基于信息理论的舰船噪声盲分离算法

目前水声信号处理方法大多基于信号和环境的特定统计假设 ,使用限制较大。当模型假设不成立时 ,会严重影响信号处理效果。本文导出了基于信息理论的舰船噪声盲分离算法。该算法利用当信号相互独立时的信息理论特性 (互信息量最小或熵最大 )作为分离准则 ,逐步学习确定分离矩阵。算法无需输入信号和混合矩阵的任何先验知识 ,实验仿真证明了算法的有效性。

An Information Theory-Based Approach to Blind Separation of Ship Noise

Most of the approaches to underwater acoustic signal processing are dependenct upon the statistical assumption of signals and environment, which greatly limits their applications. Any failure in model assumption will lead to and unsatisfactory signal processing effect. A new blind algorithm to separate ship noise is presented in this paper. The algorithm utilizes the information theoretical property as the separation criterion that is the mutual information of independent signals is minimum or the entropy is maximum, and estimates separation matrix via learning step by step. The algorithm needs no priori information about the input signal and mixture matrix. Simulations have proved the effectiveness of the approach.