小波包节点分段阈值降噪在水声监听中的应用

Wavelet packet node segmental threshold denoising for underwater acoustic monitoring

水声目标信号在发送、传播过程中,易受到环境噪声、系统自噪声等影响,因此水声监听过程中目标信号会掺杂大量噪声信息。为提高获取目标信号的准确性和可靠性,降低噪声,在已有小波分析基础上,提出小波包节点相对能量判断最优分解层,最优分解层节点系数分段阈值处理重构方法,实现水声监听信号分频段去噪。将0.1 kHz 8.4 kHz实验数据按节点频率排序划分为5个强弱不同的频段信号实现消噪提取,结果表明该方法可将噪声信号与目标信号有效分离,与全局单一阈值相比,具有较好降噪能力。该方法打破了小波阈值去噪高频处理的局限性,提高了识别精度,改善了全局单一阈值去噪存在的短板,在鱼类分析识别、舰船监听、深海探测等方面具有一定的推广和应用价值。

Underwater acoustic target signal is easy to be affected by environmental noise and system self-noise in the process of transmitting and receiving.The target signal will be doped with a lot of noise information in underwater acoustic monitoring.In order to reduce noise interference and improve the accuracy and reliability of acquiring target signal,the relative energy criterion of wavelet packet nodes is proposed based on the existing wavelet analysis.The target signals are de-noised and extracted respectively from 0.1 kHz to 8.4 kHz in the five frequency bands of the underwater acoustic monitoring experimental data.The results show that the noise band signal can be effectively separated from the target signal band by node segmentation threshold processing.Compared with the global single threshold wavelet method,it has good separation and noise reduction ability.This method breaks the limitation of wavelet threshold denoising in high frequency processing,improves the recognition accuracy,and effectively overcomes the shortcomings of global single threshold rule denoising.It has good popularization and application value in marine biology investigation,ship recognition,deep-sea exploration,etc.