An efficient channel equalization method for direct sequence spread spectrum underwater acoustic communication

Abstract

The present invention discloses an efficient channel equalization method for direct sequence spread spectrum underwater acoustic communication. For direct sequence spread spectrum underwater acoustic communication systems, the method mainly solves the problems of slow convergence, high complexity and large pilot overhead of adaptive channel equalization. The implementation steps are: first, determine the relevant parameters of the hypothesis feedback equalizer; second, train the equalizer coefficients based on the pilot symbols using a proportional adjustment type adaptive filtering algorithm and data reuse technology; third, filter unimportant equalizer coefficients based on the threshold method to obtain a partial coefficient structure equalizer; finally, use the partial coefficient structure equalizer to estimate the data symbols. The present invention makes full use of the inherent sparse characteristics of the underwater acoustic channel equalization system, effectively solves the problems of high complexity and slow tracking speed of direct sequence spread spectrum underwater acoustic communication channel equalization, and has high practical application value.

Description

Technical Field

[0001] The present invention belongs to the technical field of underwater acoustic communication, and in particular relates to an efficient channel equalization method for direct sequence spread spectrum underwater acoustic communication. Background Art

[0002] Direct Sequence Spread Spectrum (DSSS) communication technology reduces the system's signal-to-noise ratio requirements by spreading the energy of the transmitted signal over a wide spectrum, improving the stealth of communication, and has received extensive attention in the field of underwater acoustic communication. The complexity of underwater acoustic communication channels still makes it difficult to achieve reliable direct sequence spread spectrum underwater communication, and receiver algorithm design remains the focus of current research.

[0003] At present, traditional direct sequence spread spectrum underwater acoustic communication receiver technologies include classic RAKE receiving technology, time reversal (TR) technology, and a combination of the above two.

[0004] The performance of RAKE receiving technology depends on the accurate tracking of the channel state, which is difficult to guarantee under fast-changing underwater acoustic channel conditions. On the other hand, it also depends on the length of the spread spectrum sequence. If the sequence length is insufficient, it will lead to non-negligible inter-path interference, which will also affect the performance. The basic idea of ​​time reversal technology is to focus multipath signals into single-path signals to eliminate path interference, but its performance depends on the accuracy of channel state information acquisition and the number of receiving array elements. It is generally impossible to completely eliminate inter-path interference, so it can only be applied to specific scenarios.

[0005] In addition to the above-mentioned schemes, equalization-based receiving technology has also received attention. Considering that direct sequence spread spectrum communication generally works under low signal-to-noise ratio conditions, the traditional decision feedback equalization technology is prone to error propagation. Therefore, the hypothesis feedback equalization (HFE) technology is proposed in the literature to effectively solve the problem of error propagation. Nevertheless, the equalization-based receiving scheme is more complex than RAKE reception and time reversal reception, especially the use of hypothesis feedback equalization will further increase the amount of calculation, so it has been seldom used in actual systems for a long time. Summary of the invention

[0006] Technical problem: In order to solve the problems of high complexity and slow convergence of the existing adaptive hypothesis feedback equalization scheme for direct sequence spread spectrum underwater acoustic communication, the present invention utilizes the sparsity of the equalizer and proposes an efficient channel equalization method for direct sequence spread spectrum underwater acoustic communication, which improves the convergence speed while effectively reducing the equalization complexity.

[0007] Technical solution: An efficient channel equalization method for direct sequence spread spectrum underwater acoustic communication, specifically comprising the following steps:

[0008] (1) Obtain the chip-level received data sequence y(k), k = 1, 2, ..., K, where the first K p *L corresponding pilot symbols, L is the spreading code length;

[0009] (2) Determine the length of the full coefficient hypothesis feedback equalizer, including the length L of the forward equalizer w w and the length L of the backward equalizer f f , modulation symbol constellation size Q;

[0010] (3) Determine the proportional adjustment type sparse adaptive filtering algorithm and the number of data reuses N dr , and the partial coefficient hypothesis feedback equalizer coefficient selection threshold λ w and λ f ;

