WO2012041047A1 - Channel estimation method and base station - Google Patents

Abstract

A channel estimation method and a base station are disclosed in the invention, and the method includes that: the base station receives sounding signals transmitted on each carrier by a terminal, and performs the Least Squares channel estimation for each carrier according to the received sounding signals and the preset sounding signals which are corresponding to the terminal; the base station obtains correlation coefficients between the carriers according to the result of the Least Squares channel estimation; the base station obtains mean square time delays according to the correlation coefficients, quantizes the mean square time delays and the signal-to-noise ratios of the carriers, and looks up a filtering matrix corresponding to the quantization result in the corresponding relation of preset quantization values and filtering matrixes; the base station obtains the Minimum Mean Square Error channel estimation of each carrier according to the filtering matrix and the result of the Least Squares channel estimation; and the base station performs linear interpolation with the Minimum Mean Square Error channel estimation and obtains the channel estimation of each carrier on all frequency bands. The complexity of calculation is reduced with the invention.

Description

 The present invention relates to the field of communications, and in particular, to a channel estimation method and a base station. BACKGROUND OF THE INVENTION In many wireless communication technologies, Orthogonal Frequency Division Multiplexing (Orthogonal Frequency Division)

Multiplexing (referred to as OFDM) is one of the most promising technologies. In recent years, due to the rapid development of digital signal processing technology, OFDM has attracted widespread attention as a high-speed transmission technology with high spectrum utilization and good anti-multipath performance. OFDM technology has been successfully applied to Digital Audio Broadcasting (DAB), Digital Video Broadcasting (DVB), High Definition Television (HDTV), Wireless Office i or Wireless Local Area Network (WLAN) and Wireless Metropolitan Area Network (WMAN). Beamforming can be applied to OFDM systems to improve system performance. The use of this technique requires knowledge of the channel information. A mobile station (MS for short) can transmit a Sounding signal to a base station (BS), which enables the BS to know the channel response of the BS to the MS by using the nature of channel reciprocity. The algorithms for channel estimation of Sounding signals mainly include Least Squares (LS) channel estimation algorithm and Minimum Mean Square Error (MMSE) channel estimation algorithm. The LS channel estimation algorithm is simple to implement, but the estimation accuracy is not high, and it is susceptible to Gaussian noise, especially in the case of low signal-to-noise ratio. The MMSE channel estimation algorithm can achieve better performance. However, the inventors have found that the MMSE channel estimation method in the related art has a high computational complexity. SUMMARY OF THE INVENTION To solve the above technical problem, the present invention provides a channel estimation scheme to solve the problem of high computational complexity of MMSE channel estimation in the related art. In order to achieve the above object, according to an aspect of the present invention, a channel estimation method is provided, the method comprising: receiving, by a base station, a sounding signal transmitted by a terminal on each carrier, according to the received probe The measured signal and the preset detection signal corresponding to the terminal perform least square channel estimation on each carrier; the base station acquires a correlation coefficient between each carrier according to the result of the least square channel estimation; the base station obtains a mean square delay according to the correlation coefficient, and The mean square delay and the signal-to-noise ratio of the carrier are quantized, and the filter matrix corresponding to the quantized result is searched for in the correspondence between the preset quantized value and the filter matrix; the base station obtains the result of the filter matrix and the least square channel estimation Minimum mean square error channel estimation for each carrier; base station performs linear interpolation using minimum mean square error channel estimation to obtain channel estimates for each carrier on all frequency bands. The base station obtains the mean square delay according to the correlation coefficient. The base station uses R _m (Δ/) = ¹ - calculates p ^p 1 + 2πΔ/σ _τ Ζ to calculate the mean square delay, where J is the frequency interval of the adjacent carrier, Δ / is the interval of the number of carriers between the respective carriers, R^O/) is the correlation coefficient, and σ _τ is the mean square delay. Performing a least square channel estimation on each carrier according to the received sounding signal and a preset sounding signal corresponding to the terminal includes: the base station performs a least square channel estimation using a formula ^& 0) = ( ^8 ^^??, where A ^fa (m) is the result of the least square channel estimation, which is the index of the carrier carrying the sounding signal, R = [r(0), r(l), ..., r(Jl) is the received sounding signal, = [ 6(0)^(1),... (J_1)] ^T is the preset detection signal corresponding to the terminal, where, for the detection sequence loop mode, L = P, where P is the preset value, for the detection sequence The sampling mode, =1. The base station obtains the correlation coefficient between each carrier according to the result of the least square channel estimation: the base station uses the public A _pi ) = E _m { ^ls ^τήή ¹ m + to calculate the estimated interval Δ/bearing The correlation coefficient of the carrier of the sounding signal; if Δ/ is zero, MR _pp (0) = R _pp (Al) - o _n ² ; if Δ / is not zero, then ^^(Δ/) = ^(Δ/) wherein, R ^ (a /) [Delta] represents the actual correlation coefficients spaced carriers / bearers of a detection signal, averaging £ _m represents for each carrier, ^ (Δ /) denotes estimated to be Δ apart correlation / bearers of a carrier detection _signal, σ η ² is the noise carrier. The base station obtains the minimum mean square error channel estimation of each carrier according to the result of the filter matrix and the least square channel estimation: the base station calculates the minimum mean square error of each carrier using the formula H _m = WH _k Channel estimation, where is a filter matrix; channel response of each carrier carrying the sounding signal obtained by estimating the minimum mean square error channel, is a vector of the element, _& is a carrier carrying the sounding signal estimated by the least square channel Channel response, fy _ls is

