JP4381900B2 - Channel estimation and data detection method - Google Patents

Description

  The present invention relates to a communication path estimation and data detection method, and more particularly to a communication path estimation and data detection method for performing data detection by estimating a communication path in a MIMO system that performs wireless communication using a plurality of transmission / reception antennas.

  In recent years, a MIMO (Multiple-Input Multiple-Output) system using a plurality of transmission / reception antennas has attracted attention as means for realizing high-speed and large-capacity transmission in wireless communication.

  However, in the MIMO system, if the accuracy of the channel state information is low, the characteristics are greatly degraded due to interference from other antennas. Therefore, in order to detect a signal from each antenna, it is necessary to accurately estimate the state of the communication channel at the receiver and to know the state of the communication channel.

  As a channel estimation method, for example, an Expectation-Maximization (EM) algorithm may be used (see, for example, Non-Patent Document 1). This EM algorithm is used in, for example, channel estimation and data in a MISO (Multiple-Input Single-Output) system described in Non-Patent Document 6 and a space-time block coding (STBC) system described in Non-Patent Document 2. It is known to be an excellent algorithm for detection.

  If maximum likelihood (ML) estimation is difficult when estimating a communication path, the EM algorithm can approximate the ML solution by repeating the communication path estimation a plurality of times.

  The EM algorithm is a method using a log-likelihood function composed of two iteration steps, and performs the following processing.

  First, in Expectation-step (E-step), which is the first step, an expected value of the log likelihood function is calculated. In a second step, Maximization-step (M-step), a parameter that maximizes the function obtained in E-step is obtained.

  However, the conventional EM algorithm has a problem that the convergence speed becomes slow because all parameters for channel estimation and data detection are updated simultaneously. There is also a problem that it is impossible to effectively follow the fluctuation of the communication path.

  Therefore, a SAGE (Space-Alternating Generalized EM) algorithm has been proposed that overcomes the disadvantage that the convergence speed of the EM algorithm is slow by sequentially updating each parameter (see, for example, Non-Patent Document 5).

For example, SAGE is used for channel estimation and data detection in the direct spreading code division multiple access (DS-CDMA) system described in Non-Patent Document 3 and the space-time code (STC) system described in Non-Patent Document 4. The algorithm is applied.
CN Georghiades and JC Han, "Sequence estimation in the presence of random parameters via the EMalgorithm," IEEE Trans.Commun., Vol. 45, no.3, pp.300-308, March 1997. B. Lu and X. Wang and Y. Li, "Iterative receivers for space-time block coded OFDM systems in dispersive fading channels," IEEE Trans. Wireless Commun., Vol. 1, no. 2, pp.213-225, April 2002. A. Kocian and BH Fleury, "Iterative joint symbol detection and channel estimation for DS / CDMA via the SAGE algorithm," IEEE PIMRC 2000. vol. 2, pp.1410-1414. Y. Xie and CN Georghiades, "Two EM-type channel estimation algorithm for OFDM with transmitter diversity," IEEE Trans. Commun., Vol. 51, no.1, pp.106-115, Jan. 2003. JA Fessler and AO Hero, "Space-alternating generalized expectation-maximization algorithm," IEEE Trans. On Signal Processing, vol. 42, no.10, pp.2664-2677, Oct 1994. CH Aldana and J. Cioffi, "Channel tracking for multiple Input, signal output systems using EM algorithm," IEEE ICC 2001. vol.2, pp. 586-590.

  As described above, the MIMO system is effective as a means for realizing high-speed and large-capacity transmission in wireless communication, and the accuracy of channel estimation is important. However, the MIMO system reduces channel complexity while reducing the complexity of the device. There has not yet been proposed a method that can achieve an excellent error rate characteristic by following.

  The present invention has been made in consideration of the above facts, and an object of the present invention is to provide a channel estimation and data detection method that can achieve excellent error rate characteristics by following channel variations while reducing the complexity of the apparatus. It is.

In order to achieve the above object, the invention according to claim 1 is characterized in that transmission data transmitted from a plurality of transmission antennas is received by a plurality of reception antennas, and the transmission antennas are connected to each other based on the reception data received by each reception antenna. In the channel estimation and data detection method for estimating the channel gain with the receiving antenna and detecting the transmission data, the transmission data is divided into a plurality of sub-blocks, and the reception data and the predetermined transmission data estimation Obtaining a channel estimation value representing a channel gain for a channel between each transmission antenna based on the value for each reception antenna, and determining a transmission antenna whose transmission data estimation value should be updated Obtaining an expected value of the received data for each receiving antenna; and transmitting data estimated values at which the expected value is maximized for each receiving antenna. And a step of updating a transmission data estimation value corresponding to the transmission antenna to be updated for each reception antenna based on the transmission data estimation value and the communication path estimation value that maximize the expected value. a step of decoding by combining the transmitted data estimated value updated for each receiving antenna, and setting a transmission data estimation value updated as the predetermined transmission data estimates, the processing including the sub Repeat at least the number of transmission antennas for each block, and based on the channel estimation value for the last data of the sub-block, obtain the initial transmission data estimation value of the next sub-block, and calculate the obtained initial transmission data estimation value. The predetermined transmission data estimated value is set .

