MIMO-FBMC Channel Estimation Method Based on Distributed Compressed Sensing

Abstract

The present invention discloses a MIMO-FBMC channel estimation method based on distributed compressive sensing. The present invention utilizes the sparse correlation between MIMO channels, characterizes it as a joint sparse model, and studies the channel estimation method for MIMO-FBMC systems based on distributed compressive sensing; optimizes the selection of the number of relevant atoms for inner product operation through a weak selection threshold, thereby reducing the amount of inner product calculation, and combines sparse adaptivity to achieve the reconstruction of sparse signals. The simulation results confirm that, compared with the traditional OMP method and other classical distributed compressive sensing methods, the present invention has better channel estimation performance.

Description

Technical Field

[0001] The present invention belongs to the field of communication technologies, and more particularly, to a MIMO-FBMC channel estimation method based on distributed compressive sensing. Background Art

[0002] Due to its technical characteristics and advantages, the Multiple-Input Multiple-Output (MIMO) technology, after being combined with the Orthogonal Frequency Division Multiplexing (OFDM) technology, has been widely applied in 4G and 5G networks and has become a key technology for the new generation of mobile communications. Although the traditional OFDM can achieve relatively low complexity and high bandwidth efficiency, when applied to more complex dynamic or multi-user networks in future mobile communication scenarios, it has two major problems: it is difficult to achieve strict synchronization (Doppler effect in a mobile environment) and non-continuous frequency band transmission (severe spectrum leakage). The Filter Bank Multi-Carrier with Offset Quadrature Amplitude Modulation (FBMC / OQAM) technology can well solve the above problems by using a filter bank with good time-frequency focusing. First, the non-strict orthogonality of the FBMC / OQAM system enables it not to require a cyclic prefix, thus having higher spectrum efficiency. Second, the FBMC / OQAM system can flexibly control the interference between adjacent subcarriers and can make better use of fragmented spectrum resources. Finally, the FBMC / OQAM system processes channel estimation, synchronization, etc. separately on each subcarrier, making it more suitable for the communication uplink. Although the latest 3GPP organization has pointed out that the fifth-generation mobile communication system (5G) will adopt the sub-band filtered OFDM (F-OFDM) technology as the multi-carrier transmission scheme. However, due to the advantages of the FBMC / OQAM technology, scholars still maintain their research enthusiasm for the FBMC / OQAM technology, hoping to apply it to future mobile communications.

[0003] The development of mobile communication technology has always revolved around how to improve the transmission rate and quality within limited bandwidth. By combining the FBMC / OQAM technology with the MIMO technology, the communication system can possess the advantages of both technologies simultaneously, thus meeting the requirements of the development of the new generation of mobile communication technology. However, due to the inherent imaginary part interference in the FBMC / OQAM system, after combining with the spatial multiplexing MIMO technology, the system will also be affected by the interference between antennas. When performing channel estimation on the MIMO-FBMC (abbreviated as MIMO-FBMC) system, the existence of these interference terms will deteriorate the performance of traditional channel estimation methods, which seriously affects the channel estimation performance.

[0004] Currently, the research on channel estimation of the MIMO-FBMC system by domestic and foreign scholars mainly focuses on studying the corresponding channel estimation methods by combining the MIMO technology on the basis of pilot-assisted of the single-input single-output (SISO) FBMC / OQAM system. The traditional pilot channel estimation of the MIMO-FBMC system mainly includes two types of methods: interference approximation of pilots (IAM) and interference cancellation of pilots (ICM). Compared with the SISO-FBMC system, there is more imaginary part interference between pilot symbols in the MIMO-FBMC system, which results in low accuracy of pilot-based channel estimation in the MIMO system.

[0005] In recent research, some researchers have proposed to use the compressed sensing method for channel estimation of the FBMC system. Existing research has proposed a channel estimation method for the FBMC system based on Orthogonal Matching Pursuit (OMP); compared with the least squares (LS) estimation method of the traditional pilot structure, the compressed sensing method in FBMC can significantly improve the channel estimation performance. Existing research has proposed a channel estimation method for the FBMC system based on Tanimoto sparse weak selection regularization orthogonal matching pursuit; this method has better error code performance than the classical compressed sensing method. Researchers have proposed a compressed sensing channel estimation method based on the approximate message passing algorithm for the MIMO-FBMC system; existing research has proposed an effective sparse adaptive channel estimation method for the MIMO-FBMC system; both of the above two methods have better channel estimation performance than the traditional methods.