A ^fa O) is the vector of the element. For the detection sequence cyclic mode, the base station performs linear interpolation using the minimum mean square error channel estimation, and obtaining channel estimates for each carrier on all frequency bands includes: calculating channel estimates A ^fe of each carrier on all frequency bands according to the following formula:

P-l P-l

H' ^e (w) = (\--)H' ^m (bP + ) +—H' ^m ((b + \)P +

 2 2 where ^ is the position of the carrier and P is the carrier spacing of the cyclic mode of the detection signal.

 P-l

An integer satisfying the condition bP<w<(b + l)P, where c is an integer satisfying the condition c = w_bP—

 2

P-l

H ^lm (bP + ) is the minimum mean square error channel estimate;

2 For the detection sequence sampling mode, the base station performs linear interpolation using the minimum mean square error channel estimation, and obtaining the channel estimation of each carrier on all frequency bands includes: calculating the channel estimation A ^fe 0 of each carrier in all frequency bands according to the following formula:

H ^le (w) = (1 - ^-) H ^lm (dD + g) + -H ^l '"((d + \)D + g)

DD where ^ is the position of the carrier, is the carrier spacing of the mode in which the detection signal is carried, d satisfies the integer of the condition ^ D<w<(i/ + 1)D, and e is an integer satisfying the condition e = w_iffi>_g, For the actual sample offset, + is the minimum mean square error channel estimate. Quantifying the signal-to-noise ratio of the mean squared delay and the carrier includes: Using the following formula to obtain the vector quantization proximity condition: ∑ f ⁺¹ (d V( _Xi , yX W( , )) -d(W( _Xi , yX W(C _M ,)))^ (y) dy = 0 wide (d(W(x, _yj W( C _U ))—cW(x, _yj ), W(C. _j+l ))) _Ps (x)p _R (, )dx = 0 where is the quantization order of the signal-to-noise ratio, (X) is the letter The probability distribution of the noise ratio, the quantization interval of the signal-to-noise ratio is {[x _M , x _; ), l </<3⁄4}, which is the quantization order of the mean square delay, and p _R (y) is the mean square delay The probability distribution, the quantized interval of the mean square delay is {Ly^^ l ^^}, (^, . _; . is the quantized sample when the sum falls into [X,.-, , ) and D , , y _j ) respectively. Point, which is the filter matrix, and d is the quantization distortion of W; quantizes the mean square delay and the signal-to-noise ratio of the carrier according to the vector quantization neighboring condition and the centroid condition. In order to achieve the above object, according to another aspect of the present invention, a base station is provided, the base station includes: a receiving module, configured to receive a sounding signal sent by a terminal on each carrier; and a first estimating module configured to receive according to the received The detection signal and the preset detection signal corresponding to the terminal perform least square channel estimation on each carrier; the first obtaining module is configured to acquire correlation coefficients between the respective carriers according to the result of the least square channel estimation; the second acquiring module, setting To obtain a mean square delay according to the correlation coefficient, and quantize the mean square delay and the signal-to-noise ratio of the carrier, and search for a filter matrix corresponding to the quantized result in a correspondence between the preset quantized value and the filter matrix; An estimation module, configured to obtain a minimum mean square error channel estimate of each carrier according to a result of the filter matrix and the least square channel estimation; a third estimation module configured to perform linear interpolation using a minimum mean square error channel estimate to obtain each carrier on all frequency bands Channel estimation. The second obtaining module is configured to calculate a mean square delay using = ¹ -, where p ^p 1 + 2πΔ/σ _τ Ζ is the frequency interval of the adjacent carriers, and Δ/ is the interval of the number of carriers between the carriers , R _pp (M) is the number of relations, and σ _τ is the mean square delay. The first estimation module is configured to perform least square channel estimation using the formula » ^ι » = (B ^H B y ^l B ^H R , where A ^fa (m) is the result of the least square channel estimation, and is a carrier carrying the sounding signal Index; R = [r(0), r(l), ..., r(Jl) is the received probe signal; B = [b(0Xb(ll..., b(L - l)] ^T is the preset detection signal corresponding to the terminal, where, for the detection sequence loop mode, L = P, where P is the preset value, For the detection sequence sample mode, = 1. By the present invention, the corresponding relationship between the quantization parameter of the parameter and the filter matrix is pre-stored, and the filter matrix is obtained by finding the correspondence relationship, thereby solving the computational complexity of the MMSE channel estimation in the related art. The high-level problem, in turn, achieves the effect of reducing the computational complexity. The accompanying drawings, which are used to provide a further understanding of the invention, constitute a part of this application, the illustrative embodiments of the invention The invention is not to be construed as limiting the invention. In the drawings: FIG. 1 is a flowchart of a channel estimation method according to an embodiment of the present invention; FIG. 2 is a preferred flow of an MMSE channel estimation method according to an embodiment of the present invention; Figure 3 is a block diagram showing the structure of a base station according to an embodiment of the present invention; Figure 4 is a block diagram showing the structure of an MMSE channel estimation apparatus based on feature parameter quantization according to an embodiment of the present invention. The present invention will be described in detail below with reference to the accompanying drawings in conjunction with the embodiments. It is to be noted that the features of the embodiments and the embodiments of the present application may be combined with each other without conflict. Embodiments of the present invention provide a channel estimation method, which is based on MMSE channel estimation of a Sounding signal, and can be applied to an 802.16e system. FIG. 1 is a flowchart of a channel estimation method according to an embodiment of the present invention, where the method includes Step S102: The base station receives the sounding signal sent by the terminal on each carrier, and performs least square channel estimation on each carrier according to the received sounding signal and the preset sounding signal corresponding to the terminal; Step S104: The base station acquires correlation coefficients between the respective carriers according to the result of the least square channel estimation. Step S106: The base station obtains a mean square delay according to the correlation coefficient, and quantizes the mean square delay and the signal to noise ratio of the carrier. Searching for a filter matrix corresponding to the quantized result in a corresponding relationship between the preset quantized value and the filter matrix; for example, a table corresponding to the filter matrix and the quantized value is stored in the base station in advance, and the filter corresponding to the quantized result is obtained by looking up the table a matrix; step S108, the base station obtains a minimum mean square error channel estimation of each carrier according to a result of the filter matrix and the least square channel estimation; and step S110, the base station performs linear interpolation using a minimum mean square error channel estimation to obtain each carrier in all frequency bands. Channel estimation. In this embodiment, the corresponding relationship between the parameter quantization value and the filter matrix is stored in advance, and the filter matrix is obtained by finding the correspondence relationship. Since the filter matrix is calculated in advance, the embodiment reduces the calculation when performing channel estimation. The complexity. Step 4 S S 106 can also use the following implementation: The base station obtains the filter matrix using the following formula: W = R _pp (R _pp + σ) (RR ^H )~ ^ι Γ ¹ , where R _pp is the correlation coefficient obtained above The composition of the autocorrelation matrix, the composition method is the prior art, and will not be repeated here, σ „ ² is the noise of each carrier, R is the received data. ^ is the R transpose conjugate of the matrix, RR^ ^-1 is The inverse matrix of RR ^ff . Also, in step 4 S S 106, the mean square delay can be calculated by the following method: The base station calculates the mean square delay using the formula R _m (M) = ¹ - where J is the adjacent carrier Frequency interval, Δ/ is p ^p 1 + 2πΔ/σ _τ Ζ The interval between the number of carriers between each carrier, R^ (A/;> is the correlation coefficient, and σ _τ is the mean square delay. jt匕, in In step 4 S S 106, the signal-to-noise ratio of the mean square delay and the carrier can be quantized by: Using the following formula to obtain a vector quantization neighbor condition:

