多途水声信号盲恢复技术研究

针对盲解卷积只能获取未知源信号的波形,而得不到源的幅度问题,研究了一种平稳信道、多途传播条件下估计信道参数及源分离的算法.该算法引入均值和方差都已知的参考信号,在FIR模型的基础上利用解相关技术估计水声多途信道参数,进行盲信号分离.分析了实际的水声信道参数特点,提出新的信道参数及阶数的估计方法,继而利用改进的自适应逆滤波器恢复源信号,最后在瞬时混合以及不同信噪比情况下与其他算法进行比较,以验证算法的有效性.仿真结果表明,未知源估计结果与信道阶数和信噪比有关;对于卷积混合,该方法对于信道参数的估计具有稳健性,.可以同时恢复源信号的时序结构与幅度信息;对于瞬时混合,该方法的计算量和性能指标优于其他算法.

Improved blind restoration of multi-channel underwater signals

To calculate the amplitude of signals from an unknown source,a novel method was developed that evaluates channel parameters and separates unknown source signals into a stationary channel in addition to multi-path propagation channels.Based on the finite-impulse response(FIR) channel model,reference signals with zero-mean and known variance were introduced to determine the channel characteristics using decorrelation.First,a new method for estimating the parameters and the order of the channel was formulated based on analyzing sound propagation characteristics in the ocean.Next,the blind signals were restored using an improved adaptive inverse filter.Finally,to prove its validity,the algorithm was compared with four other algorithms using different SNRs and linear instantaneous mixtures of the sources.Simulation results showed that the estimation results of unknown source mainly depend on the channel order and SNR.For convolution mixture case,the algorithm is robust for the estimation of channel parameters,with both temporal structure and amplitude information of the signals resumed simultaneously.For instantaneous mixture case,the computational load and performance indexes of the proposed method are better than those of the other four algorithms.