去噪小波包能量法在水声信号识别中的应用

对小波包能量法进行改进,提出了一种新的方法———去噪小波包能量法,该算法对信号小波去噪后,再用小波包能量法提取信号的特征;应用去噪小波包能量法研究了不同的小波去噪方法对水声信号分类识别率的影响,在实测信号样本集上用BP神经网络进行了识别实验。结果显示软阈值、硬阈值、弹性阈值3种标准的小波去噪方法均能明显提高信号识别率,其中最为显著的是软阈值标准去噪,信号的识别率可由未去噪的53 .3%提高到98. 3%,表明算法的有效性。

Application of the Denoising Wavelet-Pocket Energy Method to the Recognition of Underwater Acoustic Signal

To improve the wavelet-pocket energy method,we presented a new one-denoising wavelet-pocket energy method. The arithmetic first denoised the signal and then contracted the features of it with wavelet packet energy method. Different wavelet-based denoising standard were studied and experiments on BP neural Network were made. The results show that the denoising standard based on soft, hard and stretch gating can increase recognition ratio evidently, especially the soft standard, recognition ratio raise from 53.3% to 98.3%, which proves the validity of the arithmetic.