∑ f ⁺¹ (d(W( _Xl , y), W(C _t ,)) -d(W( _Xl , y), W(C _M j)))p _s {x _t )p _R (y) d = 0 3⁄4-l

Wide (d(W(x, _yj W(C _U ))—cW(x, _yj ), W(C. _j+l ))) _Ps (x)p _R ( , )dx = 0 where is signal noise The quantization order of the ratio, (X) is the probability distribution of the signal-to-noise ratio, and the quantization interval of the signal-to-noise ratio is {[x _M , x _; ), l </<3⁄4}, which is the quantization order of the mean square delay, p _R (y) is the probability distribution of the mean square delay, and the quantization interval of the mean square delay is {[ j ≤ n _R ], (^, . _; . is and falls into [x _M , x,.) and The quantized samples at [y, ^.) are the filter matrix, the quantization distortion of the matrix; subsequently, the mean square delay and the signal-to-noise ratio of the carrier are quantized according to the obtained vector quantization neighboring condition and the centroid condition. Preferably, step 4 S102 can be implemented in the following manner: The base station uses the formula ^O CB^B ^R to perform least square channel estimation, where A ^fa (m) is the result of the least square channel estimation, and is the carrier carrying the sounding signal. Index; R = [r(0), l),..., - is the received probe signal; B = [b(0\ b(\), ..., b(L- l) is pre- The detection signal corresponding to the terminal, wherein, for the detection sequence Cyclic mode, J is a preset value, for example, L = P, P is the value of the thousand specified in the protocol; for the detection sequence sample (Decimation) ), =1 where, step 4 S410 can be implemented by: The base station calculates the correlation coefficient of the estimated carrier of the separated Δ/carrier detection signal using the formula = ^{ ^il ^ + A/); /zero, then R (0) = „(A/)- σ„ ² ; if Δ/ is not zero, then