  The present invention is an invention applied to a multi-input multi-output system having a plurality of transmission antennas and a plurality of reception antennas. The transmission data transmitted from the plurality of transmission antennas is received by the plurality of reception antennas, and each reception is performed. Based on the reception data received by the antenna, the transmission data is detected by estimating the channel gain between the transmission antenna and the reception antenna.

A channel estimation value representing a channel gain for a channel with each transmission antenna is obtained for each reception antenna based on the reception data and a predetermined transmission data estimation value. Note that, as described in claim 4 , the channel estimation value can be obtained by a minimum mean square error method.

  As the predetermined transmission data estimated value, an initial transmission data estimated value is set in the estimation of the first communication channel estimated value, and thereafter the transmission data estimated value updated last time.

The initial transmission data estimated value can be obtained by the procedure described in claim 2. That is, according to the second aspect of the present invention, the transmission data includes predetermined known data, an initial communication path estimated value is obtained based on the received data corresponding to the known data and the known data, and the obtained initial communication is obtained. An initial transmission data estimated value is obtained based on a path estimated value and received data corresponding to the known data, and the obtained initial transmission data estimated value is set as the predetermined transmission data estimated value. The known data is inserted in advance at the head of the transmission data transmitted from each transmission antenna. In addition, as described in claim 3 , the initial transmission data estimated value can be calculated by maximum likelihood detection.

  Then, the transmission antenna whose transmission data estimated value is to be updated is determined. That is, in the present invention, the SAGE algorithm is applied to a multi-input multi-output system, and the transmission data estimation values for all the transmission antennas are not updated simultaneously, but the transmission data estimation values for some transmission antennas are updated sequentially. I will do it. Thereby, the speed at which the calculation converges can be increased.

  Next, as an E-step process of the SAGE algorithm, an expected value of received data is obtained for each reception antenna, and as an M-step process, a transmission data estimated value with the maximum expected value is obtained for each receive antenna.

  Then, based on the transmission data estimation value and the channel estimation value with the maximum expected value, the transmission data estimation value corresponding to the transmission antenna to be updated is updated for each reception antenna, and updated for each reception antenna. The transmission data estimated value is synthesized and decoded.

  The updated transmission data estimated value is set as a predetermined transmission data estimated value and used for the next iterative calculation.

These processes are repeated at least as many as the number of transmission antennas for each subblock obtained by dividing the transmission data into a plurality of subblocks , and the initial value of the next subblock is determined based on the channel estimation value for the last data of the subblock. A transmission data estimated value is obtained, and the obtained initial transmission data estimated value is set as the predetermined transmission data estimated value . That is, in the SAGE algorithm, the transmission data estimation values of all the transmission antennas are not updated at the same time, but the transmission data estimation values are sequentially updated. The transmission data estimated value is updated once per antenna.

In this way, since the SAGE algorithm is applied to the multi-input multi-output system to estimate the channel estimation value and the data detection, it is possible to achieve excellent error rate characteristics while reducing the complexity of the apparatus. Also, by exchanging channel estimation values between sub-blocks and applying the SAGE algorithm for each sub-block, it is possible to obtain excellent error rate characteristics following channel variations while reducing the amount of computation.

  As described above, the present invention has an excellent effect that it is possible to achieve excellent error rate characteristics by following communication path fluctuations while reducing the complexity of the apparatus.

  Hereinafter, an example of an embodiment of the present invention will be described in detail with reference to the drawings.

  FIG. 1 shows a schematic block diagram of a wireless communication system 10 according to the present invention. As shown in FIG. 1, the wireless communication system 10 includes a transmitter 12 and a receiver 14.

  The transmitter 12 includes a symbol mapping unit 16 and a demultiplexer 18, and the receiver 14 includes a communication path estimation and data detection unit 20, a multiplexer 22, and a symbol demapping unit 24. Yes.

N (N ≧ 2) transmission antennas 26 _{1 to} 26 _N are connected to the demultiplexer 18, and M (M ≧ 2) reception antennas 28 ₁ are connected to the communication path estimation and data detection unit 20. ~ 28 _M is connected.

