Applying SBL and Non-Linear Dynamics Features for Detecting Deception from Speech Signal
Extracting novel non-linear dynamics (NLD) feature sets and applying SBL classifier for deception detection based speech processing is the primary aim of this study. As the NLD features provide additional information regarding the dynamics and structure of deceptive speech, here, 24 NLD features that show significant correlations to deception are selected. The features have been computed partially, and represent so far unknown acoustical perceptual concepts. After a correlation-filter feature subset selection, Sparse Bayesian Learning (SBL) classification model was trained. SBL algorithm is turned out to gain a satisfactory performance for detecting deceptive speakers on the NLD feature sets. Compared with the classical model of SVM and RBFNN, the proposed model achieves high classification accuracy in detecting deception.
Keywordsdeception detection speech signal Sparse Bayesian Learning (SBL) non-linear dynamics (NLD)
Unable to display preview. Download preview PDF.
- 7.Vrij, A.: Detecting Lies and Deceit: Pitfalls and Opportunities, 2nd edn. The Psychology of Crime, Policing and Law. John Wiley & Sons, Ltd., West Sussex (2008)Google Scholar
- 8.Zheng, C., Zhang, L., Vincent, T.-Y.N..., Huang, D.S.: Metasample-based sparse representation for tumor classification. J. IEEE/ACM Transactions on Computational Biology and Bioinformatics 8(5), 1273–1282 (2011)Google Scholar