[0011] (4) Based on the pilot data, the coefficients of the hypothesis feedback equalizer are trained using an adaptive filtering algorithm and data reuse technology;

[0012] (5) Based on the full coefficient hypothesis feedback equalizer obtained in the training phase, a threshold is applied to determine the retained coefficients, and the remaining coefficients with small contributions are discarded to obtain a partial coefficient adaptive hypothesis feedback equalizer;

[0013] (6) Based on the partial coefficient adaptive hypothesis feedback equalizer, the data symbols are equalized at the chip level, and the chip-level equalized symbols are despread to obtain the final symbol estimation.

[0014] In step (4), the full coefficient hypothesis feedback equalizer is trained based on the pilot data using the improved proportional normalized least mean square adaptive algorithm. Since the training uses pilot symbols and the decision feedback symbol vector is completely known, only one equalizer needs to be trained, that is, the hypothesis feedback equalizer degenerates into a traditional decision feedback equalizer. For the convenience of discussion, we do not distinguish which transmission symbol a certain chip-level symbol belongs to. The update formulas for the forward part and the reverse part of the equalizer are respectively

[0015]

[0016]

[0017] Where y(k)=[y(k+L w -1),...,y(k)] t is the received signal vector, d(k)=[d(k-1),...,d(kL f )] tis the chip-level decision symbol vector, is the chip-level symbol estimation error, * indicates conjugation, is the chip-level equalization symbol, μ is the update step size, and δ is a very small positive number to prevent meaningless division. For phase estimation, it can be obtained by phase-locked loop (PLL) technology, that is,

[0018]

[0019] in K f1 and K f2 is the PLL parameter. w (k) and G f (k) is a diagonal matrix, whose diagonal elements are

[0020]

[0021]

[0022] where w l (k) and f l (k) are the lth elements of w(k) and f(k), respectively. ε is a small positive number, |α|≤1. The above training process is repeated N times. dr times, and obtain the final full-coefficient equalizer vector and

[0023] In step (5), based on the threshold λ w and λ f , we obtain the partial coefficient index set as follows: First, we define two empty sets Λ w and Λ f , and then do the following: for 1≤l≤L w ,if Then w =Λ w ∪l; similarly for 1≤l≤L f ,if Then f =Λ f ∪l, the final set of partial coefficient indexes is and Their sizes are L' w and L' f Based on Λ w and Λ f , obtain the partial coefficient equalizer vector and

[0024] In the step (6), the block diagram of the hypothesis feedback equalizer for estimating data symbols is as follows: Figure 3 As shown in the figure, we can see that in order to obtain the estimate of the lth symbol in the nth symbol period, the equalizer has Q parallel branches, and the i-th branch corresponds to the i-th hypothesis H of the n-th symbol. i :s(n)=a i , where a i is the i-th symbol of the constellation diagram. The initialization values ​​of the forward and backward equalizers of each branch are set to the partial coefficient equalizer w obtained in step 5 s and f s , and after that, each time a symbol is equalized, they are reset to the equalizer corresponding to the best hypothesis branch. i Under these conditions, the following chip-level equalization symbols are obtained:

[0025]

[0026] where w s,i (n,l) and f s,i (n,l) are the forward and reverse equalizer vectors corresponding to the lth symbol in the nth symbol period, is the corresponding phase estimate, H i After all L code elements in the nth symbol are equalized, the code element equalization symbols obtained under different hypotheses are despread. Hypothesis H i The corresponding despread symbol is estimated to be in is the spreading code. Finally, the best decision index is obtained based on the despread symbol estimation. The final symbol estimate is

[0027] Beneficial effects: The present invention utilizes the inherent sparse characteristics of the underwater acoustic channel equalizer to reduce the number of effective taps of the equalizer, which on the one hand improves the convergence speed of the equalizer and saves pilot symbol expenses, and on the other hand reduces the computational complexity. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 is a flow chart of the method of the present invention;

