小波尺度相关滤波及其改进算法在电火花声源信号处理中的应用

Application of wavelet scale correlation filtering and its improved algorithm in signal processing with a spark sound source

分析提取宽带电火花声源信号,受到船舶辐射噪声的严重干扰,在5 kHz以下的能量集中区信噪比低,且二者Lipschitz指数特性相近,传统基于相邻尺度相关的滤波算法抗干扰能力有限。根据宽带电火花声源信号不同频带所受干扰的不同,信噪比较高的小尺度高频小波系数,采用相邻尺度相关的滤波算法;信噪比较低的大尺度中低频小波系数,采用跨尺度相关的滤波算法,并对算法中阈值系数的选取方法进行修正。结果表明,该算法滤波效果良好,有效的提取了电火花声源信号,适合窄带强干扰背景噪声下的宽带水声信号处理。

It is seriously interfered by ship noise when analyzing and extracting broadband spark sound source signal. In the energy concentrated domain which is below 5 kHz, the traditional scale correlation filtering algorithm, which is based on adjacent-scale correlation, has limited anti-interference ability due to the low signal-to-noise ratio （SNR） and similar Lipschitz exponent characteristic of each other. However, because different frequency bands of the broadband electric spark signal have different noise interferences, the filtering algorithm based on adjacent-scale correlation is adapted to high SNR and small-scMe high-frequency wavelet coefficients filtering; the filtering algorithm based on cross- scale correlation is adapted to low SNR and large-scale low-frequency wavelet coefficients filtering, and the threshold coefficient selection method had been corrected in the algorithm. It is shown that the filtering Mgorithm has a good filtering effect and extracts the broadband spark sound source signM effectively; it is applicable to broadband underwater acoustic signal processing in the presence of narrow-band strong interference background noise.