ,...,l _M ]∈C ^2×M , and its rank (Rank) is 2, so only It is necessary to send two uncorrelated marker symbols L _s =[l _s1 ,l' _s1 ]∈C ^2×2 to reconstruct the complete marker matrix, that is, L=L _s ·A, where A _ij represents the reconstructed marker The weight coefficient of l _sj when symbol l _i . Although the received signal y(n) corresponding to each group of transmitted symbols is composed of signals received in two time slots, for each pair of transmitted symbols [s ₁ ,s ₂ ] ^T , through the first time slot The signal received in the second time slot and the signal matrix reconstructed by it can recover the signal sent in the second time slot, so only two mutually uncorrelated marker symbols need to be sent.

The invention further proposes a marking design method with error correction capability. This method sends v repetitions of Nt marker symbols per epoch, and before marker reconstruction, the same marker symbol vectors are averaged to reduce noise interference.

The present invention takes the expectation maximization (EM) algorithm in Gaussian mixture model clustering method as an example to illustrate the main idea of the symbol detection system based on clustering. At the same time, the present invention takes the equal probability of modulation symbols and the invariance of noise variance as examples to illustrate the main idea of using the inherent characteristics of the communication system as a priori to accelerate the detection algorithm.

The likelihood function of the received signal is:

where ψ=[{π ₁ ,θ ₁ },{π ₂ ,θ ₂ },...,{π _M ,θ _M }], θ _k ={μ _k ,Σ _k }, because the communication system modulates symbols etc. probability, and the noise covariance matrix of multiple antennas is the same, so Ψ=[{π ₀ ,θ ₁ },{π ₀ ,θ ₂ },…,{π ₀ ,θ _M }],θ _k ={μ _k ,Σ ₀ },π ₀ =1/M. Because each data point y(n) must belong to a certain Gaussian component, the hidden variable z _n ∈{0,1} ^M is introduced, Where z _nk is the element m of z _n , that is, only one element of z _n is 1, and the rest are 0. The expectation of the joint distribution of the received signal and latent variables is:

Under the assumption that the parameter Ψ of the Gaussian mixture distribution is known, the posterior probability of y(n)=C _k is:

Under the assumption that γ _nk is known, by deriving (8) with respect to the mean and covariance matrix and taking 0, the update formula can be obtained:

and

By iterating formula (9) and formula (10) (11), the parameters of the Gaussian mixture distribution can be obtained and used for subsequent classification. If it is further considered that the noises of different time slots are independent of each other, that is, the noise covariance matrix is a diagonal matrix σ ² I, the update formula of the covariance matrix can be written as: Using this update formula can greatly increase the calculation speed.

The present invention further proposes a hierarchical clustering-based STBC system decoding method. Taking the Alamouti coding system and QPSK modulation as an example, the received signals will form 16 categories in total. All received signals are first clustered into 4 categories, and then each category is clustered separately. This method can greatly increase the decoding speed.

The present invention further proposes an STBC decoding system based on Co-regularized Multi-view Clustering. Taking the Alamouti coding system as an example, since the transmitted symbols corresponding to the received signals y ₁ (n) and y ₂ (n) of the two time slots corresponding to the same symbol group, the received signals of the two time slots in this period are respectively When doing clustering, the results obtained by the clustering should tend to be consistent. Taking the GMM calculated by EM as an example, the calculated (9) should be approximate. Let there be a total of L time slots, when clustering for the rth time slot, (7) is equivalent to minimizing the following formula:

in u _rnk is equal to γ _nk calculated in the rth time slot. The calculation result approximation of each time slot is constrained by the following formula:

make Further, it can be obtained that clustering all time slots at the same time is equivalent to the following formula:

The update rule can be obtained by derivation as:

and

Figure 3 shows the traditional pilot design method, where the gray antennas are antennas that transmit 0, and the black antennas are antennas that transmit 1. Each antenna transmits 1 once, and a total of N _t symbols need to be transmitted. Since a single pilot is affected by noise with a large variance, it needs to be repeatedly sent multiple times to estimate accurately.