R _pp {M) = R _pp {M), where R^(A/) represents the actual correlation coefficient of the carrier separated by Δ/bearing detection signals, and £ _m represents the average of each carrier A, ^(Δ /) indicates the estimated correlation coefficient of the carrier separated by Δ/bearing detection signals, σ ² is the noise of each carrier. In the subsequent steps, channel estimation is performed using R _pp (Δ/). Preferably, step S108 This can be achieved by the base station calculating the minimum mean square error channel estimate for each carrier using the formula H _m =w _fa , where is the filter matrix; H _m is the most The channel response of each carrier carrying the sounding signal obtained by the small mean square error channel estimation, H _m is

The vector H _{fa where} H ^to ( ) is an element is the channel response of each carrier carrying the sounding signal estimated by the least square channel, and A _fa is a vector of the element, where m is the position of each element in the vector. For the detection sequence cyclic mode, each carrier on all frequency bands can be calculated according to the following formula

Channel estimation of Pl Pl fl ^le (w): H ^le (w) = (1 -—) H ^lm (bP + ) +—H' ^m ((b + \)P + )

 2 2 where ^ is the position of the carrier, P is the carrier spacing carrying the detection signal cyclic mode, 6 is satisfied

 P-1

Condition bP<w<(b + V)P an integer, c is an integer satisfying the condition c = w-bP-

 2

P-l

H' ^m (bP + ) is the minimum mean square error channel estimate. For the detection sequence sampling method, the channel estimation ^fe O) of each carrier in all frequency bands can be calculated according to the following formula: H' ^e (w) = (1 - ^)H ^lm (dD + g) + -^H ^lm ( (d + \)D + g) , where ^w is the position of the carrier, "the carrier spacing for the mode of carrying the sounding signal, for the integer satisfying the condition of fifi><w<(i/ + l)D, e is An integer satisfying the condition e = w-iffi>-g is the actual sample offset, and + is the minimum mean square error channel estimate. Embodiment 2 FIG. 2 is a preferred flowchart of the MMSE channel estimation method according to the embodiment of the present invention. As shown in FIG. 2, the MMSE channel estimation based on the sounding signal according to the embodiment of the present invention mainly includes the following steps: Step 201: Perform LS channel estimation on each carrier that carries the Sounding signal according to the Sounding signal and the received signal. The signal is a signal sent by the terminal to the base station, and the sounding signal corresponding to the terminal is pre-stored in the base station, and the received signal is a Sounding signal transmitted through the channel. During the transmission, the Sounding signal may change, and the received signal may be sent by the terminal. letter Not the same. Specifically, the sent Sounding sequence Decimation method is orthogonal by Frequency Division Multiplexing (FDM), and the sent Sounding sequence 'J Cyclic mode is orthogonal to the Code Division Multiplexing (CDM) method. Using the orthogonal properties of the Sounding sequence, the LS channel estimation can be performed by the following formula: H ^ls (m) = (B ^H B) - B ^H R (1) where 釆 is used for LS channel estimation; w Indicates the index carrying the Sounding signal carrier; R = [r(0), r(l),..., r(Z - l) is the vector of the received data; = [6(0), 6(1) ,..., 6 ( - 1)] ^τ is the Sounding sequence sent by the user. For the Sounding sequence Cyclic mode = , for the Sounding sequence Decimation mode = 1. Step 203: Calculate a correlation coefficient between carriers carrying the So Ding signal by using the LS channel estimation obtained in the previous step. Specifically, the channel response of the 7-bearing Sounding signal carrier can be obtained by the LS channel estimation algorithm in the previous step, and the correlation coefficient between the carriers carrying the Sounding signal can be estimated by the following formula: R _ρρ (Δ/) = E _m [ H ^1s (m)H ^LS (m + Δ/)} (2) where ^^Δ/) represents the estimated correlation coefficient of the Δ/carriers carrying the Sounding signal; fl ^ls (m) represents the LS channel Estimating the obtained channel response carrying the Sounding signal carrier; £ _m means averaging each carrier w. The result obtained by LS channel estimation can be expressed as follows: H ^k (m) = H(m) + N(m) (3) where HO) is the true channel response, Ν(ηή is noise. If the sounding signal carrier is assumed to be carried The channel response and noise are independent of each other, then the correlation coefficient in (2) can be expressed as follows: R (M) = R (M) + R _Z (M) where ^(Δ/) is the correlation coefficient between the estimated carriers carrying the ding signal, which is the correlation coefficient between the true bearer signal carriers. , R _Z (A/) is the correlation coefficient between noises. If the noise on the different carriers carrying the Sounding signal is assumed to be uncorrelated, then it can be known that 3⁄4( /) is an i to function. Therefore, the formula (4) can be expressed as follows:

_ R _PP (0) + & _n Δ/ = 0

 (5)

