Research On Ensemble Learning Emotion Recognition Framework Based On ECG Signal Deep Features Of CNN And Bidirectional GRU

Emotion is a state of mind that is inextricably linked to thoughts,feelings,behavioral responses,etc.Emotions are complex,they are states of feeling that can affect our behavior,physical and psychological changes.In People's Daily work and life,the role of emotion is everywhere.Since the middle of the 20 th century,the rapid development of machine learning and deep learning has greatly promoted the development of artificial intelligence.Automatic emotion recognition technology has also advanced by leaps and bounds.ECG(Electrocardiogram)is the use of electrodes placed on the skin over a period of time the process of recording the electrical activity of the heart.ECG is not only easy to acquire,but also can react emotional state quickly,which makes it one of the important research directions of automatic emotion recognition.However,after years of development,the research has encountered a certain bottleneck.It is mainly reflected in three aspects:(1)The existing ECG division methods usually need a long time(>30s)to achieve a certain accuracy;(2)In the existing research on emotion recognition,the context of emotional state change is rarely considered,that is,the temporal correlation of ECG feature sequences is not utilized.(3)In the existing research on emotion recognition,most of the features are extracted by hand.The effectiveness is limited and extremely complex,which requires a lot of time.To solve the above three problems,this paper designs a framework of emotion recognition based on time-frequency features of ECG signal and an automatic emotion recognition algorithm based on deep feature fusion of ECG signal.The details are as follows :(1)an emotion recognition framework based on time-frequency features of ECG signal is designed.By extracting time domain features,frequency domain linear features,frequency domain nonlinear features and time-frequency domain features of short time(10s),and applying them to machine learning algorithm,the average accuracy rate of positive and negative emotion binary classification experiment was finally reached to 79.51%.(2)an ensemble learning emotion recognition framework based on the deep features of ecg signal CNN and bidirectional GRU is designed.The convolutional neural network model and the bidirectional gated loop unit(GRU)model were designed to extract and fuse the two depth features.The framework of the algorithm is based on the original ECG signal to extract features,and the temporal correlation of ECG signals is taken into account.The ensemble learning algorithm is added,which not only improves the accuracy of classifier to 87.71%? but also enhances the generalization performance of the model.The algorithm provided in this paper can directly extract features based on the original ECG signal without complicated manual feature extraction process,and has a breakthrough in accuracy.It is a meaningful exploration of automatic emotion recognition based on ECG signal and provides a new exploration direction.