  The symbol mapping unit 16 performs symbol mapping on the input signal according to the modulation scheme, and outputs the transmission symbol matrix X to the demultiplexer 18.

  For example, when a transmission signal is modulated by QPSK (Quadrature Phase Shift Keying) and transmitted, one transmission symbol is generated with 2 bits. Therefore, assuming that (transmission bit, transmission symbol), i is an imaginary number, The four types of transmission symbols can be expressed on the complex plane, such as (10, 1 + i), (00, -1 + i), (01, -1-i), and (11, 1-i). In this case, if the sign of either the real part or the imaginary part of the transmission symbol is inverted, it can be determined that one of the two transmission bits is incorrect, and both the real part and the imaginary part Can be determined that all two transmission bits are wrong. As described above, the symbol mapping unit 16 maps the transmission signal to any transmission symbol on the complex plane according to the gray mapping.

The demultiplexer 18 generates transmission symbol sequences X _{1 to} X _N to be transmitted by the transmission antennas 26 _{1 to} 26 _N from the transmission symbol matrix X, and outputs them to the corresponding transmission antennas 26 _{1 to} 26 _N , respectively. Thereby, a transmission symbol sequence corresponding to each transmission antenna is transmitted from each transmission antenna.

Note that the transmission symbol sequence (matrix) X _n (n = 1, 2,... N) has a frame length of a transmission frame transmitted from each transmission antenna as L, and X _n = [X _n ¹ , X _n ² , ... represented by X _n ^{L] T.} Here, the superscript T represents transposition, and X _n ^L represents the Lth transmission symbol transmitted from the nth transmission antenna.

Further, p known symbols, which are determined in advance, are inserted at the beginning of each transmission symbol sequence X _n and transmitted.

In the receiver 14, signals transmitted from the transmission antennas 26 _{1 to} 26 _N are received by the reception antennas 28 _{1 to} 28 _M and taken into the communication path estimation and data detection unit 20.

The kth received symbol Y _m ^k received by the mth receiving antenna is expressed by the following equation.

H ^k _{n, m} is a complex Gaussian random variable, which is a channel gain between the n-th transmitting antenna and the m-th receiving antenna, and W ^k _m is zero average and noise variance σ ² . It has Gaussian noise.

Here, the received symbol sequence _{(matrix) Y m (m = 1,2,} ... M) ^{_{a, Y m = [Y m 1}} , Y m 2, ... Y m L] putting ^{T, the} above equation (1) Can be expressed as:

Here, X = [diag (X ₁ )... Diag (X _N )] is an L × NL transmission symbol matrix, and H _m = [H _{1, m} ... H _{N, m} ] ^T is an NL × 1 channel vector. _{, H n, m = [H} 1 n, m ... H L n, m] T is channel vector of _{L × 1, W m = [} W 1 m ... W L m] T is a noise vector of L × 1 is there.

In channel estimation and data detection unit 20 will be described in detail later, communication between the transmission antennas and the reception antennas based on the received symbol sequence Y ₁ to Y _M and the like that are received by the receiving antenna 28 ₁ ~ 28 _M A channel vector H _m representing a channel gain is estimated, and a transmission symbol sequence X _n is estimated based on the channel vector H _m .

  Next, before describing the SAGE algorithm that is an algorithm used for channel estimation according to the present embodiment, the EM algorithm described in Non-Patent Document 1, Non-Patent Document 6, and the like will be described.

  First, bεB is a parameter vector estimated from the observation vector yεY using the probability distribution p (y | b). The EM algorithm obtains an estimated value of b by repeating the calculation when ML estimation of b is difficult. The derivation of the algorithm depends on the unobservable complete data zεZ, and if this is observable, the ML estimate can be easily calculated.

  The observed random variable y called as incomplete data in the framework of the EM algorithm has a relationship of mapping z | → y (z) as shown in FIG.

  The EM algorithm uses an incomplete data and desired estimated data to calculate an expected value of complete data, an E-step (expectation step), and an M-step (maximization step) that maximizes the function obtained in the E-step. ) Of two iteration systems.

The calculation formula for the i-th iterative calculation of the conditional expected value Q (b | b ^[i] ) in E-step is expressed by the following equation.

In M-step, as shown by the following equation, a parameter vector that maximizes the function obtained by the above equation (3) is obtained, and this is updated as an estimated value b ^{[i + 1]} in the (i + 1) -th iteration calculation. To do.

  Since the complete data z cannot actually be observed, the maximization of Λ (z | b) is performed using the incomplete data y and the latest estimated value of the parameter vector b.

