Lie Detection Technology Based On Voice And Radar Dual Sensors

Lying is a common and complicated behavior in human daily life,which will have a bad influence on the cognition of others,and the serious circumstances may even constitute a crime.The research of polygraph technology is of great significance to the fields of psychology,physiology,behavior,criminal investigation,and justice.As the direct carrier of lies,voice is one of the mainstream research directions of polygraph technology.However,due to the extremely strong individual differences in voice,the limitations of polygraph detection are large.Physiological signals have strong objectivity and regularity,and play an important role in polygraph detection,but its traditional contact measurement methods will affect the results.Therefore,this paper proposes a polygraph system and technology based on voice and radar dual sensors,which can obtain three signals of speech,heartbeat and respiration in a non-contact manner.With the help of feature extraction,feature fusion and machine learning technology to achieve polygraph detection,it can overcome the limitation of a single sensor and improve the accuracy of polygraph detection.The main work of this paper is:1.A polygraph system based on voice and radar dual sensors is proposed,and the working principle and components of the radar are introduced.Secondly,it introduces the polygraph paradigm and common databases,and builds a database suitable for the research of dual-sensor polygraph technology in this article.Finally,the machine learning algorithm models used in this article are introduced.2.Research on voice-based polygraph technology.First,the speech signal is preprocessed based on the combined noise reduction of SOX and improved spectral subtraction,and the detection of sound events based on the energy-zero-ratio method.Secondly,according to the characteristics of voice changes when lying,use open SMILE to extract voice features.Finally,the voice polygraph classification is carried out.The results show that the accuracy rate after feature selection based on Boruta algorithm can reach 64.8%.3.Research on radar-based polygraph technology.First,the radar signal is preprocessed by arctangent demodulation,filtering and crop synchronization to obtain the heartbeat and breathing signals.Second,perform traditional feature extraction on heartbeat and breathing signals.Based on the short-term change characteristics of physiological signals when lying,the frame-level features of heartbeat and breathing signals are extracted.Finally,the classification of heartbeat and breathing polygraphs is carried out.The results show that the accuracy rate of polygraph detection based on breathing signals can reach 64.6%,and that of polygraph detection based on heartbeat signals can reach 68.1%.4.Based on the polygraph technology model of voice and radar dual sensors,a feature fusion algorithm based on PCA dimensionality reduction and improved canonical correlation analysis is proposed.This algorithm fusion three feature sets of voice,heartbeat and breathing by maximizing the correlation between feature sets after dimensionality reduction.The results show that the accuracy of polygraph detection after fusion reaches 71%.Compared with the other two algorithms,the accuracy of the feature fusion algorithm proposed in this paper is improved by at least 2%,and the fusion classification time efficiency is improved by at least30.8%.