– Δ/≠0 where σ „ ² is the noise, ie the value of R _z (0). Step 205: Estimate the parameters of the channel at this time based on the correlation coefficient of the carrier carrying the Sounding signal: Mean Square Delay. The two characteristic parameters of delay and signal-to-noise are quantized, and the filter matrix is obtained by looking up the table. Specifically, in the following derivation process, a main assumption is made: Multipath power attenuation obeys a negative exponential distribution. It can be obtained by simulation and actual measurement. Assuming that the multipath power attenuation is distributed from a negative exponent, the power delay distribution can be approximated by equation (6):

其中 σ_τ为均方时延。 对 (^τ)进行 Fourier变换， 可以得到信道频域相关 函数：

Where σ _τ is the mean square delay. By performing a Fourier transform on ( ^τ ), the channel frequency domain correlation function can be obtained:

那么两个相隔 个载波的相关系数为： R _H (△/) =丄Γ ^ex (-―) exp(- _/'2πΔ/τ)ί/τ (7) By arranging the above results, you can get:

Then the correlation coefficients of two separated carriers are:

R _g (A/)= (9)

1 + j2 Alo _T L where is the frequency spacing of adjacent carriers. According to equations (5) and (9), the mean square delay σ _τ can be obtained. It can be seen that the filter matrix is completely determined by the following two parameters: mean square delay σ _τ , signal power ratio noise power (ie, signal to noise ratio)

设/)对 χ,.的偏导为 0, 可得下式： ( RR ^H lo ² _n ) „ If these two parameters are quantized, the filter matrix can be calculated in advance. When performing MMSE channel estimation based on sounding signals, the matrix can be obtained by looking up the table to avoid real-time matrix inversion and matrix. The multiplication operation greatly reduces the operation. The following describes the process of parameter quantization. Let the quantization order of the signal ratio noise snr be n _s , and the 4 rate distribution be p _s (x , the quantization interval is {[x _M ,x,),l </<3⁄4}„ Let the quantized order of the mean square delay σ _τ be n _R , the probability distribution be p _R (y) , and the quantization interval be “[[ -! , ), 1 }. If the sum falls into [ , 3⁄4 ) and [; , , ) respectively, the quantized sample is = ς .. In the above case, the distortion of the overall quantization can be expressed as follows: F Use the Frobenius norm to measure the quantization distortion of the matrix .

Let /) for χ, the partial guide of 0, can get the following formula:

^- =∑ ί ⁺¹ (diwix^yl w(C..)) -diwix^yl w(C _i+lJ )))p _s (x _t )p _R (y)d (H)

=0 Set /) to, the partial derivative of . is 0, you can get the following formula: (U r 3⁄4+l

— = 2^J ( ^d ( ^w ( ^x ' yj )' ^w( - ^c ij )) ~ ^d ( ^w ( ^x ' yj )' ^w ( ^c , j+i )))3⁄4 ( ^χ )3⁄4 (i2) c(y

=0 A new vector quantization proximity condition can be obtained by solving the nonlinear equations (11) and (12). According to this new vector quantization proximity condition and centroid condition (select a point in the quantization interval [x^xJUD,,^.), the average distortion of the vector falling within the above quantization interval to be the smallest), The LBG ( Linde-Buzo-Gray ) algorithm can obtain the optimal parameter quantization of mean square delay and signal to noise ratio. Unconstrained optimal parameter quantization has no limit on the quantization order of each parameter. In the case where the storage space is limited, it is necessary to limit the quantization order of each parameter to meet the limitation of the storage space. At the same time, it is necessary to minimize the quantization distortion to improve system performance. The above problem can be described as follows: min D (13) subject to n _R n _s ≤ M (14) where M represents the limitation of the storage space, and the minimum value of D is obtained under the condition of the formula (14). If each set of quantized orders satisfying the constraint (14) is solved using the method described above, the computational complexity is O(M ³ ). Thus, when the value of the storage space M is large, the amount of calculation is large. The calculation complexity of the steepest descent method described below is O(3M), which can reduce the amount of calculation to a large extent. In order to better describe the algorithm, some variables are set below. Let /) ( ) = indicate the distortion caused by the quantization order being ( , ). M(P) = Tr _s n _R represents the storage space required for the quantization order. Where / is a strictly decreasing function and M(n) is a strictly increasing function. Let ^ denote a vector of 1x2, where the _/th element is 1, and the remaining elements are 0. Then the gradient in the direction of the jth element is: The specific steps of the steepest descent method are as follows: 1 Initial ^^ setting = (1, 1).

