阅读说明：本技术 一种复数域水声信道自适应均衡方法 (Complex domain underwater acoustic channel self-adaptive equalization method ) 是由 殷敬伟 田亚男 刘清宇 韩笑 葛威 生雪莉 于 2019-09-18 设计创作，主要内容包括：本发明提供的是一种复数域水声信道自适应均衡方法。(1)接收端将通带信号解调为基带复信号作为均衡器输入；(2)基带复信号与均衡器系数卷积得到均衡器的输出；(3)计算期望信号与均衡器输出之间的误差；(4)利用均衡误差定义新的代价函数,按照CAP-LMS/F算法,均衡器根据每个系数的大小施加不同约束自适应更新均衡器抽头系数。本发明将原有的适用于处理实信号的LMS/F算法拓展到复数域,以便处理基带上的水声复信号；每次迭代时,对均衡器的每一个抽头系数自适应地分配稀疏惩罚项,加快小系数收敛速度的同时减小大系数的收敛误差,提高均衡性能。(The invention provides a complex domain underwater acoustic channel self-adaptive equalization method. (1) The receiving end demodulates the passband signal into a baseband complex signal as an equalizer input; (2) convolving the baseband complex signal with the coefficient of the equalizer to obtain the output of the equalizer; (3) calculating an error between the desired signal and the equalizer output; (4) and defining a new cost function by using the equalization error, and applying different constraints to the equalizer to adaptively update the tap coefficients of the equalizer according to the magnitude of each coefficient according to the CAP-LMS/F algorithm. The invention expands the original LMS/F algorithm suitable for processing real signals to a complex field so as to process underwater acoustic complex signals on a baseband; and in each iteration, each tap coefficient of the equalizer is adaptively distributed with a sparse penalty item, so that the convergence speed of a small coefficient is accelerated, the convergence error of a large coefficient is reduced, and the equalization performance is improved.)

1. A complex domain underwater acoustic channel self-adaptive equalization method is characterized by comprising the following steps:

(1) the receiving end demodulates the passband signal into a baseband complex signal as an equalizer input;

(2) convolving the baseband complex signal with the coefficient of the equalizer to obtain the output of the equalizer;

(3) calculating an error between the desired signal and the equalizer output;

(4) and defining a new cost function by using the equalization error, and applying different constraints to the equalizer to adaptively update the tap coefficients of the equalizer according to the magnitude of each coefficient according to the CAP-LMS/F algorithm.

2. The complex-domain underwater acoustic channel adaptive equalization method as defined in claim 1, wherein: the new cost function is defined by adding p-norm of feedforward equalizer and feedback equalizer.

3. The complex-domain underwater acoustic channel adaptive equalization method as claimed in claim 1 or 2, characterized in that: the equalizer applies different constraints to self-adaptively update the tap coefficients of the equalizer according to the size of each coefficient, and is realized by comparing each equalizer coefficient with the mean value of all equalizer coefficients, applying no constraint to the large coefficients and applying strong constraint to the small coefficients.

Technical Field

The invention relates to an underwater acoustic signal processing method, in particular to a sparse underwater acoustic channel equalization method.

Background

Intersymbol interference caused by multipath spreading in the underwater acoustic channel and doppler shift caused by relative motion make the channel equalization at the receiving end a challenging task. A minimum mean square error (LMS) equalizer is widely used because of its simple operation and small computational complexity. However, the equalizer is sensitive to the input signal and the signal-to-noise ratio, and the performance is severely degraded especially under low signal-to-noise ratio conditions. The least-squares-error (LMF) equalizer can overcome the defects of the LMS by using the higher moment of the estimation error as a cost function, and better inhibit the interference of noise. However, LMF equalizers are very computationally complex. The LMS/F equalization algorithm is provided by combining the advantages of the LMS algorithm and the LMF algorithm, and the LMS/F equalization algorithm can effectively improve the equalization performance of the LMS and keep the simplicity and the stability of the LMS.

The underwater acoustic channel equalizer exhibits sparse characteristics, which means that most coefficients of the equalizer are close to zero, and only a few values are non-zero. This feature of the equalizer is caused by the sparse physical characteristics of the hydroacoustic channel itself. With this sparse characteristic, the equalizer performance will improve.

