论文部分内容阅读
提出一种水下目标回波的特征提取方法。该方法根据平稳小波变换的冗余性和奇异值的稳健性,将湖底回波信号的平稳小波变换系数矩阵的奇异值作为特征向量。它本质上是利用平稳小波变换将信号分解到多个子空间,再采用K-L变换实现对子信号的特征压缩。实测数据分析表明,本文方法与子带能量特征法相比: (1)在相同的样本集和类内距条件下,得到的类间距大于后者, (2)不论所选测试样本和训练样本是否属于同次湖试所得,分类正确识别率均高于后者, (3)随样本集的变动,其正确识别率抖动程度远小于后者。因此,该方法能得到更加稳健、有效的特征以及更好的分类效果。
A feature extraction method for underwater target echo is proposed. Based on the redundancy of stationary wavelet transform and the robustness of singular values, the method takes the singular value of the stationary wavelet transform coefficient matrix of the bottom echo as the eigenvector. It is essentially the use of stationary wavelet transform the signal is decomposed into multiple subspaces, and then use K-L transform to achieve the characteristics of the sub-signal compression. The measured data analysis shows that the proposed method compares with the sub-band energy feature method: (1) with the same sample set and class spacing, the distance between the two classes is larger than the latter; (2) no matter whether the selected test sample and the training sample Belong to the same lake test, the correct recognition rate of the classification are higher than the latter, (3) with the sample set changes, the correct recognition rate jitter is much less than the latter. Therefore, the method can get more robust and effective features and better classification results.