2 When M (when the constraint (14) is not satisfied, the value returns to the previous cycle and terminates. 3 is the maximum value. (Index, set + jump to step 2. Since D(P) is a A strictly decreasing function, while each component of the quantization order Ή is strictly increased in each cycle, so the final result should be global optimal rather than local optimal. Step 207: Calculate the LS The channel estimation and filtering matrix calculates the MMSE channel estimation of each carrier carrying the Sounding signal. The MMSE channel estimation based on the sounding signal can be applied as follows:

H _mmse = WH _ls (16) where = (R _pp + _σ (RR ^H , where is the filter matrix; is the channel response of each carrier carrying the Sounding signal obtained through the MMSE channel estimation, where is the vector of the element, /m Indicates that the MMSE channel estimation is used; A _fa is the channel response of each carrier carrying the Sounding signal estimated by the LS channel, where A _fa is a vector with ¹ » as an element, indicating that the LS channel estimation is used. „ ² is the noise, and R is the received sequence. Step 209: Finally, according to the MMSE channel estimation, linear interpolation is performed to obtain the channel estimation of the carriers on all frequency bands. Specifically, if the correlation between the carriers performing linear interpolation is large, then Linear interpolation will give better performance; if the correlation between carriers with linear interpolation is small, then linear interpolation will not achieve better performance. According to the channel estimation obtained in step 207, the channel estimation of the carrier on all frequency bands is obtained by linear interpolation by the following formula: For the Sounding sequence

P-l P-l

H' ^e (w) = (\--)H' ^m (bP + ) +—H' ^m ((b + \)P + (17)

2 2 where ^w is the position of the carrier, and P is the carrier spacing of the Cyclic mode carrying the Sounding signal.

 P-l b is an integer satisfying the condition bP<w< (b + l)P, and c is an integer satisfying the condition c = w_bP-,

 2

P-l

H ^lm (bP + ) is the MMSE channel estimate determined in step 20Π. For the Sounding sequence Decimation mode, use the following formula: H ^ls (w) = (1 - ^-) H ^lm (dD + g) + ^-H^{{d + \)D + g) where ^w is the position of the carrier, "For the carrier spacing of the Decimation mode carrying the Sounding signal, for the integer satisfying the condition ^ D<<(i/ + 1)D, e is an integer satisfying the condition e = w-iffi>- g, and g is the actual decimal offset The A' ^m (6E» + g) is the MMSE channel estimation determined in step 207. The embodiment of the present invention further provides a base station, which is used to implement the foregoing method. Figure 3 is a base station according to an embodiment of the present invention. As shown in FIG. 3, the base station includes: a receiving module 302, configured to receive a sounding signal sent by the terminal on each carrier; a first estimating module 304 coupled to the receiving module 302, configured to receive the detected signal according to the received signal The least square channel estimation is performed on each carrier with the preset detection signal corresponding to the terminal. The first obtaining module 306 is coupled to the first estimation module 304, and is configured to acquire each carrier according to the result of the least square channel estimation and the noise of each carrier. Correlation coefficient between; second acquisition mode Block 308, coupled to the first obtaining module 306, configured to obtain a mean square delay according to the correlation coefficient, and quantize the mean square delay and the signal to noise ratio of the carrier, in a correspondence between the preset quantization value and the filter matrix. Searching for a filter matrix corresponding to the quantized result; the second estimating module 310 is coupled to the second obtaining module 308, and configured to The result of the filter matrix and the least square channel estimation obtains a minimum mean square error channel estimate of each carrier; a third estimation module 312, coupled to the second estimation module 310, configured to perform linear interpolation using a minimum mean square error channel estimate to obtain all frequency bands Channel estimation for each carrier. Preferably, the second obtaining module 308 is configured to use the calculation of the mean square p ^p 1 + 2πΔ/σ _τ Ζ delay, where J is the frequency interval of the adjacent carriers, and Δ/ is the number of carriers between the carriers. The interval, R^(A/) is the correlation coefficient, and σ _τ is the mean square delay. Preferably, the first estimation module 304 is configured to perform a least square channel estimation using the formula ^O CB^B) - ^^??, where is the index of the carrier carrying the sounding signal; R = [r (0), r ( l), ..., r(L- l) is the received detection signal B = [b(0\ b(\), ..., b(L- l) is the preset detection corresponding to the terminal a signal, wherein, for the detection sequence loop mode, L = P, where P is a preset value, and for the sounding sequence sample mode, =1. According to an embodiment of the present invention, an MMSE channel estimation apparatus based on a sounding signal is further provided. The apparatus may be used to implement the probe signal based MMSE channel estimation method provided by the present invention. FIG. 4 is a structural block diagram of an MMSE channel estimation apparatus based on feature parameter quantization according to an embodiment of the present invention, as shown in FIG. The LS channel estimation module 401, the correlation coefficient calculation module 403, the mean square delay estimation and parameter quantization module 405, the MMSE channel estimation module 407, and the linear interpolation module 409. The LS channel estimation module 401 corresponds to the first estimation module. 304, set to match the Sounding signal and the received signal The respective carriers of the Sounding signal perform LS channel estimation; the correlation coefficient calculation module 403, corresponding to the first obtaining module 306, calculates the correlation coefficient between the carriers carrying the Sounding signal by using the LS channel estimation of the carrier carrying the Sounding signal; The delay estimation and feature parameter quantization module 405, corresponding to the second acquisition module 308, estimates the parameters that can characterize the channel characteristics according to the correlation coefficient of the carrier carrying the Sounding signal: a mean square delay. Simultaneously, the mean square delay and the signal The MMSE channel estimation module 407, corresponding to the second estimation module 310, is configured to calculate the MMSE channel estimation of each carrier carrying the Sounding signal according to the LS channel estimation and the filtering matrix; linear interpolation Module 409, corresponding to the third estimation module 312, is set to root According to the MMSE channel estimation of each carrier carrying the Sounding signal, linear interpolation is performed to obtain channel estimation of carriers on all frequency bands. In summary, the MMSE channel estimation method based on the sounding signal provided by the present invention has the following advantages: 1. When only the LS channel estimation and the linear interpolation channel estimation method are used, the channel response of the carrier position carrying the Sounding signal is susceptible to The effect of Gaussian noise. When the noise power is too large, the channel response on the other carriers is calculated according to the channel response at the carrier position of the sounding signal obtained by the LS channel estimation, and the performance of the system is significantly reduced. However, when the MMSE channel estimation method is used, part of the noise can be removed by the filter matrix, so that the channel estimation of other carriers is closer to the real channel response, and the performance of the system is also improved.