If b ^[i] is a sequence of estimated values generated by the initial value b ^{[0] of the} EM algorithm, Λ (y | b ^[i] ) does not decrease (monotonic characteristic). The EM algorithm depends on the initial value, and the convergence speed is inversely proportional to the ratio of missing information. This convergence speed is also inversely proportional to the size of the complete data space.

  Next, the SAGE algorithm will be described. Here, as an example, the SAGE algorithm proposed by Fessler et al. Described in Non-Patent Document 4 will be described.

In the SAGE algorithm, not all parameters are updated simultaneously at the i-th iteration, but only the subset b _{S of} b indexed by S = S [i] is updated, and the other parts shown in the following equation The set is not updated.

  This partially updating operation makes the convergence speed of the SAGE algorithm faster than that of the EM algorithm.

  The formula for the i-th iterative calculation in E-step is expressed by the following formula.

Also, only b _S is updated in the repetitive calculation in M-step. A calculation formula for updating only b _S is expressed by the following formula.

  Next, the channel estimation and data detection method executed in the channel estimation and data detection unit 20 of the receiver 14 according to the present embodiment, that is, the transmission symbol is estimated by applying the SAGE algorithm to the MIMO system. A method of detecting the will be described.

First, initial estimation of a communication path using known symbols will be described. Conventionally known MMSE (Minimum Mean Square Error) channel estimation is applied as the channel estimation. Further, the known symbol series is assumed to be X ^tr .

Since the channel vector H _m and the noise vector N _m are uncorrelated, the channel vector H ^tr for the known symbol block X ^{tr is} expressed by the following equation (8) as described in Non-Patent Document 7 shown below. It is represented by

  (Non-Patent Document 7) S. Haykin, Adaptive Filter Theory, Englewood Ciffs, N. J .: Prentice-Hall, third ed., 1996.

Here, it is assumed that I _p is a p × p unit matrix, and R _hh ^tr is a communication channel auto-covariance matrix for known symbols and is known. Superscript H represents Hermitian transpose. Also, σ ² is the variance of additive white Gaussian noise, and is assumed to be known in this embodiment. Y _m ^tr represents a known symbol sequence received by the m-th receiving antenna 28 _m .

  The channel estimation value obtained by the above equation (8) is averaged in the time direction of known symbols. The averaged channel estimation value is expressed by the following equation.

  Then, using the average channel estimation value E obtained as described above, the initial transmission symbol sequence estimated value represented by the following equation (10) is subjected to ML detection (maximum likelihood detection) represented by the following equation (11). Ask for.

  X is a predicted value of a transmission symbol, and when a transmission signal is modulated by QPSK and transmitted, it is one of four types of ± 1 ± i.

The initial transmission symbol sequence estimated value obtained by the above equation (11) is used as the initial transmission symbol sequence estimated value (X ^[0] ) in the SAGE algorithm described later.

  Next, channel estimation and data detection by the SAGE algorithm will be described. Note that, in the transmission symbol portion other than the known symbols, the channel estimation process and the data detection process in each reception antenna are the same.

  Like the EM algorithm described above, the parameter vector to be estimated is b = X, and the incomplete data is y = Y. When complete data is selected as z = {Y, H}, the log likelihood function of z is expressed by the following equation as described in Non-Patent Document 3.

  Note that the second term of the above equation (12) is independent of the transmission signal and can be ignored because it is a constant value. [Lambda] (Y | H, X) in the above equation (12) can be expressed by the following equation (13) based on a Gaussian distribution as described in Non-Patent Document 8 shown below.

  (Non-Patent Document 8) J. G. Proakis, Digital Communication, Fourth Edition, New York: McGraw-Hill, 2001.

This conditional log-likelihood function Λ (Y ^k | H, X) can be expressed as the following equation if a constant term is ignored.

Superscript * indicates a complex conjugate. Here, choosing a subset b _S of b in SAGE algorithms and transmission frame X _n from the n-th transmit antenna (b _S = X _n). Then, excluding terms independent of X _n ^k , the function in E-step in the i-th iterative calculation, that is, the conditional expected value of the received symbol Y is expressed by the following equation.

  In the i-th calculation in the repetitive calculation of the above equation (15), only symbols transmitted from the nth (n = (i mod N) +1) transmission antenna are updated, and symbols transmitted from other transmission antennas are not updated. . That is, as i increases, the index of the transmission antenna of the symbol to be updated also increases, and the symbol of the transmission antenna that was updated first is updated again in the (N + 1) th iterative calculation.