[0029] Figure 2 This is a block diagram of a direct sequence spread spectrum underwater acoustic communication system;

[0030] Figure 3 The structure diagram of the adaptive hypothesis feedback equalizer;

[0031] Figure 4 is the channel impulse response estimated based on experimental data;

[0032] Figure 5 : The equalized symbol constellation diagram is obtained under the condition of a signal-to-noise ratio of about 4.5dB, where (a) and (b) are the code chip symbol constellation diagram and symbol constellation diagram obtained by the full coefficient hypothesis feedback equalizer, respectively, and (c) and (d) are the code chip symbol constellation diagram and symbol constellation diagram obtained by the partial coefficient hypothesis feedback equalizer of the present invention, respectively.

[0033] Figure 6 Comparison diagram of the symbol mean square error of the full coefficient and partial coefficient equalizer;

[0034] Figure 7 This is a comparison chart of the symbol error rate (SER) of 17 receiving channels using full coefficient and partial coefficient hypothesis feedback equalizers under the experimental conditions. DETAILED DESCRIPTION

[0035] The present invention will be further explained below in conjunction with the accompanying drawings.

[0036] According to Figure 2 The block diagram of the underwater acoustic communication system is shown in the experiment. The number of transmitting hydrophones is N t is 1, the number of receiving hydrophones is N r is 17, the carrier frequency f c The frequency is 5.5kHz, QPSK modulation is used, and the spreading code is an m-sequence of length L = 7. Each transmitted data block includes 1300 symbols, of which the number of pilot symbols is K. p is 276, the number of data symbols K d is 1024, symbol period T s The sampling rate at the receiving end is 2.7ms. s The underwater acoustic channel impulse response estimated based on experimental data is as follows Figure 4 As shown, it can be seen that the underwater acoustic channel in this experimental environment has obvious 2-path, and the chip-level channel length is about 200.

[0037] Firstly, the improved proportional normalized least mean square adaptive filtering algorithm is used to train the full coefficient equalizer. The parameters in the training stage are set as follows: α = -0.5, μ = 0.1, ε = 1, δ = 0.00001, and the amplitude threshold of the partial coefficient equalizer λ w and λ f Both are 0.0067, and the second-order phase-locked loop parameters are: K f1 =0.007, K f2 =K f1 / 10. According to the channel length, set the full coefficient forward equalizer length L w is 400, the length of the reverse equalizer is L f is 200. Number of data reuses N dris 8. Threshold is applied to the trained full coefficient equalizer to obtain a partial coefficient equalizer, whose forward and backward equalizer lengths are L' w =85 and L' f =188, from which the complexity ratio of the partial coefficient equalizer and the full coefficient equalizer can be calculated as

[0038]

[0039] Therefore, partial coefficient equalization can save nearly 43% of the computational effort.

[0040] The data symbols are equalized by full coefficient hypothesis feedback equalization and partial coefficient hypothesis feedback equalization. In the data symbol equalization stage, the adaptive equalizer parameters are set to μ = 0.15, ε = 0.1, and the phase-locked loop parameters are set to K f1 =0.006, K f2 =K f1 / 10, and the rest of the parameters are the same as in the training phase.

[0041] Figure 5 The chip-level and symbol-level equalization symbol constellation diagrams corresponding to the two equalization methods are shown. The chip-level signal-to-noise ratio of the received data is about 4.5dB, N dr =4. Figure 5 (a) and (c) are the chip-level equalization symbol constellation diagrams of full coefficient and partial coefficient hypothesis feedback equalization, respectively. Figure 5 (b) and (d) are the symbol-level constellation diagrams of the full coefficient and partial coefficient hypothesis feedback equalizers, respectively. Thanks to the spreading gain, the despread equalization symbols are more accurate than the chip-level equalization symbols. From the constellation diagram, the performance of the two equalizations is comparable.

[0042] Figure 6 This is a comparison chart of the mean square error of the equalized symbols of the two equalizers. It can be seen that when the number of data reuses is greater than 3, the partial coefficient equalization obtains a lower mean square error.

