听觉小波包在舰船噪声分解与重构中的应用

舰船噪声的分解和重构对于舰船噪声特征提取与听音分析都具有重要作用。采用多分辨分析理论对噪声进行分解和重构只能对低频进行精细化处理,但对高频却难以达到理想效果。正交小波包分析可以在任何欲处理的频带内进行可允许的任意层次分解,但分解的小波包树的结构难以确定,没有明确的构造导引方法。将Bark频率群的理论引入小波包树结构的选择,提出构造听觉小波包树结构的方法,构建了基于db6小波的采样率为44.1 kHz的听觉小波包。采用该听觉小波包结构对四型舰船噪声信号进行分解、子带加权和重构,频谱分析和听音分析都表明重构噪声特征更加清晰。

Application of auditory wavelet packet to decomposition and reconstruction of ship-radiated noise

Decomposition and reconstruction of ship-radiated noise plays an important role in the feature extraction and listening analysis.Low-frequency part of signals can be finely processed by multi-resolution analysis,while high-frequency part not so.Signals can be decomposed at any level by orthogonal wavelet packet analysis,while there’s no guide to build the wavelet packet tree.The theory of Bark domain frequency bands is introduced and the method of building the auditory wavelet packet is proposed.The structure of auditory wavelet packet for signals at the sampling rate of 44.1kHz is built based on db6 wavelet.Four types of ship-radiated noise are processed by the auditory wavelet packet.The results of frequency spectrum and listening analyses indicate that the corresponding reconstructed noise is easier to be recognized.