2. The derivation can be obtained by the following two parameters: Mean Square Delay and Signal Power Ratio Noise Power. The MMSE channel estimation method based on the sounding signal provided by the invention quantizes the two parameters and calculates the filter matrix in advance. In this way, when performing MMSE channel estimation, the filter interpolation matrix can be obtained by looking up the table, thereby avoiding matrix inversion and matrix multiplication in real time, which greatly reduces the operation. Moreover, the channel signal response can be obtained more accurately by using the sounding signal-based MMSE channel estimation method according to the present invention. Compared with the LS channel estimation algorithm and the linear interpolation channel estimation algorithm, the embodiment of the present invention can obtain a performance improvement of about 1.0 dB without significantly increasing the operation. Obviously, those skilled in the art should understand that the above modules or steps of the present invention can be implemented by a general-purpose computing device, which can be concentrated on a single computing device or distributed over a network composed of multiple computing devices. Alternatively, they may be implemented by program code executable by the computing device, such that they may be stored in the storage device by the computing device and, in some cases, may be different from the order herein. The steps shown or described are performed, or they are separately fabricated into individual integrated circuit modules, or a plurality of modules or steps are fabricated as a single integrated circuit module. Thus, the invention is not limited to any specific combination of hardware and software. The above is only the preferred embodiment of the present invention, and is not intended to limit the present invention, and various modifications and changes can be made to the present invention. Where in the invention ^" God and Within the principles, any modifications, equivalent substitutions, improvements, etc., are intended to be included within the scope of the present invention.

Claims

Claims A method of channel estimation, including:  The base station receives the sounding signal sent by the terminal on each carrier, and performs minimum square channel estimation on the respective carriers according to the received sounding signal and a preset sounding signal corresponding to the terminal;  And obtaining, by the base station, a correlation coefficient between the respective carriers according to a result of the least square channel estimation;  Obtaining, by the base station, a mean square delay according to the correlation coefficient, and performing quantization on the mean square delay and a signal to noise ratio of the carrier, and searching and corresponding in a correspondence between a preset quantization value and a filter matrix a filter matrix corresponding to the quantized result;  And obtaining, by the base station, a minimum mean square error channel estimate of each of the carriers according to a result of the filter matrix and the least square channel estimation; The base station performs linear interpolation using the minimum mean square error channel estimate to obtain channel estimates for each carrier on all frequency bands. The method according to claim 1, wherein the obtaining, by the base station, the mean square delay according to the correlation coefficient comprises: using, by the base station, ¹ - calculating a mean square delay, where J is a phase p ^p 1 + 2πΔ/σ _τ Ζ the frequency interval of the adjacent carrier, Δ/ is the interval of the number of carriers between the respective carriers, R _pp (Δ/) is the correlation coefficient, and σ _τ is the mean square delay . The method according to claim 1, wherein performing least square channel estimation on the respective carriers according to the received detection signal and a preset detection signal corresponding to the terminal comprises: using, by the base station, a formula A ^fa (m) = (B ^H B y ^x B ^H R to perform the least square channel estimation, where A ^fa (m) is the result of the least square channel estimation, and is the index of the carrier carrying the sounding signal, R = [ r(0), l),..., -l)] ^T is the received probe The signal is measured, B = [b(0\ b(\), ..., b(L - l) is the preset detection signal corresponding to the terminal, wherein, for the detection sequence loop mode, L = P , where P is the default value, and = 1 for the detection sequence. 4. The method according to claim 3, the acquiring, by the base station, the correlation coefficient between the respective carriers according to the result of the least square channel estimation comprises: using, by the base station, a formula k _pp (M) = E _m {H ^l m)H ^l m + Δ/)} Calculate the estimated correlation coefficient of the carrier Δ/the carrier carrying the detection signal; if Δ/ is zero, then R„(0) = „(Δ/) - σ „ ² ; if Δ/ is not zero, then R _pp (M) = R _pp (M) , where R^(A/) represents the actual correlation coefficient of the carrier separated by Δ/bearing detection signals, and E _m represents the average of each carrier w, k _pp ( Δ/) represents the estimated correlation coefficient of the carrier Δ/the carrier signal carrying the detection signal, and σ _η ² is the noise of the carrier. The method according to claim 3, the obtaining, by the base station, the minimum mean square error channel estimation of the respective carriers according to the filter matrix and the result of the least square channel estimation comprises: using, by the base station, a formula H _me = The WH _ls calculates a minimum mean square error channel estimate of each of the carriers, where is the filter matrix; and is a channel response of each carrier carrying the sounding signal obtained by estimating a minimum mean square error channel, and is A ^ta ( m) is a vector of elements, A _fa is a channel response of each carrier carrying the sounding signal estimated by the least square channel, and fl _ls is a vector with ^O) as an element. 6. The method of claim 5, For the probe sequence loop mode, the base station uses the minimum mean square error channel estimate linear interpolation to obtain all the channels of each carrier frequency band estimating comprises: calculating a total of the respective carriers on the frequency band in accordance with the following formula channel estimate A ^fe (t : Pl Pl H ^le (w) = (\ _— )H ^lm (bP + ) +—H' ^m ((b + \)P + ) 2 2 where ^w is the position of the carrier and P is the carrier spacing of the detection signal cyclic mode  P-l b is an integer satisfying the condition bP<w<(fi + l)P, and c is a condition satisfying the condition c = w_bP-  2 P-l An integer, H ^lm (bP + ), is the minimum mean square error channel estimate; 2 For the sounding sequence sampling mode, the base station performs linear interpolation using the minimum mean square error channel estimation, and obtaining channel estimates of each carrier on all frequency bands includes: calculating channel estimation A ^fe of each carrier in all frequency bands according to the following formula: ( ): H ^ls (w) = (1 - ^-) H ^lm (dD + g) + -H ^l '"((d + \)D + g) DD where ^ is the position of the carrier, is the carrier spacing of the mode in which the detection signal is carried, d is an integer satisfying the condition dD<w<(d + \)D, and e is an integer satisfying the condition e = w- - g, /) is the actual sample offset,

