A low-complexity sparse adaptive turbo equalization method for underwater acoustic communication

Abstract

The present invention discloses a low-complexity underwater acoustic communication sparse adaptive turbo equalization method. For a single-carrier underwater acoustic communication system, the method mainly solves the problems of slow convergence and high computational complexity of the existing minimum symbol error rate type adaptive turbo equalization. The implementation process is divided into two stages: 1) In the training stage, a full-tap equalizer is used, and the tap coefficients of the equalizer are brought to a convergence state by using a minimum symbol error rate type adaptive algorithm combined with data multiplexing technology, and the coefficients are selected based on the average output vector of the equalizer to obtain a partial tap equalizer; 2) In the data detection stage, a partial tap equalizer is used to implement turbo equalization. The present invention applies the partial tap equalizer to the minimum symbol error rate type adaptive turbo equalization, reduces its computational complexity while ensuring the turbo equalization performance, is simple to implement, and has a high practical application value.

Description

Technical Field

[0001] The present invention belongs to the technical field of single-carrier underwater acoustic communication, and in particular relates to a low-complexity underwater acoustic communication adaptive turbo equalization method. Background Art

[0002] In single-carrier underwater acoustic communication, the strong multipath effect of the underwater acoustic channel causes serious inter-symbol interference at the receiving end, which poses a challenge to reliable data transmission. Equalization technology can effectively suppress inter-symbol interference and is one of the keys to the receiving end. There are currently two main types of equalization methods: equalization based on channel estimation and direct adaptive equalization. Compared with the former, direct adaptive equalization has low complexity and environmental tracking capabilities, and has become the mainstream equalization scheme for single-carrier underwater acoustic communication. Giving the equalizer soft input and soft output capabilities can further realize turbo equalization, that is, iterative interaction of soft information between the equalizer and the decoder to gradually improve the performance of the receiver.

[0003] Adaptive turbo equalization (ATEQ) technology based on adaptive equalization has been widely used in the field of single-carrier underwater acoustic communication. Representative schemes include adaptive turbo equalization schemes based on least mean squares (LMS) and recursive least squares (RLS) adaptive algorithms and adaptive turbo equalization schemes based on minimum symbol error rate (MSER) ​​adaptive algorithms. Simulation and measured underwater acoustic communication results confirm the superiority of adaptive turbo equalization over traditional one-time equalization decoding schemes. Nevertheless, adaptive turbo equalization is more complex than traditional equalization due to the use of multiple iterative processing. In particular, with the continuous improvement of the requirements for underwater acoustic communication transmission distance and transmission rate, the number of taps of the equalizer increases accordingly, reaching hundreds or even thousands, which further increases the complexity of turbo equalization.

[0004] At present, some complexity reduction schemes have been proposed for LMS and RLS adaptive turbo equalization, such as reducing the number of effective taps by utilizing the sparsity of the equalizer. However, there is still a lack of relevant research on MSER adaptive turbo equalization. Summary of the invention

[0005] The present invention proposes a low-complexity underwater acoustic communication sparse adaptive turbo equalization method to solve the problem of high complexity of a minimum symbol error rate class adaptive turbo equalizer under long multipath and high rate conditions.

[0006] In order to solve the above technical problems, the specific technical solutions of the present invention are as follows:

[0007] A low-complexity underwater acoustic communication sparse adaptive turbo equalization method comprises the following steps:

[0008] (1) Determine the number of transmitter transducers N. The length of a transmitted symbol block on the nth transmitter transducer is K, including K p Pilot Symbols and K d Data Symbols

[0009] (2) Determine the number of hydrophones at the receiving end, M. The receiving block on the mth hydrophone is denoted as y m =[y m,1 ,…,y m,K+L-1 ] T , where L represents the length of the underwater acoustic channel;

[0010] (3) Determine the equalizer parameters, including the feedforward (FF) equalizer parameters K1, K2 (thus determining the full-tap FF equalizer length L w =M(K1+K2+1)) and soft interference canceller (SIC) equalizer parameters K3, K4 (thus determining the full-tap SIC equalizer length L f =N(K3+K4+1)-1), and their corresponding adaptive update step sizes μ w and μ f , the tap selection threshold γ w ,γ f , determine the second-order phase-locked loop parameter K f1 and K f2 , Data Reuse (DR) times N dr , number of turbo iterations N iter , initialize the equalizer and phase-locked loop;

[0011] (4) Based on the pilot sequence, the full-tap equalizer is trained using the minimum symbol error rate (MSER) ​​adaptive filtering algorithm combined with data multiplexing technology;

[0012] (5) based on the average output vector of the equalizer, respectively select taps for the FF and SIC equalizers trained in step (4) to obtain partial tap equalizers, and retrain the partial tap equalizers based on the pilot sequence;

[0013] (6) using the partial tap equalizer obtained in step (5) as an initial value, based on the received signal and the prior soft information from the channel decoder, detecting the data symbols and generating extrinsic soft information;

[0014] (7) The channel decoder decodes based on the external soft information from the equalizer. If the number of turbo iterations has reached N iter Then the decoder outputs the information bit hard decision, otherwise the decoder generates the coded bit external verification soft information and feeds it back to the equalizer, repeating steps (4)-(7).

