基于LPC谱和支持向量机的船舶辐射噪声识别

Recognition of radiated noises of ships based on LPC spectrum and support vector machines

船舶辐射噪声是非常复杂的,寻找新的特征是目前水下目标识别中的一项非常迫切而艰巨的任务。基于线性预测编码(LPC)原理提出了一种加权交叠平均的LPC谱估计算法,同时给出了支持向量机解决多类分类问题的一对多方法。利用得到的LPC谱特征矢量用支持向量机分类器和BP神经网络分类器对海上实测的三类目标噪声数据进行了分类识别,并与一般的LPC谱特征进行了对比。结果表明,加权交叠平均的LPC谱特征对三类目标的总体正确识别概率在95.02%以上,并且比一般的LPC谱特征具有更好的分类性能,支持向量机的分类性能也优于BP神经网络的分类性能。

Radiated noises of ships are very complicated.It is an imperative and difficult task to look for the new feature from ships noises in the classification of underwater targets.On the basis of the principle of the Linear Predictive Coding（LPC）,a weighting overlaps average LPC power spectrum density estimation algorithm is proposed,and an algorithm （One-Against-All, OAA） of multi-class Support Vector Machines （SVMs） is defined. Then the extracted feature vectors and SVMs,BP neural network are used to classify three different classes of targets.Results show that the statistical classification corrective rate exceeds 95.02% ,the weighting overlaps average LPC spectrum has better discernment performance than the general LPC spectrum Feature,and the SVMs performance is better than that of BP neural network.