Research On Emotion Recognition Based On Short-time ECG

Emotions are closely related to people's health.Negative emotions can be detected through emotion recognition technology,which helps maintain physical and mental health.At the same time,emotion recognition is also an important part of realizing human-computer emotional interaction.This paper summarizes the relevant research at home and abroad,and focuses on the real-time nature of emotion recognition.Finally,short-term emotional ECG with a length of 10 s is selected as the research object to carry out relevant research on emotion recognition.The main research work of this paper is as follows:In this paper,21 emotionally and physically healthy students were collected through videoinduced methods,and together with the Bio Vid Emo DB emotion database,five emotions including joy,sadness,disgust,fear and anger were formed.Emotional database within.The database contains emotional ECG data and the facial expression videos of the subjects during the experiment.The facial expression videos will be used as annotations of the emotional ECG data.In the preprocessing stage of ECG signal,sym wavelet is selected to reduce the noise of ECG signal.The emotion data collected in the experiment often contains a considerable amount of nonemotion content.At the same time,because this paper uses short-term ECG signals to conduct relevant research on emotion recognition,it is necessary to design a corresponding labeling system to mark the effective part of the emotion data.In this paper,a labeling system based on facial expression recognition is designed.The face recognition part is realized by a multi-task convolutional neural network,and the expression recognition part is realized by a residual network.Using this system can effectively mark the location of short-term emotional ECG signals with a length of 10 s.Then,the Pan & Tompkins differential threshold algorithm is used to identify the R wave of the short-term emotional ECG signal to obtain the RR interval of the short-term ECG signal.Taking the RR interval sequence as the research object,multiple indexes of heart rate variability were obtained through time domain,frequency domain and nonlinear analysis.And through the statistical analysis method,the differences of the five heart rate variability indicators under the emotional state were studied.Finally,based on a variety of machine learning methods,this paper establishes a corresponding recognition model for one-to-one emotion recognition and recognition between multiple emotions,and optimizes the model through 10-fold cross-validation and network search algorithms.Experiments show that the support vector machine and the random forest algorithm can achieve a high recognition effect in one-to-one emotion recognition.At the same time,the random forest algorithm also performs well in the classification of five emotions.Finally,based on the above research results,this paper completed the interface design that can analyze and recognize emotions of short-term emotional ECG signals.