[0006] The theory of distributed compressed sensing points out that the joint reconstruction can be realized by using the inherent correlation of the sparse structures of multiple signals, which can greatly improve the reconstruction efficiency. Some research has found that the existing research on distributed compressed sensing channel estimation mainly focuses on the OFDM system, and the distributed compressed sensing channel estimation in the FBMC system is still in the exploratory research stage. Summary of the Invention

[0007] The object of the present invention is to provide a MIMO-FBMC channel estimation method based on distributed compressive sensing to make up for the deficiencies of the prior art.

[0008] The present invention first utilizes the correlation between MIMO channels, characterizes it as a joint sparse model, and transforms the channel estimation into a joint sparse signal reconstruction problem; then, optimizes and selects the number of relevant atoms for inner product operation through threshold priority, and combines sparse adaptivity to achieve the reconstruction of sparse signals.

[0009] To achieve the above object, the present invention is realized through the following specific technical solutions:

[0010] The MIMO-FBMC channel estimation method based on distributed compressive sensing includes the following steps:

[0011] S1: Collect MIMO-FBMC signals to obtain a signal reception matrix;

[0012] S2: Use the distributed compressive sensing reconstruction algorithm to transform the MIMO-FBMC channel estimation problem into solving the signal h j ;

[0013] S3: Use the threshold priority algorithm for specific operations, and finally obtain the signal h j , and complete the channel estimation.

[0014] Furthermore, in S2, the channel estimation of the MIMO-FBMC system is simplified to the superposition of multiple single-input single-output systems, then there is

[0015] Y j = Φ j h j + η j , and the signal h j can be recovered through the distributed compressive sensing reconstruction algorithm; where Y j is the measurement signal, Φ j is the observation matrix, and η j is the noise matrix.

[0016] Even further, for any K-sparse signal h j , Φ j must satisfy

[0017] where 0 < δ K < 1.

[0018] Furthermore, in S3, the threshold priority algorithm is specifically as follows:

[0019] S3-1: Set a threshold ε; if the reconstructed signal h jSatisfy ||h j，t -h j ||2 ≤ ε, then stop the iteration; otherwise, continue with S3-2;

[0020] S3-2: Calculate the inner products of all J residual vectors and the column vectors of the observation matrix; where Φ j，λ represents the λ-th column of the observation matrix;

[0021] S3-3: Set a threshold Th = 0.5 * max(abs(u λ )) and select the values in u λ that are greater than the threshold. Record the positions of the corresponding column vectors in Φ j in the index set λ t ;

[0022] S3-4: Update the support set Λ t = Λ t-1 ∪ λ t ; Update the residual vector r j，t , Calculate where r j，t = y j - Φ j h j，t ;

[0023] S3-5: Compare the updated residual with the residual from the previous iteration. If ||r j，t ||2 ≥ ||r j，t-1 ||2, update the step size stage = stage + 1, s = s × stage, and return to S3-2; otherwise, r j = r j，t , t = t + 1, and jump to S3-1.

[0024] Compared with the prior art, the advantages and beneficial effects of the present invention are as follows:

[0025] The present invention utilizes the sparse correlation between MIMO channels, characterizes it as a joint sparse model, and studies the channel estimation method for MIMO-FBMC systems based on distributed compressive sensing; optimizes the selection of relevant atoms for inner product operations through threshold priority, thereby reducing the amount of inner product calculation, and combines sparse adaptivity to achieve the reconstruction of sparse signals.

[0026] The present invention explores the channel estimation problem of MIMO-FBMC systems based on distributed compressive sensing, and proposes a sparse adaptive distributed compressive sensing channel estimation method based on threshold priority. The present invention optimizes the selection of relevant atoms for inner product operation through backtracking thought and threshold priority, and combines sparse adaptiveness to realize the reconstruction of sparse signals. The channel estimation performance of the proposed method is simulated and analyzed under two multipath fading channels. The simulation results confirm that, compared with the traditional OMP method and other classical distributed compressive sensing methods, the distributed compressive sensing sparse adaptive algorithm proposed in this paper has better channel estimation performance. Description of the Drawings

[0027] Figure 1 It is the PA channel parameter estimation diagram.

[0028] Figure 2 It is the comparison diagram of the mean square error performance of different algorithms under the PA channel.

[0029] Figure 3 It is the comparison diagram of the bit error rate performance of different algorithms under the PA channel.

[0030] Figure 4 It is the comparison diagram of the mean square error performance of different algorithms under the EPA channel.

