Systematic Study On Emotion Recognition And Classification Based On ECG And PPG Fusion

In the field of psychological clinical medicine,emotional problems have become one of the biggest causes of endangering human physical and mental health and life.Since the 21 st century,psychoclinical medicine has gradually begun to develop towards the integration of big data,artificial intelligence,machine learning,etc.,and emotion recognition is a key factor in the field of psychoclinical medicine.With the popularity of computer devices and wearable devices,ECG(electrocardiogram signal)and PPG(electroencephalogram signal)emotion recognition have attracted wide attention again.At present,the common methods are to use a single signal source for identification,but the shortcomings are that the feature expression is not ideal,the information is missing,and the artificial feature of the time domain signal has redundancy.This thesis "Systematic Research on Emotion Recognition and Classification Based on ECG and PPG Fusion" is a project carried out by the author in Beijing Zhongke Lingrui Innovation Base affiliated to the Chinese Academy of Sciences.On the basis of previous projects of Chinese Academy of Sciences,this paper proposes a wavelet time-frequency recognition algorithm based on ECG and PPG fusion(hereinafter referred to as the "algorithm").This algorithm uses order of magnitude fusion to get the fused signal,and then uses wavelet packet to extract the time-frequency image after fusion and saves it.Then Googlenet transfer learning in deep neural network was used to train the time-frequency image to complete the emotion recognition task.This method fuses ECG and PPG signals,and the obtained time-frequency images can better reflect the essential attributes of the signals compared with the time-domain ones.The features obtained through deep learning and self-learning have less redundant information,which improves the recognition rate and enhances the generalization performance.In this paper,based on the emotional fluctuation data of different individuals when watching different video clips,the ECG and PPG signal data received are identified to judge the current emotional state of the subjects.In order to judge the effectiveness and reliability of the proposed algorithm,it is compared with QPSO-BP,SVM algorithm and ECG signal separately.The results of the comparison are evaluated by confusion matrix,and finally verified in the validation set.The comparison results showed that the optimal classification accuracy of QPSO-BP was 91.2%,the optimal classification accuracy of SVM was 93.6%,the classification accuracy of single ECG was 90.8%,and the classification accuracy of Softmax function training classifier used in this paper was 96.8%.It is proved that the proposed algorithm is superior to the traditional single algorithm in precision and has certain advantages and effectiveness.