[0015] In the step (1), on the nth transmission branch, the information bit stream is sequentially coded, interleaved, and symbol mapped; every q interleaved bits is mapped to a data symbol x n,k , x n,k Taken from the constellation point set Each constellation point α i With a certain bit sequence s i =[s i,1 ,s i,2 ,…,s i,q ],s i,j ∈{0,1} corresponds.

[0016] In step (2), the received signal at the kth time on the mth (1≤m≤M)th hydrophone is expressed as follows:

[0017]

[0018] where x n,k-l represents the transmission symbol of the nth transducer, h n,m,k (l) and φ n,m,k denote the lth tap and phase rotation of the subchannel between the nth transducer and the mth hydrophone, respectively, n m,k represents additive noise.

[0019] In step (4), based on the input received data and pilot symbols, the estimation of the kth transmitted symbol of the nth transducer output by the full-tap MSER type adaptive equalizer is as follows:

[0020]

[0021] in represents the received signal vector after phase compensation, The part beyond the actual received sequence is padded with zeros. is the decision symbol vector, where

[0022]

[0023] During the training phase, Known pilot symbols can be used; w n,k and f n,kare the FF equalizer coefficient vector and the SIC equalizer coefficient vector corresponding to the nth transducer, which are updated using the MSER-type adaptive algorithm; the phase Tracking is done through a second-order phase-locked loop. The tracking formula is as follows

[0024]

[0025] in, Indicates taking the imaginary part; to ensure convergence, according to Figure 2 The workflow of the partial tap adaptive turbo equalizer (PT-ATEQ) shown in FIG. 1 is a flowchart of a partial tap adaptive turbo equalizer (PT-ATEQ) in which the equalizer training process is repeated N times on the pilot signal block. dr After the training, we get the full-tap FF equalizer. and SIC equalizer

[0026] In the step (5), based on the average output vector of the equalizer, the full-tap equalizer and The coefficient selection is carried out as follows: First, calculate the time period [K P -L p +1,K p ]Internal full tap FF equalizer and SIC equalizer The average output vector The expression is as follows

[0027]

[0028]

[0029] The symbol ⊙ represents the Hadamard product; then define the partial tap support set Initialize them to empty sets and update them as follows

[0030]

[0031]

[0032] in and Respectively and The infinite norm of Then we can get the partial tap equalizer: Their lengths are L' w and L' f , it is obvious that L' w <L w ,L'f <L f ; Finally, based on the pilot data, the tap equalizer and Retrain according to step (4).

[0033] In step (6), based on the received signal and the prior soft information from the decoder, a partial tap MSER type adaptive turbo equalizer is used to detect the data symbol, and the symbol estimation obtained is as follows:

[0034]

[0035] in, Partial tap equalizer and The MSER-type adaptive algorithm is still used for updating; for the first round of turbo iteration, since there is no prior information from the decoder available at the equalizer input, the hard decision symbol For the t≥2th turbo iteration, the following a posteriori soft decision symbols are used:

[0036]

[0037] The above conditional probability Obey the following Gaussian distribution

[0038]

[0039] in, for Hard judgment

[0040] In formula 5, the prior probability P(x n,k =α i ) is calculated as

[0041]

[0042] in The bit log-likelihood ratio prior information provided to the channel decoder, Constraints can be Formula 5 not only uses the prior P(x n,k =α i ), and also uses the previous round of estimated symbols The conditional probability of the soft decision symbol is improved, thereby improving the reliability of the soft decision symbol and reducing error propagation; after completing the detection of the entire data symbol block, the equalizer generates the corresponding external soft information As the input of the channel decoder, the a posteriori information The calculation is as follows

[0043]

[0044] Beneficial effects: The present invention adopts data multiplexing technology to effectively improve the training quality of the full-tap minimum symbol error rate type adaptive equalizer, uses the average output vector of the equalizer as the basis for selecting important taps of the equalizer, and comprehensively considers the contribution of the equalizer's own structure and its input to symbol estimation. The partial tap equalizer obtained thereby has not only low complexity but also high robustness. The present invention effectively reduces the complexity of the minimum symbol error rate type adaptive turbo equalization without affecting its performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 is the estimated channel impulse response (CIR) amplitude diagram;

[0046] Figure 2 It is a working flow diagram of a partial tap adaptive turbo equalizer;

[0047] Figure 3 Comparison of FF equalizer taps between improved proportional normalized minimum symbol error rate (IPNMSER) ​​partial tap adaptive turbo equalizer PT-IPNMSER-ATEQ and full tap adaptive turbo equalizer IPNMSER-ATEQ;

[0048] Figure 4 The comparison chart of the number of error free blocks (NEFB) between PT-IPNMSER-ATEQ and IPNMSER-ATEQ is shown in Figure 1.

