Emotion recognition method using millimetre wave radar based on deep learning

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