Lie Detection Based on a DIFCW Radar with Machine Learning

Lying is a common behavior that humans deceive and mislead others by concealing plots, false statements, and distorting facts which is important for intelligence service, bank, and justice department. In this paper, a system based on a DIFCW radar with machine learning is presented to perform polygraph. The radar is utilized to achieve non-contact measurement of heartbeat and respiratory signals. Based on the two signals, several features are extracted. And a model is trained through machine learning to determine whether the subject lie. In the experiment, a database of 11 males and 10 females with 435 lies and 562 truths are established. The accuracy rate of heartbeat and respiratory polygraph detection attach to 63.2% and 61.5%.