[0031] Figure 5 It is the comparison diagram of the bit error rate performance of different algorithms under the EPA channel. Detailed Embodiments

[0032] The technical solution described in the present invention will be further described and illustrated below in conjunction with the embodiments and the drawings.

[0033] Embodiment 1;

[0034] A MIMO-FBMC channel estimation method based on distributed compressive sensing includes the following steps:

[0035] S1: Collect MIMO-FBMC signals and process the signals as:

[0036]

[0037] In the formula, is the MIMO channel matrix, is the equivalent transmission symbol vector form of the MIMO-FBMC signal, where,

[0038] S2: Simplify the measurement equation of the above MIMO system into the superposition of multiple single-input single-output systems, then there is

[0039]

[0040] In the formula,

[0041] Define the set Θ that contains all antennas, Θ = {β = (n t , n r ) | n t = 1,..., N t , n r = 1,..., N r}. Let Then the MIMO-FBMC channel estimation problem is transformed into solving the following equation,

[0042] Y β = Φ β h β + η β , β ∈ Θ (21)

[0043] Here, h β can be recovered through the distributed compressive sensing reconstruction algorithm. Among them, Y j is the measurement signal, Φ j is the observation matrix, and η j is the noise matrix; and for any K-sparse signal h j , Φ j must satisfy

[0044]

[0045] S3: Finally, use the threshold priority algorithm for channel estimation:

[0046] Input: measurement signal y j , observation matrix Φ j , where j = 1, 2,..., J, step size s;

[0047] Output: K-sparse reconstructed channel h j ;

[0048] Step 1: Initialize the iteration number t = 1; initialize the residual r j，0 = y j ; initialize the step size s = 1, stage = 1; initialize the index set

[0049] Step 2: Set the threshold ε, and ε = 10 -7 ; if the reconstructed signal h j satisfies ||h j，t - h j ||2 ≤ ε, then stop the iteration, otherwise continue to Step 3.

[0050] Step 3: Calculate the inner products of all J residual vectors with the column vectors of the observation matrix; where Φ j，λ represents the λ-th column of the observation matrix.

[0051] First, select the 2K maximum values in u λ ; then set a threshold Th = 0.5 * max(abs(u λ )); select the values in u λ that are greater than the threshold, and record the positions of the corresponding column vectors in Φ j in the index set λ t .

[0052] Step 4: Update the support set Λ t = A t-1 ∪λ t .

[0053] Step 5: Update the residual vector r j，t , calculate where r j，t = y j - Φ j h j，t .

[0054] Step 6: Compare the updated residual with the residual of the previous iteration. If ||r j，t ||2 ≥ ||r j，t-1 ||2, update the step size stage = stage + 1, s = s × stage, and return to Step 3; otherwise, r j = r j，t , t = t + 1, and jump to Step 2.

[0055] Example 2:

[0056] This example simulates and compares the mean square error (MSE) and bit error rate (BER) performance of different channel estimation methods in the MIMO-FBMC system under time-frequency doubly selective channels. Four algorithms, namely the least squares method (LS), OMP, distributed OMP (DCS-OMP), and distributed sparse adaptive matching pursuit (DCS-SAMP), are selected for analysis and comparison with the present invention. The specific simulation parameters are shown in Table 1.

[0057] The specific channel parameters are as follows:

[0058] 4-path pedestrian channel (PA)

[0059] Multipath delay: [0 110 190 410] ns

[0060] Relative power: [0 -9.7 -19.2 -22.8] dB;

[0061] 7-path Extended Pedestrian A (EPA) channel

[0062] Multipath delay: [0 30 70 90 110 190 410] ns

[0063] Relative power: [0 -1 -2 -3 -8 -17.2 -20.8] dB.

[0064] Table 1 Simulation parameters

[0065]

[0066] First, the estimation accuracy of the multipath channel parameters of the present invention under the PA channel was analyzed. Figure 1 Shows the comparison of the original and estimated channel power delay profiles, where the present invention is the marked line with asterisks. Obviously, the present invention can accurately estimate the number of multipaths and the relative power of each path in the PA channel.

[0067] Then, the channel estimation performance of the present invention was analyzed and discussed under two multipath fading channels, PA and EPA.

[0068] Figure 2 and Figure 3 Gives the performance comparison of the mean square error and bit error rate of different algorithms in the 4-path PA channel. It should be noted that the initial channel estimation of all methods is based on the IAM pilot structure. As can be seen from the two figures, the present invention achieves the best results in terms of both mean square error performance and bit error rate performance, and the channel estimation performance of the compressive sensing algorithm is significantly better than that of the traditional LS method.

