Research on feature extraction of engine abnormal sound signal based on linear prediction analysis

Engine abnormal sound signal can be regarded as convolution of impulse response of glottal excitation signal and abnormal sound channel signal, so linear prediction algorithm is proposed as deconvolution method in this paper and a joint feature vector is established by extracting the features of abnormal sound signal. Then the low-dimensional fault diagnosis feature with maximum fault signature information is obtained by using principal component analysis. Experimental results demonstrate the effectiveness of the proposed method.