  Then, the conditional expected value of the communication channel is expressed by the following equation using the MMSE estimation method.

Here, R _hh is a self-covariance matrix of the communication path and is assumed to be known. σ ² is the variance of additive white Gaussian noise, and is assumed to be known in this embodiment. I _N is an N × N unit matrix. Then, as shown by the following equation, the solution of M-step can be obtained by differentiating the above equation (15) with respect to (X _n ^k ) ^* .

Therefore, the transmission symbol sequence estimated value X _n ^{k [i + 1]} in the (i + 1) -th repetitive calculation can be expressed by the following equation.

  The transmission symbol sequence estimation value is estimated for each receiving antenna. The transmission symbol sequence estimation values estimated for the respective receiving antennas are combined by a maximum ratio combining method (MRC: Maximum Ratio Combining).

  The maximum ratio combining is performed by weighting each transmission symbol with its channel estimation value as shown in the following equation.

  Here, F {·} means a hard decision. The maximum ratio synthesis is performed every time iterative calculation is performed once.

  In the hard decision, when the transmission signal is modulated by QPSK and transmitted, the positive and negative of the real part and the imaginary part are respectively determined. If positive, +1 (+ i) is determined. judge. As a result, the transmission symbol is decoded.

  Then, the transmission symbol sequence estimated value synthesized by the above equation (20) is used in the next iterative calculation of each receiving antenna.

  In this way, by applying the SAGE algorithm to the MIMO system and estimating the communication channel and detecting the transmission symbol, the bit is superior to ML detection that detects the transmission symbol by estimating the communication channel using a known symbol. An error rate (BER) can be obtained.

  Next, a Sim-SAGE algorithm that divides a transmission frame into a plurality of sub-blocks, applies a SAGE algorithm to each sub-block, estimates a communication path, and detects a transmission symbol will be described.

In the SAGE algorithm, MMSE that performs channel estimation requires an inverse matrix corresponding to the frame length L of the transmission frame, and the order of the amount of computation is O (L ³ ). In addition, since the initial data detection accuracy in the rear part of the frame deteriorates, when the SAGE algorithm is applied to the entire transmission frame, if there is a change in the communication path, it may not be possible to follow this communication path change well.

  Therefore, one transmission frame is divided into sub-blocks (SB) for each l symbol, and the SAGE algorithm is applied to each sub-block. That is, as shown in FIG. 3, a transmission symbol sequence X including L symbols is divided into X = {X [1],..., X [B],. A SAGE algorithm is applied to each sub-block. In the following, B is an index of a sub-block.

In the Sim-SAGE algorithm, in order to be able to follow the channel variation, the channel estimation value for the l-th symbol of the previous sub-block (the last symbol of the sub-block) is used as the initial estimation value of the next sub-block. Used as Thus, by dividing the transmission frame into sub-blocks and applying the SAGE algorithm to each sub-block, the order of the amount of computation associated with the channel estimation of all sub-blocks in one frame is O ((L / l) Xl ³ ) = O (Ll ² ), and the amount of calculation can be reduced. Further, by using the channel estimation value of the previous sub-block as the initial channel estimation value of the next sub-block, the effect of following the channel can be obtained.

Note that the initial transmission symbol estimated value X ^{k [0]} [B] = [X ^k ₁ [B],..., X ^k _N [B]] can be obtained by the following equation by ML detection.

  Next, the flow of processing when the communication path estimation and data detection unit 20 performs communication path estimation and data detection using the Sim-SAGE algorithm will be described with reference to the flowchart shown in FIG.

First, in step 100, based on the received symbols Y _m ^tr and the known symbol series X ^tr received by the receiving antennas 28 _{1 to} 28 _M , the channel estimation value H for the known symbol series X ^tr according to the above equation (8). ^{Find tr} .

In step 102, an average channel estimation value E [H _m ^tr ] is obtained by averaging the channel estimation value H ^tr obtained in step 100 in the time direction of known symbols as shown in the above equation (9).

  In step 103, a sub-block length l (= frame length L / number of sub-blocks) is set.

In step 104, “1” is assigned to the index B of the sub-block. In the next step 106, the average channel estimation value E [H _m ^tr ] obtained in step 102, the received symbols Y _m ^k (k = p + 1 to L) after the known symbols received by the receiving antennas 28 _{1 to} 28 _M. ), The initial transmission symbol sequence X ^k represented by the above equation (10) is ^obtained by ML detection using the above equation (11). The obtained initial transmission symbol sequence X ^k is used as X ^[0] [0] in the SAGE algorithm after the next step.

