Emotion recognition method using millimetre wave radar based on deep learning

Abstract

Non-contact emotion recognition is a new research field. Using millimetre-wave radar does not require users to wear any equipment, and there is no privacy violation. This study proposes a new emotion recognition method, which uses millimetre-wave radar to transmit frequency-modulated continuous wave signals and extracts and separates the time domain signals of heartbeat and respiration of subjects through the echo of human reflection. According to the characteristics of collected signals, a deep learning framework combining one-dimensional convolutional neural network and Bidirectional Long Short-Term Memory neural networks is designed for feature extraction and classification. Experiments show that this method has high average recognition accuracy for four emotions in the case of person-independent. The excellent range resolution, higher system integration, and lower power consumption of millimetre-wave radar will profoundly impact the development of human–computer interaction, modern medical care, and education.