[0049] Figure 5 The following is a comparison chart of the uncoded bit error rate (UBER) of PT-IPNMSER-ATEQ and IPNMSER-ATEQ after one and three rounds of turbo iterations. DETAILED DESCRIPTION

[0050] The present invention will be further explained below in conjunction with the accompanying drawings.

[0051] The experimental data was used to verify the method of the present invention. The experiment was a 2×4 multiple-input multiple-output MIMO single-carrier underwater acoustic communication, that is, N=2, M=4. The channel coding was a low-density parity check code LDPC with a code rate of 1 / 2, a transmission distance of 2km, a carrier frequency of 17kHz, a symbol period of 0.2ms, and a modulation method of QPSK. The experiment performed a total of 5 burst transmissions, each of which transmitted 50 blocks of length K=1024, with the first K of each block being p = 200 pilot symbols, followed by K d =824 are data symbols, and the signal-to-noise ratio at the receiving end is about 10dB. Figure 1 The estimated channel impulse response amplitude diagram shows that the channel has a certain sparsity and the delay spread is about 70 symbol periods.

[0052] use Figure 2 The adaptive turbo equalizer shown in FIG. 1 processes the measured received data. The length parameters of the FF and SIC equalizers are K1=40, K2=50 and K3=40, K4=120 respectively. For each transmission signal of 2×4 MIMO transmission, in the case of full tap, the length of the FF equalizer is L w =364, the length of the SIC equalizer is L f =319. The second-order phase-locked loop parameters are set to K f1 =0.07, K f2 =K f1 / 10. The improved proportional normalized minimum symbol error rate IPNMSER adaptive algorithm is used to track the equalizer coefficients. The step size parameter of IPNMSER-ATEQ is set to: μ w =μ f =0.45. Set two groups of partial tap equalization thresholds: (γ w =0.01,γ f =0.01) and (γ w =0.02,γ f =0.02), two partial tap adaptive turbo equalization schemes are obtained: PT-IPNMSER-ATEQ-I and PT-IPNMSER-ATEQ-II. The performance metrics include the number of error-free data blocks NEFB and the undecoded bit error rate UBER.

[0053] In order to demonstrate the beneficial effects of the present invention, PT-IPNMSER-ATEQ and IPNMSER-ATEQ are compared below. First, the structures and complexities of the equalizers corresponding to the two schemes are compared. Without loss of generality, the 24th block of the first burst is taken as an example. Figure 3The comparison results of the FF equalizer structure at the end of the block between IPNMSER-ATEQ and PT-IPNMSER-ATEQ-II after one turbo iteration are shown, where the tap values ​​that are actually discarded in PT-IPNMSER-ATEQ-II are set to zero. Obviously, the number of taps in PT-IPNMSER-ATEQ-II is less than half of that in the full-tap IPNMSER-ATEQ. After four rounds of turbo iterations, the number of taps in the PT-IPNMSER-ATEQ-II equalizer is further reduced, where the FF equalizer has L' w = 67 taps, SIC equalizer includes L' f =7 taps, a total of 74 taps. At this time, the complexity ratio of the partial tap equalizer and the full tap equalizer is

[0054]

[0055] Therefore, PT-IPNMSER-ATEQ-II saves more than 70% of the computational effort compared to IPNMSER-ATEQ.

[0056] Figure 4 The NEFB comparison diagram between IPNMSER-ATEQ and PT-IPNMSER-ATEQ-I / II, where the number of turbo iterations is 1. It can be seen that the performance of PT-IPNMSER-ATEQ-I is always better than that of IPNMSER-ATEQ, and the comparison between PT-IPNMSER-ATEQ-II and IPNMSER-ATEQ is similar to that of N dr For all balancing schemes, the performance increases with N. dr Increased, especially for PT-IPNMSER-ATEQ-I, when N dr =6, the number of error-free blocks detected exceeds 220, accounting for nearly 90% of the total 250 detected blocks.

[0057] Figure 5 The UBER comparison diagrams of PT-IPNMSER-ATEQ-I / II and IPNMSER-ATEQ after one and three turbo iterations are shown in Figure 2. Figure 4 In addition, the UBER corresponding to three turbo iterations is significantly better than that of one turbo iteration, which verifies the superiority of turbo equalization over traditional schemes.

