Deceptive Chinese speech detection based on sparse decomposition of cepstral feature

In order to improve the performance of deception detection based on Chinese speech signals, a method of sparse decomposition on spectral feature is proposed. First, wavelet package transform is applied to divide the speech signal into multiple sub-bands, and the band cepstral feature of wavelet packet are obtained by operating discrete cosine transform on logarithmic of each sub-band energy, and cepstral feature is generated by combing Mel Frequency Cepstral Coefficient and Wavelet Packet Band Cepstral Coefficient. Then K-singular value decomposition algorithm is employed to achieve the training over-complete mixture dictionary based on both truth and deceptive feature sets,and orthogonal matching pursuit algorithm is used for sparse coding according to the mixture dictionary to get sparse feature. Finally, recognition experiments are performed with various classified modules. Experimental results show that the sparse decomposition method has better performance than conventional dimension reduced method. The recognition accuracy of the sparse cepstral features proposed in this paper is 78.34% higher than other features, improving the recognition ability of deception detection system significantly.