Research Of Emotion Recognition Based On Respiratory And Electrocardiogram Signals

Emotion is a functional response of the body through cognition and evaluation of external objective stimuli.It is the physical,psychological and behavioral performance of human beings.With the rapid development of the current society,people's life is facing great mental pressure.In this state for a long time,it is easy to produce negative emotions and even cause symptoms like insomnia and loss of appetite.More will increase the incidence of anxiety,depression and other psychological diseases,and then threaten people's health and even life.In addition,emotions can affect the cardiovascular system and lung function,and long-term negative emotions may even aggravate the occurrence and deterioration of cardiovascular diseases and lung basic conditions.Effective emotion recognition can instantly understand their physical and mental health and reduce the negative impact of negative emotions.At the same time,it can also play an intervention role in the treatment of physiological and psychological diseases,so as to promote the recovery of patients.The generation of emotion is closely related to the activity of the autonomic nervous system.ECG and respiratory signals can indirectly describe the activity of the autonomic nervous system,and then reflect the state of emotion.Based on this,the paper extracted time domain,frequency domain and nonlinear features from Heart rate variability and respiratory rate variability,and analyzed the influence of emotion on the cardiopulmonary system by combining the coupling effects of heart and lung under different emotions.The main research contents of this paper are as follows:(1)A reasonable emotional induction experiment was designed to extract ECG and respiratory signals from 60 college students under six emotional states.Wavelet transform and Butterworth low-pass filter were used to de-noise the resampled ECG and respiratory signals.(2)The adaptive difference threshold method was used to identify the pre-processed R point from ECG and respiratory signal peak point respectively.Thus,RR interval and BB interval sequences were constructed.Linear features,nonlinear features and coupling entropy features based on two time series and ECG and respiratory signals were extracted.(3)The results showed that the sympathetic function of the autonomic nervous system was strengthened in the positive emotional state of happiness,and the degree was higher than that of the weakened Vagus nerve.In the two negative emotions of sadness and fear,the intensity of the action of the Vagus nerve increases,and the regulation mechanism of the autonomic nervous system in the state of anger and disgust is strengthened,and its activity is strengthened.(4)ReliefF algorithm was used to screen the features.KNN and Random Forest were used to establish two emotional classification models for emotion classification.The results showed that the sentiment classification model based on Random Forest had a good classification effect.The correct rate of one-to-one sentiment classification was above 90%,and the highest rate was 97.5%.In multi-emotion classification,the accuracy of this model could reached 83.61%,which further indicated that the features extracted in this paper could be used for emotion classification.