Fig. 4 shows a marking symbol design method of the present invention. In this method, M marker symbols are inserted before sending data symbols, and each marker symbol vector is a possible sending symbol vector. Taking a 2×2 antenna array and QPSK modulation as an example, there are 4 ² possible transmit symbol vectors. Taking the transmitted data symbol as [1+j, 1+j] ^T as an example, since the data symbol and the marker symbol pass through the same channel during this period, all the transmitted symbols of [1+j, 1+j] ^T Data symbols and token symbols [1+j,1+j] ^T will be clustered into the same category. The relationship between the clustering category and the transmission symbol is established through the label symbol, and then all the data signals of this category are judged to be the same transmission symbol as the label symbol.

FIG. 5 shows the results of clustering by the method shown in FIG. 4 and the positions of the markers.

FIG. 6 shows another marker design method proposed by the present invention that only needs to send N _t marker symbols instead of M marker symbols in a time period. Taking 4×4 antenna array and 16QAM modulation as an example, M=16 ⁴ marker symbols form a marker matrix Its rank (Rank) is N _t = 4, so only N _t linearly independent marker symbols need to be sent Then the complete marker matrix can be reconstructed, that is, L=L _s ·A, where A _ij represents the weight coefficient of l _sj when l _i is reconstructed.

Fig. 7 shows another design method of a marker symbol with error correction function proposed by the present invention. Due to noise effects, markers may appear far from the mean of the corresponding class and affect performance. By repeating the same mark v times, a total of vN _t mark symbols need to be sent, thereby improving the anti-noise capability.

Fig. 8 shows the algorithm flow of the clustering algorithm used as an example in the present invention. First initialize the parameters of the Gaussian mixture distribution, and then iteratively calculate the posterior probability of y(n)=c _k

and Gaussian mixture distribution parameter updates, namely:

and

By iterating the formula above, the parameters of the Gaussian mixture distribution can be obtained and used for subsequent classification.

FIG. 9 shows the flow of the multi-view clustering algorithm applied to the STBC system proposed by the present invention. First, the parameters of the Gaussian mixture distribution are initialized, and then the parameters of the Gaussian component are updated for each time slot, and then the posterior probability is updated for each time slot.

Fig. 10 shows the performance comparison between the method proposed by the present invention and the traditional method. Wherein the STBC-GMM curve uses the third marker symbol design method and the corresponding clustering algorithm proposed by the present invention, the marker symbols are repeated twice, and the MLD-channel estimation curve is based on the MLD of channel estimation, the number of pilots and the STBC-GMM The number of marker symbols used is the same, and the MLD-known CSI curve is an MLD that fully knows the information information. After the SNR is higher than a certain threshold, the algorithm of the present invention can accurately know the channel information without knowing the channel information. MLD performance, better than MLD performance based on channel estimation.

The EM algorithm used in the present invention is not limited to the scenario of sending symbols with equal probability, and meanwhile, the variance of the noise can also change with time. In addition, the present invention is not limited to using the EM algorithm to obtain the parameters of the Gaussian mixture distribution. The hierarchical clustering STBC system proposed by the present invention is also not limited to the two-stage clustering method. The method proposed by the present invention can be easily extended to any number of transmitting antennas and receiving antennas, and is not limited to the Alamouti coding system.

Claims (1)

1. space-time block code system signal method of sending and receiving, it is characterised in that:

In transmitting terminal: transmitter is inserted into N before sending data information_t< M irrelevant label symbols, M is natural number, The content of the label symbol is known to receiving end；

In receiving end: the N sent using transmitting terminal_tThe corresponding reception signal of a irrelevant label symbol vector passes through linear Combined mode reconstructs the label matrix of M label symbol composition, to the data-signal and label square received in a period Battle array carries out cluster by clustering algorithm and forms M classification, marks the corresponding hair of each classification by reconstructing obtained label matrix Symbol is sent, symbol is sent it is believed that number making decisions recovery according to cluster result and label logm；

The N sent with transmitting terminal_tA irrelevant label symbol vector reconstruction goes out to mark matrix method particularly includes:

If the label matrix that M label symbol is formed are as follows:

Its order is N_t, therefore only need to send N_tA orthogonal label symbol vectorIt is i.e. restructural complete Whole label matrix, i.e. L=L_sA, wherein A_ijIt indicates to rebuild label symbol l_iWhen l_sjWeight coefficient.