Disclosure of Invention

The invention aims to provide a complex domain underwater acoustic channel self-adaptive equalization method with high convergence rate and small equalization error.

The purpose of the invention is realized as follows:

(1) the receiving end demodulates the passband signal into a baseband complex signal as an equalizer input;

(2) convolving the baseband complex signal with the coefficient of the equalizer to obtain the output of the equalizer;

(3) calculating an error between the desired signal and the equalizer output;

(4) and defining a new cost function by using the equalization error, and applying different constraints to the equalizer to adaptively update the tap coefficients of the equalizer according to the magnitude of each coefficient according to the CAP-LMS/F algorithm.

The present invention may further comprise:

1. the new cost function is defined by adding p-norm of feedforward equalizer and feedback equalizer.

2. The equalizer applies different constraints to self-adaptively update the tap coefficients of the equalizer according to the size of each coefficient, and is realized by comparing each equalizer coefficient with the mean value of all equalizer coefficients, applying no constraint to the large coefficients and applying strong constraint to the small coefficients.

The invention discloses a sparse underwater acoustic channel equalization method based on a complex field least mean square/fourth power error (LMS/F).

The invention provides a new complex domain LMS/F equalization algorithm for complex receiving signals on a baseband. And in combination with the sparse characteristic of the equalizer, adding p-norm constraint in the cost function to improve the equalization performance. Compared with the traditional LMS/F algorithm, the complex number domain self-adaptive punishment LMS/F algorithm (CAP-LMS/F) in the invention distributes sparse constraint in a self-adaptive mode according to the size of the equalizer coefficient in each equalization process. For the equalizer coefficient with smaller amplitude, sparse constraint exists to accelerate the convergence speed of the equalizer coefficient; for equalizer coefficients of larger amplitude, the constraint disappears to improve equalization performance. Therefore, the performance of the newly proposed algorithm is improved in convergence speed and equalization error.

The invention has the advantages that:

(1) expanding the original LMS/F algorithm suitable for processing real signals to a complex field so as to process underwater acoustic complex signals on a baseband;

(2) and in each iteration, each tap coefficient of the equalizer is adaptively distributed with a sparse penalty item, so that the convergence speed of a small coefficient is accelerated, the convergence error of a large coefficient is reduced, and the equalization performance is improved.

Drawings

Fig. 1 is a schematic diagram of complex domain underwater acoustic channel adaptive equalization;

FIG. 2 is a flowchart of the CAP-LMS/F algorithm;

FIG. 3 is a comparison of equalizer tap coefficient magnitudes at an arctic under-ice experimental communication distance of 500 m;

FIG. 4 is a comparison of error rate performance of equalization algorithms when the communication distance is 500m under the arctic ice;

FIG. 5 is a comparison of the tap coefficients of the equalizer at a communication distance of 4km under ice in the arctic;

FIG. 6 is a comparison of error rate performance of the equalization algorithms when the communication distance is 4km under the arctic ice.

Detailed Description

The invention is described in more detail below by way of example.

1. The invention mainly comprises the following steps with reference to fig. 1:

(1) the input signal of the equalizer at n moments is r_n＝[r_n+Kr_n+K-1r_n+K-2… r_n]^TWherein r is a demodulated baseband complex received signal; k is the length of the feedforward equalizer;

(2) the output signal of the equalizer at n moments is

Wherein theta is_nCompensating the influence caused by phase deflection, wherein the size of the compensation is controlled by a second-order phase-locked loop; w is a_n＝[w_n,0w_n,1w_n,2… w_n,K]^TThe length of the tap coefficient of the feedforward equalizer is K + 1; f. of_n＝[f_n,0f_n,1f_n,2… f_n,L]^TThe length of the feedback equalizer tap coefficient is L + 1;is the symbol judged before the n time;

(3) calculating the error between the desired signal and the estimator output

Wherein x_nIs the desired signal at time n. In the training phase, x_nTo transmit a signal; in the tracking phase, x_nA decision symbol at time n;

(4) defining a new cost function using errors

Wherein epsilon > 0 is a threshold parameter which affects convergence speed and equalization performance; gamma ray₁,γ₂≧ 0 is used to measure the magnitude of the sparsity constraint imposed on the equalizerSmall;

0 ≦ p ≦ 1 denotes the p-norm of the equalizer tap coefficients. Couple cost functionsAnd obtaining a CAP-LMS/F channel equalization updating formula by derivation.