[0069] From Figure 2 it can be seen that the present invention has the best mean square error performance. At the same time, by comparing the two methods of OMP and DCS-OMP, it can be concluded that the mean square error performance of the distributed compressive sensing method is better than that of the traditional compressive sensing method, and the simulation results are consistent with the conclusions in the introduction analysis.

[0070] From Figure 3 it can be seen that the present invention has the optimal bit error rate. Compared with the traditional distributed sparse adaptive algorithm, the bit error performance of the present invention has been significantly improved. At the benchmark bit error rate of 10 -2 compared with the OMP algorithm and the DCS-OMP algorithm, the present invention improves the bit error performance by 4 dB and 1 dB respectively; compared with the DCS-SAMP algorithm, it improves the bit error performance by 3.9 dB.

[0071] Figure 4 and Figure 5The mean square error and bit error rate performance curves of different algorithms in the 7-path EPA channel environment are given. The results show that the performance curve trends of all algorithms in the EPA channel are the same as those in the PA channel, that is, the present invention still has the optimal channel estimation performance in the EPA channel.

[0072] Compare Figure 4 with Figure 2 and it is found that due to the increase in the number of channel multipaths, the performance curves of the five algorithms have a more obvious floor effect.

[0073] Similarly, in Figure 5 , the bit error rate performance curves of different algorithms in the EPA channel also tend more to the floor effect. This shows that the increase in the number of channel multipaths will affect the accuracy of the algorithm channel estimation. As the signal-to-noise ratio increases, the bit error performance of the OMP algorithm gets better and better, while the other algorithms are more approaching a flat line. The bit error performance of the OMP algorithm at high signal-to-noise ratio is close to the bit error performance of the present invention. This is because the traditional compressive sensing method processes each channel separately, while the distributed compressive sensing method processes jointly, and the performance of the joint processing will be affected by the number of this joint.

[0074] On the basis of the above embodiments, the present invention continues to describe in detail the technical features involved therein and the functions and roles played by these technical features in the present invention, so as to help those skilled in the art fully understand the technical solution of the present invention and reproduce it.

[0075] Finally, although this specification is described according to the embodiments, not every embodiment only contains an independent technical solution. This narrative way of the specification is only for clarity. Those skilled in the art should regard the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. A MIMO-FBMC channel estimation method based on distributed compressive sensing, characterized in that, The channel estimation method includes the following steps: S1: Collect MIMO-FBMC signals to obtain a signal reception matrix; S2: Use the distributed compressive sensing reconstruction algorithm to transform the MIMO-FBMC channel estimation problem into solving the signal h j ; S3: Perform specific operations using the threshold optimization algorithm, and finally obtain the signal h j , and complete channel estimation.

2. The MIMO-FBMC channel estimation method according to claim 1, wherein In S2, the channel estimation of the MIMO-FBMC system is simplified to the superposition of multiple single-input single-output systems, so there is Y j = Φ j h j + η j The signal h can be recovered by the distributed compressive sensing reconstruction algorithm j ; where Y j is the measurement signal, Φ j is the observation matrix, and η j is the noise matrix.

3. The MIMO-FBMC channel estimation method according to claim 2, wherein For any K-sparse signal h j , Φ j must satisfy 4. The MIMO-FBMC channel estimation method according to claim 1, characterized in that, In S3, the threshold optimization algorithm is specifically: S3-1: Set the threshold ε; if the reconstructed signal h j satisfies ||h j,t -h j ||2 ≤ ε, stop the iteration; otherwise, continue with S3-2; S3-2: Calculate the inner products of all J residual vectors and the column vectors of the observation matrix; where Φ j,λ represents the λ-th column of the observation matrix; S3-3: Set a threshold Th = 0.5 * max(abs(u λ )) and select the values in u λ that are greater than the threshold. Record the positions of the column vectors in Φ j corresponding to these values in the index set λ t ; S3-4: Update the support set Λ t = Λ t-1 ∪ λ t ; Update the residual vector r j,t , Calculate where r j,t = y j - Φ j h j,t ; S3-5: Compare the updated residual with the residual of the previous iteration. If ||r j,t ||2 ≥ ||r j,t-1 ||2, update the step size stage = stage + 1, s = s × stage, and return to S3-2; otherwise, r j = r j,t , t = t + 1, and jump to S3-1.

Images