[0043] Figure 7 The figure is a comparison of the symbol error rate (SER) of the two equalizers on 17 receiving channels. It can be seen that the overall performance of the partial coefficient equalization scheme is comparable to that of the full coefficient equalization scheme, and there is an additional performance gain on individual channels.

[0044] In summary, the present invention can reduce the computational complexity while ensuring good equalization performance.

[0045] The above is only a preferred embodiment of the present invention. It should be noted 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

[0046] It is the protection scope of the present invention.

Claims

1. An efficient channel equalization method for direct sequence spread spectrum underwater acoustic communication, characterized in that: The steps include: (1) Obtain the chip-level received data sequence y(k), k = 1, 2, ..., K, where the first K p *L corresponding pilot symbols, L is the spreading code length; (2) Determine the length of the full coefficient hypothesis feedback equalizer, including the length L of the forward equalizer w w and the length L of the backward equalizer f f , modulation symbol constellation size Q; (3) Determine the proportional adjustment type sparse adaptive filtering algorithm and the number of data reuses N dr , and the partial coefficient hypothesis feedback equalizer coefficient selection threshold λ w and λ f ; (4) Based on the pilot data, the coefficients of the hypothesis feedback equalizer are trained using an adaptive filtering algorithm and data reuse technology; (5) Based on the full coefficient hypothesis feedback equalizer obtained in the training phase, a threshold is applied to determine the retained coefficients, and the remaining coefficients with small contributions are discarded to obtain a partial coefficient adaptive hypothesis feedback equalizer; (6) Based on a partial coefficient adaptive hypothesis feedback equalizer, the data symbols are equalized at the chip level, and the chip-level equalized symbols are despread to obtain the final symbol estimation; In step (4), the update formulas for the forward part and the reverse part of the equalizer are Where y(k)=[y(k+L w -1),K,y(k)] t is the received signal vector, d(k)=[d(k-1),K,d(kL f )] t is the chip-level decision symbol vector, is the chip-level symbol estimation error, * represents conjugation, is the chip-level equalization symbol, μ is the update step size, and δ is a positive number; is the phase estimation, obtained by PLL technology, that is, in K f1 and K f2 is the PLL parameter; G w (k) and G f (k) is a diagonal matrix, whose diagonal elements are Among them, w l (k) and f l (k) is the lth element of w(k) and f(k), ε is a positive number, |α|≤1; the above training is repeated N times dr times, and obtain the final full-coefficient equalizer vector and In step (5), based on the threshold λ w and λ f , we obtain the partial coefficient index set as follows: First, we define two empty sets Λ w and Λ f , and then do the following: for 1≤l≤L w ,if Then w =Λ w ∪l; similarly for 1≤l≤L f ,if Then f =Λ f ∪l, the final set of partial coefficient indexes is and Their sizes are L' w and L' f ; Based on Λ w and Λ f , obtain the partial coefficient equalizer vector and In step (6), to obtain an estimate of the lth symbol in the nth symbol period, the equalizer has Q parallel branches, and the i-th branch corresponds to the i-th hypothesis H of the n-th symbol. i :s(n)=a i , where a i is the i-th symbol of the constellation diagram; the initialization value of the forward and backward equalizers of each branch is set to the partial coefficient equalizer w s and f s , after which each symbol is equalized, they are reset to the equalizer corresponding to the best hypothesis branch; in hypothesis H i Under these conditions, the following chip-level equalization symbols are obtained: where w s,i (n,l) and f s,i (n,l) are the forward and reverse equalizer vectors corresponding to the lth symbol in the nth symbol period, is the corresponding phase estimate, H i The chip decision vector under the condition; when all L code symbols in the nth symbol are equalized, the code symbol equalization symbols obtained under different hypotheses are despread. Hypothesis H i The corresponding despread symbol is estimated to be in is the spreading code; finally, the best decision index is obtained based on the despread symbol estimation The final symbol estimate is

Images