[0058] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A low-complexity sparse adaptive turbo equalization method for underwater acoustic communication, characterized in that: The steps include: (1) Determine the number of transmitter transducers N. The length of a transmitted symbol block on the nth transmitter transducer is K, including K p Pilot Symbols and K d Data Symbols (2) Determine the number of hydrophones at the receiving end, M. The receiving block on the mth hydrophone is denoted as y m =[y m,1 ,…,y m,K+L-1 ] T , where L represents the length of the underwater acoustic channel; (3) Determine the equalizer parameters, including the feedforward FF equalizer parameters K1, K2 and the soft interference cancellation SIC equalizer parameters K3, K4, and their corresponding adaptive update step sizes μ w and μ f , the tap selection threshold γ w and γ f , determine the second-order phase-locked loop parameter K f1 and K f2 , data multiplexing DR times N dr , number of turbo iterations N iter , initialize the equalizer and phase-locked loop; (4) Based on the pilot sequence, the full-tap equalizer is trained using the minimum symbol error rate (MSER) ​​adaptive filtering algorithm combined with data multiplexing technology; (5) performing tap selection on the FF and SIC equalizers trained in step (4) based on the average output vector of the equalizer, obtaining partial tap equalizers, and retraining the partial tap equalizers based on the pilot sequence; (6) using the partial tap equalizer obtained in step (5) as an initial value, based on the received signal and the prior soft information from the channel decoder, detecting the data symbols and generating extrinsic soft information; (7) The channel decoder decodes based on the external soft information from the equalizer. If the number of turbo iterations has reached N iter Then the decoder outputs the information bit hard decision, otherwise the decoder generates the coded bit external verification soft information and feeds it back to the equalizer, repeating steps (4)-(7).

2. The low-complexity underwater acoustic communication sparse adaptive turbo equalization method according to claim 1, characterized in that: In step (1), on the nth transmission branch, the information bit stream is encoded, interleaved, and symbol mapped in sequence; every q interleaved bits is mapped to a data symbol x n,k , x n,k Taken from the constellation point set Each constellation point α i With a certain bit sequence s i =[s i,1 ,s i,2 ,…,s i,q ],s i,j ∈{0,1} corresponds.

3. The low-complexity underwater acoustic communication sparse adaptive turbo equalization method according to claim 1, characterized in that: In step (2), the received signal at the mth hydrophone at the kth moment is expressed as follows: where x n,k-l represents the transmission symbol of the nth transducer, h n,m,k (l) and φ n,m,k denote the lth tap and phase rotation of the subchannel between the nth transducer and the mth hydrophone, respectively, n m,k represents additive noise.

4. The low-complexity underwater acoustic communication sparse adaptive turbo equalization method according to claim 1, characterized in that: In step (4), based on the input received data and pilot symbols, the estimation of the kth transmitted symbol of the nth transducer output by the full-tap MSER-type adaptive equalizer is as follows: in represents the received signal vector after phase compensation, The part beyond the actual received sequence is padded with zeros. is the decision symbol vector, where During the training phase, Known pilot symbols can be used; w n,k and f n,k are the FF equalizer coefficient vector and the SIC equalizer coefficient vector corresponding to the nth transducer, which are updated using the MSER-type adaptive algorithm; the phase Tracking is done by a second-order phase-locked loop; the equalizer training process is repeated N times on the pilot signal block. dr After the training, we get the full-tap FF equalizer. and SIC equalizer 5. The low-complexity underwater acoustic communication sparse adaptive turbo equalization method according to claim 1, characterized in that: In step (5), the full-tap equalizer is calibrated based on the average output vector of the equalizer. and The coefficient selection is carried out as follows: First, calculate the time period [K p -L p +1,K p ]Internal full tap FF equalizer and SIC equalizer The average output vector The expression is as follows The symbol ⊙ represents the Hadamard product; then define the partial tap support set Initialize them to empty sets and update them as follows in and Respectively and The infinite norm of L w and L f are the lengths of the full-tap FF equalizer and the SIC equalizer respectively; determine Then we get the partial tap equalizer: Their lengths are L' w and L' f ; Finally, based on the pilot data, the tap equalizer and Retrain according to step (4).

6. The low-complexity underwater acoustic communication sparse adaptive turbo equalization method according to claim 1, characterized in that: In step (6), based on the received signal and the prior soft information from the decoder, a partial tap MSER type adaptive equalizer is used to detect the data symbols, and the symbol estimation is as follows: in, Partial tap equalizer and Adopt MSER type adaptive algorithm to update; for the first round of turbo iteration, use hard decision symbol For the t≥2th turbo iteration, the following a posteriori soft decision symbols are used: The above conditional probability Obey the following Gaussian distribution in, for The hard decision of formula 4 is the prior probability P(x n,k =α i ) is calculated as in The bit log-likelihood ratio prior information provided to the channel decoder, By constraints to seek; After completing the detection of the entire data symbol block, the equalizer generates the corresponding external verification soft information As the input of the channel decoder, the a posteriori information The calculation is as follows

Images