Research on Emotion Recognition Based on Multisource Signals

Current physiological signal-based emotion recognition systems, physiological signals are basically collected through wearable devices, but contact physiological signal collection has a great inconvenience. The existing non-contact emotion recognition system based on vital sign signals has the problem of poor recognition accuracy of single sensor. In this paper, a contactless dual-modal emotion recognition system based on radar and video sensors is proposed to address the above problems. The recognition system based on radar and video sensors exposes poor performance in detecting video heartbeat signals in a weak light environment. The heartbeat signal optimisation algorithm based on light intensity is proposed, which can select a more accurate heartbeat signal source under different light intensities and effectively improve the accuracy of heartbeat signal detection in weak light intensity environments. The experimental results show that after the fusion algorithm, the accuracy of decision-level recognition is 91.4% and the accuracy of feature-level recognition is 90.8%.