In step 108, “0” is substituted for the number of iterations i of the iteration calculation. In step 110, based on the transmission symbol sequence X ^[i] [B] and the reception symbol sequence Y, the conditional expected value E {H [B] | Y [B], X of the channel vector according to the above equation (17). ^[i] [B]}, that is, the channel estimation value H ^[i] [B] of equation (16) is obtained for each receiving antenna.

  In step 112, the transmission antenna number n whose parameter is to be updated is determined by modulo calculation (n = (i mod N) +1).

In step 114, as the E-step processing in the SAGE algorithm, based on the channel estimation value H ^[i] [B], the transmission symbol sequence X ^[i] [B], and the reception symbol sequence Y [B] ( 15) The conditional expected value Q (X _n | X ^[i] [B]) of the received symbol sequence Y [B] is obtained for each receiving antenna.

In step 116, as M-step processing in the SAGE algorithm, processing for differentiating the conditional expected value of the received symbol sequence Y [B] obtained in step 114 by (X _n ^k ) ^* by the above equation (18). This is done for each receiving antenna. As a result, a value that maximizes the conditional expected value of the M-step solution, that is, the received symbol sequence Y [B], can be obtained.

The transmission symbol estimated value X _n ^{k [i + 1]} [B] is calculated for each of the receiving antennas 28 _{1 to} 28 _M.

In step 120, each of the transmission symbol estimated values X _n ^{k [i + 1]} [B] calculated for each receiving antenna by the above equation (20) is used as the corresponding channel estimated value H _n ^[i] [ B] is weighted and added to perform maximum ratio composition processing and hard decision.

In step 122, it is determined whether or not the number of repetitions (i + 1) is equal to or greater than I _max, that is, whether or not the calculations in steps 110 to 120 are repeated I _max times, and the number of repetitions (i + 1) is equal to or greater than I _max In step S124, the process proceeds to step 124. On the other hand, if the number of repetitions (i + 1) is less than I _max , the process proceeds to step 126, the number of repetitions i is incremented, the process proceeds to step 110, and the same processing as described above is repeated.

Since it is necessary to estimate transmission symbols for all transmission antennas, I _max is a value at least equal to or greater than N, and is set to a value sufficient for convergence of the channel estimation value.

  In step 126, it is determined whether or not B + 1 exceeds L / l, that is, whether or not the detection of the transmission symbol series has been completed for all the sub-blocks. If B + 1 exceeds L / l, it is determined that transmission symbol sequence estimation has been completed for all sub-blocks, and this routine is terminated. On the other hand, if B + 1 is equal to or less than L / l, it is determined that transmission symbol sequence estimation has not been completed for all sub-blocks, and the process proceeds to step 128 to increment B, which is the index of the sub-block.

In step 130, the next sub-block (B-th sub-block) is calculated using the channel estimation value H ^l [B-1] of the previous sub-block (B-1-th sub-block) according to the above equation (21). The initial transmission symbol estimated value X ^{k [0]} [B] is calculated. That is, the channel estimation value for the l-th transmission symbol of the previous sub-block (the last transmission symbol of the sub-block) is used for calculating the initial transmission symbol estimation value of the next sub-block. Thereby, even when the channel fluctuation is large, it is possible to follow the channel fluctuation and to obtain an excellent error rate characteristic.

Then, the process returns to step 108 to reset the repetition number i, and the same processing as described above is performed. By performing these processes for all the sub-blocks, the transmission symbol sequence X _n transmitted from each of the transmission antennas 26 _{1 to} 26 _N is detected.

  FIG. 5 is a block diagram illustrating the channel estimation and data detection unit 20 when the Sim-SAGE algorithm is executed.

  As shown in FIG. 5, the channel estimation and data detection unit 20 performs channel estimation and symbol detection by a tracking unit 20A that performs ML detection of an initial transmission symbol estimation value and the like for each sub-block, and a SAGE algorithm. And a SAGE unit 20B that performs a maximum ratio combining process, a hard decision process, and the like. A brief description will be given below.

In tracking unit 20A, ML detection unit 30 detects initial transmission symbol estimated value X ^[0] [B] by ML detection. This corresponds to the processing of steps 106 and 130 in FIG. Note that the channel estimation value H ^tr is used for the first sub-block (B = 1), and H ^l [B−l] is used for the subsequent sub-blocks.

In the SAGE unit 20B, the channel estimation unit 32 calculates the channel estimation value H ^[i] [B] and outputs it to the symbol detection unit 34. This process corresponds to the process of step 110 in FIG. In the initial channel estimation, the initial transmission symbol estimation value X ^[0] [B] from the tracking unit 20A is used, and the transmission symbol estimation value X detected by the symbol detection unit 34 in the subsequent iterative calculation. ^[i] [B] is used.

