Speech Deception Detection Based on EMD and Temporal Neural Network

Deceptive behaviour is a common phenomenon in human society. Research has shown that humans are not good at distinguishing deception, so studying automated deception detection techniques critical task. Most of the relevant technologies susceptible to personal and environmental influences: EEG-based need large expensive equipment, facial-based sensitive with camera’s perspective, these reasons have somewhat limited development applications for technologies. In contrast, equipment required speech cheap easy use, capture highly covert. Based on application signal decomposition algorithms other fields such as EEG signals emotion recognition, this paper proposed reconstruction method based EMD process better performance was obtained by improving quality. The comparison results showed algorithm most suitable our method. Across many different classification algorithms, accuracy improved an average 2.05% F1 score 1.7%. addition, new detector, called TCN-LSTM network, paper. Experiments network organically combines processing capability TCN LSTM time series data; recognition rate greatly improved, highest reaching 86.2% 86.0% under EMD-based research paper, be further optimised more used task should tried.