Emotion Recognition Based On Millimeter Wave Sensing Technology

With the continuous development of Internet of Things technology and artificial intelligence,the application scenarios of intelligent devices continue to increase,human-computer interaction tasks become increasingly complex.As another development direction of intelligent human-computer interaction,emotion recognition shows a broad prospect.Though emotion recognition based on image,speech and vital sign has achieved good results,the emotion recognition based on image and speech cannot identify the user's deep emotional state and the emotion recognition based on physiological signal generally uses wearable sensors to collect vital sign,which will interfere with the emotional state.In this case,the contactless emotion recognition based on millimeter wave sensing technology shows its advantages and prospects.For this reason,this paper realizes the emotion recognition system based on millimeter wave sensing technology,and achieves a good emotion recognition accuracy rate.The emotion recognition system based on millimeter-wave perception uses millimeter-wave radar to transmit FMCW signals,which are reflected by the chest and embedded with two vital signs of respiration and heartbeat.According to the characteristics of IF,the system uses FFT,phase extraction,bandpass filtering and other methods to extract respiration and heartbeat,and combined with heartbeat beat segmentation algorithm to extract the traditional HRV features from heartbeat.Finally,the convolutional neural network,long-short-term memory network and attention module are used to build an intelligent emotion recognition algorithm to realize the system.In addition,we collected physiological signals from 17 people for a total of 3,480 minutes to build a dataset of signals.The system uses these data sets to build the emotion recognition network model,and the average emotion recognition rate under the individual model is 80.59%.