The symbol detection unit 34 transmits a transmission symbol sequence X _n ^{[i +]} transmitted from the transmission antenna number n determined by the modulo calculation unit 36 based on the channel estimation value H ^[i] [B] and the reception symbol sequence Y. ^1] [B] is obtained for each receiving antenna and output to the maximum ratio combining and hard decision unit 38. This process corresponds to the process of step 118 in FIG.

The maximum ratio combining and hard decision unit 38 combines the transmission symbol sequence X _n ^{[i + 1]} [B] obtained for each receiving antenna by maximum ratio combining and performs hard decision processing to determine a transmission symbol sequence. To do. This process corresponds to the process of step 120 in FIG.

Then, it is determined whether or not iterative calculation has been performed I _max times by the repetition number determination unit 40. If the number of repetition calculations is less than I _max , i is incremented by the incrementing unit 42 and the repetitive calculation is similarly performed. On the other hand, when the number of repetitive calculations reaches I _max , the determination unit 44 of the tracking unit 20A determines whether or not the estimation of transmission symbol sequences has been completed for all sub-blocks. , B is incremented by the increment unit 46, and the next sub-block is repeatedly calculated in the same manner as described above.

  In the present embodiment, the case of performing channel estimation using MMSE as the channel estimation method has been described. However, the present invention is not limited to this, and channel estimation is performed using a simple algorithm such as ZF (Zero Forcing). You can go.

  Next, examples of the present invention will be described.

The simulation parameters were set as follows. Number of transmitting antennas N = 2, number of receiving antennas M = 2, data modulation method is QPSK (Quadrature Phase Shift Keying) modulation method, communication channel is flat Rayleigh fading channel, symbol length T _S = 5 μs, transmission frame length L = 40, the number of known symbols p = 4, and the number of data symbols was 36.

  Further, since it is assumed that the transmission symbol estimation values from all the transmission antennas are updated and the iteration calculation is completed in one stage, the iteration calculation is completed for the number N of transmission antennas, and thus the iteration calculation is completed in one stage. Become.

FIG. 6 shows the number of iterative calculation steps by the SAGE algorithm and the mean square error (MSE) of the channel estimation value when the signal power to noise power ratio E _b / N ₀ per bit is 10 dB and 20 dB. The result of simulating the relationship is shown.

  As is apparent from FIG. 6, it can be seen that the convergence has been sufficiently achieved by performing only one iteration of the calculation. Therefore, the following simulation shows the results when one iteration of the calculation is performed.

  As notations, ML-only is the characteristic of ML detection using known symbols, SAGE is the characteristic of the proposed algorithm that does not divide the transmission frame, and Sim-SAGE is the characteristic of the proposed algorithm that divides the transmission frame into a plurality of sub-blocks.

Figure 7 is, ML estimates, for each of Simp-SAGE algorithm according to the present invention, the sub-block length (SBL) l and bit error rate (BER: Bit Error Rate) and the maximum Doppler frequency F _d × symbol length the relationship The simulation results with varying T _s are shown. F _d T _s represents the speed of fluctuation of the communication path, and simulation was performed for three patterns of 1.0 × 10 ⁻³ , 5.0 × 10 ⁻⁴ , and 1.0 × 10 ⁻⁴ . Further, in FIG. 7, “ML-only” indicates that the simulation is performed only by ML estimation, and “Simp-SAGE” indicates that the simulation is performed using the Sim-SAGE algorithm.

  When the sub-block length l is equal to the transmission frame length L, that is, when l = L = 40, the SAGE algorithm and the Sim-SAGE algorithm are equivalent.

  As shown in FIG. 7, in the case of the Sim-SAGE algorithm, it can be seen that there is an optimum sub-block length. This is because if the sub-block length is shortened, that is, if the number of divisions for one transmission frame is increased, the effect of tracking channel fluctuations can be obtained, but the number of data series samples used for channel estimation decreases, and channel estimation On the other hand, if the subblock length is increased, the effect of tracking channel fluctuations decreases, but the number of data series samples used for channel estimation increases and the accuracy of channel estimation increases. by.

As is apparent from FIG. 7, the optimum sub-block length is 8 symbols in the case of F _d T _S = 1.0 × 10 ⁻³ , F _d T _S = 5.0 × 10 ⁻⁴ , 1.0 × In the case of 10 ⁻⁴ , it can be seen as 10 symbols. In an environment where communication path fluctuations are severe, the optimum sub-block length is shortened so that the tracking effect can be obtained effectively.

