声矢量信号双谱与互双谱估计算法

提出了声矢量信号双谱与互双谱估计算法，给出了算法的具体步骤。将算法应用于两类水中目标的特征提取，并用所提取特征构造了LMBP神经网络的输入向量集，对矢量水听器实测的水中目标进行了分类识别。识别结果验证了所提出算法的有效性。实验表明，B类目标识别率优于A类目标，原因是由于B类目标特征频率较集中，而A类目标特征频率较分散所致。互双谱特征分类结果优于双谱特征分类结果这个事实是与声压振速联合信号处理优于声压或振速单一信号处理相吻合的。

The algorithms of estimating bispectrum and cross-bispectrum based on acoustic vector signal

The algorithms of estimating bispectrum and cross-bispectrum based on acoustic vector signal are proposed and the detail steps are given. The proposed algorithms are applied to feature extraction from two types of underwater acoustic targets. And the features construct input vector set of LMBPNN. The identification of underwater acoustic targets can then be made based on the extracted features. The identification results show the effectiveness of the proposed algorithms. And it has been seen by experiments that the recognition rate for Class B targets is superior to that for Class A targets. This is because the characteristic frequencies of Class B targets are relatively more concentrated than that of Class A targets. Furthermore, the fact that the classification performance of using the cross-bispectrum is superior to that of using the bispectrum is in coincidence with the fact that the performance of the joint processing of pressure and particle velocity is superior to that of the single processing of pressure or particle velocity.