摘要
基于声纳员的感受被动声纳可以认为是一个发声体,这个发声体可以表示为激励噪声源与发声体冲激响应的卷积。在这种情况下,使用倒谱和复倒谱的形式分析被动声纳目标噪声的时域特征,得到的目标特征不够明显,因此提出了利用指数倒谱和指数复倒谱的频谱特性来提取目标噪声的特征,进行分类识别。设计了BP神经网络分类器,利用实测数据对三类目标进行分类。分析比较了两种方法的分类结果,验证了基于倒谱和复倒谱的指数运算的被动声纳目标特征提取方法的可行性。
In term of the sonar operators' feeling,the passive sonar can be regarded as a sounding body.And the target noise can be expressed as a convolution of the activated noise source and the impulse response of vocal body.In this case,by using cepstrum and complex cepstrum to analyse the time-domain characteristics of target noise,the obtained target feature is not enough for target classification.Therefore,a method of using the spectral characteristics of the in-dex cepstrum and index complex cepstrum to extract the features of target-noise for classification is proposed.A BP neural network classifier is designed to classify three categories of targets from the measured data.Through analysis and comparison of two classification methods it is verified that the method based on the index operations of cepstrum and complex cepstrum is feasible.
出处
《声学技术》
CSCD
2011年第3期280-283,共4页
Technical Acoustics
关键词
倒谱
复倒谱
目标识别
被动声纳
特征提取
cepstrum
complex cepstrum
target identification
passive sonar
feature extraction