基于1（1/2）维谱与K-L变换的被动声纳目标识别

采用基于1（1/2）维谱分析与K-L变换相结合的特征提取方法,获取被动声纳噪声信号的有效识别信息,对被动声纳的目标信号进行分类。首先对被动声纳噪声进行1（1/2）维谱子带能量的特征提取,然后运用K-L变换实现高维特征向量的降维,剔除冗余特征,并以BP神经网络作为分类器对三类目标进行识别与分类。计算机仿真结果表明,该方法具有较好的分类效果和稳健性。

Passive sonar identification based on 1(1/2)-D spectrum and K-L transform

In order to obtain valid information of passive sonar noise signal and achieve target signal recognition,a new method of passive sonar noise feature extraction is presented based on 1(1/2)-D spectrum and K-L transform.The feaure of the sub-band energy for passive sonar noise is extracted as initial feature vector by using 1(1/2)-D spectrum,and then K-L transform is used to reduce dimension of high dimension feature vector,remove the redundant feature and get the final feature.BP neural network is taken as the classifier to recognize and classify the passive sonar noise.The simulation results show that the method of features extraction has the characteristics of better classification effects and stability.