Research On Emotion Recognition Method Based On Continuous Wave Sound Signal

Emotion recognition has been applied in many fields,such as safe driving,medical care and social security.Generally,emotion recognition methods can be divided into two categories.One is the use of human physical signals such as facial expressions,voice,gestures and postures,which has the advantage of easy collection and has been studied for many years.However,the reliability can not be guaranteed,because it is relatively easy for people to control physical signals such as facial expression or voice to hide their true emotions,especially in communication.The other is the traditional method of emotion detection,which usually uses some sensors to measure physiological signals.The user experience is poor,and the sensor itself may affect the wearer's emotion.Therefore,this paper focuses on the non-contact and user experience of emotion recognition,and proposes a novel method of emotion recognition using continuous wave sound to solve many problems such as inconvenient traditional measurement.However,emotion recognition using continuous wave signals is a challenging topic.The research contents and contributions of this paper include:First of all,a mobile application is developed,which can send a pre-designed continuous wave signal and receive the reflected echo through a microphone.Then,beamforming mechanism is used to sense the change of user's chest displacement.Secondly,this paper uses the method of cross-correlation function to eliminate the time delay between the transmitted signal and the received signal,generates and analyzes the continuous wave signal,and uses the method of band-pass filter to remove most of the noise.The received signal is down converted into baseband signal,and then the amplitude of sound signal is obtained.The general respiratory waveform is obtained by down sampling,and the conversion from sound signal to respiratory signal is completed.Finally,a novel lightweight neural network is proposed to extract and classify emotion features.In addition,video is used as a stimulus to collect a large amount of data,which are breath signals collected under calm,anger,fear and joy.Two types of emotion classifiers are trained based on different people's data.The performance of breath monitoring and emotion classification in different scenes are evaluated,and each scene gets higher accuracy,which can meet the common breath monitoring and emotion recognition.