+ g) is the minimum mean square error channel estimate. The method according to claim 1, wherein quantizing the mean square delay and the signal to noise ratio of the carrier comprises:  Use the following formula to get the vector quantization proximity condition: ∑ f ⁺¹ (d(W(x^y\ W( ) -d(W(x^yl W(C _i+l (y) dy = 0 J ' (d(W(x, _yj W(C _U :)) -dO x, _yj ), W(C. _j+l )))p _s (x)p _R ( _yj )dx = 0 where FFT is the quantization order of the signal-to-noise ratio, (Χ) a probability distribution of the signal to noise ratio, wherein the quantization interval of the signal to noise ratio is {[χ _Μ , χ, .), 1 ≤ ≤ ; [, is the quantization order of the mean square delay, p _R (y For the probability distribution of the mean square delay, the quantization interval of the mean square delay is "[[ -! , y _J ll ≤ _j ≤ n _R ], and C _u is x _s and _y _R respectively Quantization samples for [χ _Μ , χ, . ) and [;^, , _yj ), for the filter matrix, for the quantization distortion; The mean squared delay and the signal to noise ratio of the carrier are quantized according to the vector quantization neighboring condition and the centroid condition. 8. A base station, including:  a receiving module, configured to receive a sounding signal sent by the terminal on each carrier; the first estimating module is configured to minimize the each carrier according to the received detecting signal and a preset sounding signal corresponding to the terminal Square channel estimation;  a first acquiring module, configured to acquire a correlation coefficient between the respective carriers according to a result of the least square channel estimation;  a second acquiring module, configured to obtain a mean square delay according to the correlation coefficient, and quantize the mean square delay and a signal to noise ratio of the carrier, in a correspondence between a preset quantization value and a filter matrix Finding a filter matrix corresponding to the quantized result;  a second estimation module, configured to obtain a minimum mean square error channel estimate of each of the carriers according to the filter matrix and the least square channel estimation result;  A third estimation module is arranged to perform linear interpolation using the minimum mean square error channel estimate to obtain channel estimates for respective carriers on all frequency bands. 9. The base station according to claim 8, wherein the second obtaining module is configured to use a calculation of a mean square delay, where J is a frequency interval of adjacent carriers, p ^p 1 + 2πΔ/σ _τ Ζ Δ/ is the interval of the number of carriers between the respective carriers, and R^(A/) is the correlation coefficient. The mean square delay. 10. The base station according to claim 8, wherein the first estimation module is configured to perform the least square channel estimation using a formula, where A ^fa (m) is the least square channel estimation The result is an index of a carrier carrying the sounding signal; R = [r(0), r(l), ..., r(J -l) is the received detection signal; = [ 6(0)^(1),... (J _1)] ^T is the preset detection signal corresponding to the terminal, wherein, for the detection sequence loop mode, L = P, where P is a preset Value, for the probing sequence, = 1,.