基于子波奇异性检测的水声信号去噪方法研究

Research of Underwater Acoustic Signal De-noising Method Based on Singularity Detection with Wavelet

基于低信噪比、非平稳水声信号中,信号的有效频谱是固定的、噪声的频谱是随机的、信号频谱由代表水声信号特征的各基波和其对应的各次谐波构成的特点,提出了利用子波奇异性检测特性实现水声信号去噪的方法。该方法首先对水声信号进行多分辨分解,求取各尺度上信号的频谱成分;然后,根据最大尺度上最高的频谱幅度,确定该尺度上信号的有效频谱成分,再按照基波与谐波的关系,确定其它各级尺度上的有效频谱成分,去除奇异频谱;最后对去除奇异频谱后的信号进行频域-时域变换及子波重构。经对实测水声信号进行仿真,获得了较好的去噪效果。

Considering the characteristics of low signal noise ratio and non-stationary underwater acoustic signal, such as the fixed signal spectrums, the random noise spectrums, the mixed signal spectrums consisting of the fundamental waves and their harmonic etc., this paper proposes a method for underwater acoustic signal de-noising, which is based on the singularity detection with wavelet. This method includes three steps. The first step is the multi-resolution decomposition of the underwater acoustic signal, then to calculate the power spectrums on every scales. The second step is to confirm the largest amplitude of power spectrum on the largest scale, according to deciding the effective power spectrums on this scale. Then the effective power spectrums on every other scale are decided to wipe off the singularity power spectrums, based on the relationship between the fundamental waves and their harmonic. The third step is to transform the frequency domain signals into time domain signals on every scale for wavelet reconstruction. The simulation results of using real underwater acoustic data show that this method can get perfect effect.