Referring to fig. 2, a flow of implementing the CAP-LMS/F algorithm in step (4) is described by taking a feedforward equalizer as an example, and an equalization process of the feedback equalizer is similar to that of the feedforward equalizer.

(1) Initializing a feedforward equalizer coefficient w (0) to be 0;

(2) calculating average value m of feedforward equalizer at n moments_w,n＝E(|w_nIn the formula, E represents the mean value;

(3) each equalizer coefficient w_n,iAnd m_w,nComparing, wherein i is more than or equal to 0 and less than or equal to K.

When | w_n,i|≤m_w,nWhen the temperature of the water is higher than the set temperature,in the formula, mu_wFor iterative step size, superscript^*The expression takes conjugation, sign (-) is a sign function. At the moment, sparse constraint terms exist to accelerate the convergence speed of the small equalizer coefficient;

when | w_n,i|＞m_w,nWhen the temperature of the water is higher than the set temperature,

at the moment, no constraint is applied to the large coefficient to ensure the accurate convergence of the large coefficient;

(4) judging whether the balance is finished or not; if not, returning to the step (2) for continuing.

2. Experimental study:

the CAP-LMS/F algorithm proposed in the present invention was verified by using the communication test data of the ninth North Pole scientific investigation. In the experiment, the transmitted signal consisted of a chirp sequence and a modulated BPSK or QPSK signal, where the chirp signal was used to synchronize the received signal. The frequency band of the transmitting transducer is 2-4 kHz, the center frequency of a signal is 3kHz, the sampling frequency is 48kHz, and the symbol rate is 1 ksymbols/s. The receiving end is a hydrophone with 4 array elements. The selected communication distances are 500m and 4km, respectively.

When the communication distance is 500m, the lengths of the equalizers are empirically set to K30 and L50, respectively. The step size in the algorithm is all selected to be 0.001. For BPSK modulation, gamma is selected₁＝3e-1,γ₂8e-3, ε 2; selecting gamma for QPSK modulation₁＝4e-2,γ₂1e-3, ε 2. The magnitude comparison of the equalizer tap coefficients obtained using the conventional LMS/F algorithm and the CAP-LMS/F algorithm is shown in fig. 3. It can be seen that the taps obtained by the CAP-LMS/F algorithm are sparse, which means that most tap coefficient values are close to 0. The method is characterized in that a term for improving the sparsity is added in the cost function, and the convergence speed of the algorithm is improved by enabling the small tap coefficient to be continuously close to 0 in the convergence process.

FIG. 4 compares the error rate performance of LMS, LMS/F, CAP-LMS/F algorithms under different training sequence lengths in BPSK or QPSK modulation. It can be seen that: (1) the LMS/F algorithm can overcome the problem that the LMS algorithm is sensitive to input signals and signal-to-noise ratios, and the balance performance is superior to that of the LMS algorithm; (2) with the increase of the training sequence, the error rates of LMS/F and CAP-LMS/F are in a descending trend regardless of BPSK or QPSK; (3) the error performance of CAP-LMS/F is better than that of LMS/F.

Secondly, when the communication distance is 4km, selecting gamma when the CAP-LMS/F algorithm is used for BPSK modulation₁＝3e-2,γ₂5e-3, 1; selecting gamma in QPSK modulation₁＝2e-3,γ₂1e-3, 1.5. The resulting equalizer tap amplitudes are shown in fig. 5. The structure of the equalizer of fig. 5 is more complex than that of fig. 3, but it still has a sparse characteristic. CAP-LMS/F enhances equalizer sparsity, making most tap coefficients 0, and retaining only a few significant tap values.

When the communication distance is 4km, the variation trend of the error rate along with the length of the training sequence is shown in figure 6. It can be seen that the result is consistent with the result when the communication distance is 500 m.