  Table 1 below shows the amount of calculation in the case of the Sim-SAGE algorithm when the subblock length is 8 symbols and 10 symbols in the case of the SAGE algorithm. As described above, the transmission frame length L = 40 and the number of known symbols p = 4.

  As apparent from Table 1, it can be seen that the calculation amount is reduced when the sub-block length is shortened. It can also be seen that the amount of computation is greatly reduced by dividing the transmission frame.

FIG. 8 shows the result of simulating the relationship between E _b / N ₀ and BER when F _d T _S = 1.0 × 10 ⁻³ and 1.0 × 10 ⁻⁴ . In the case of the Sim-SAGE algorithm, the simulation was performed with F _d T _S = 1.0 × 10 ⁻³ and 1.0 × 10 ^{−4 with} the optimum sub-block length l = 8 and 10, respectively.

As is apparent from FIG. 8, in the case of the SAGE algorithm, it can be seen that E _b / N ₀ is improved by about 1 dB compared to ML-only. Further, in the case of the Sim-SAGE algorithm, compared with ML-only, at F _d T _S = 1.0 × 10 ⁻³ , BER = 10 ⁻² is about 2.5 dB, and BER = 10 ⁻³ is about 10 dBE. _b / N ₀ is improved, the _{F d T S = 1.0 × 10} -4, it can be seen that about 2dBE _b / N ₀ is improved. That is, it can be seen that the improvement effect of the Sim-SAGE algorithm is larger in the fast fading environment than in the slow fading environment. In ML-only, only the channel estimation value obtained from the known symbol at the head of the frame is used, whereas in the Sim-SAGE algorithm, channel channel fluctuation is obtained by passing the channel estimation value between sub-blocks. It is because it is able to follow against.

  Thus, it has been found that the Sim-SAGE algorithm can achieve excellent error rate characteristics by following the channel fluctuation effectively and with a low amount of computation even in a high-speed fading environment where the channel fluctuation is large.

1 is a block diagram of a wireless communication system. It is a figure for demonstrating mapping. It is a figure for demonstrating the division | segmentation of a flame | frame. It is a flowchart of the process routine performed by a communication path estimation and data detection part. It is a block diagram of a communication path estimation and data detection part. It is a diagram which shows the relationship between the number of steps of repeated calculation, and MSE. It is a diagram which shows the relationship between subblock length and a bit error rate. It is a diagram which shows the relationship between a signal power to noise power ratio and a bit error rate.

Explanation of symbols

10 wireless communication system 12 transmitter 14 receiver 16 symbol mapping unit 18 demultiplexer 20 data detecting unit 22 multiplexers 24 symbol demapping unit 26 ₁ ~ 26 _N transmitting antennas 28 ₁ ~ 28 _M receive antennas

Claims (4)

The transmission data transmitted from the plurality of transmission antennas is received by the plurality of reception antennas, and the channel gain between the transmission antenna and the reception antenna is estimated based on the reception data received by each reception antenna, In the channel estimation and data detection method for detecting transmission data,
Dividing the transmission data into a plurality of sub-blocks;
Obtaining a channel estimation value representing a channel gain for a channel with each transmission antenna based on the reception data and a predetermined transmission data estimation value for each reception antenna;
Determining a transmit antenna for which the transmit data estimate is to be updated;
Obtaining an expected value of the received data for each receiving antenna;
Obtaining a transmission data estimate for which the expected value is maximized for each receiving antenna;
Updating the transmission data estimation value corresponding to the transmission antenna to be updated for each reception antenna based on the transmission data estimation value and the communication path estimation value with the maximum expected value;
Synthesizing and decoding the transmission data estimate updated for each receiving antenna; and
Setting the updated transmission data estimated value as the predetermined transmission data estimated value;
Is repeated for at least the number of transmission antennas for each sub-block, and an initial transmission data estimation value for the next sub-block is obtained based on the channel estimation value for the last data of the sub-block, and obtained. The initial transmission data estimation value is set as the predetermined transmission data estimation value .   The transmission data includes predetermined known data, obtains an initial channel estimation value based on the reception data corresponding to the known data and the known data, and corresponds to the obtained initial channel estimation value and the known data. 2. The channel estimation and data detection method according to claim 1, wherein an initial transmission data estimated value is obtained based on received data, and the obtained initial transmission data estimated value is set as the predetermined transmission data estimated value. The initial transmission data estimate claim 1 or claim 2 channel estimation and data detection method, wherein the calculating the maximum likelihood detection. The channel estimate, channel estimation and data detection method according to any one of claims 1 to 3, characterized in that determined by the least mean square error method.

Images