基于高阶谱的LOFAR谱图特征在被动声纳信号自动识别中的应用

水下目标的特征提取一直是水声信号处理的难题。由于实际的声纳目标的发声机理和反射声波的机理十分复杂，成份多样，造成水声信号具有较强的非平稳性和非高斯性。为了提高对水下目标识别的正确识别率，突破以往研究中关于信号平衡性的假设和高斯性的假设，本文在简要介绍了ＬＯＦＡＲ谱图和高阶谱（ＨＯＳ）估计方法之后，提出了一种基于高阶谱的ＬＦＡＲ谱图和高阶谱（ＨＯＳ）估计之后，提出了一种基于高阶谱的ＬＯＦＡＲ谱图特征

An Application of Features Extracted from LOFAR Spectrum and Higher Order Spectra to Passive Sonar Target Automatic Recognition

Feature extraction of underwater target is always a difficult problem in acoustic signal processing.Because the mechanism of sonar target phonation and the echo are very complicated ,acoustic signals take on apparent nonstationarities and non-Guassianness.In order to enhance the correct rate of underwater target recognition and throw off stationary and Gaussian assumption,this paper brings forward a new feature extraction method from LOFAR spectrum based on higher order spectra (HOS-LOFAR spectra).This method is not restricted by either stationary or Gaussian assumption.The results of experiment demonstrate that using this method to extract features from sonar signals is satisfying and feasible.