Physiological Emotion Recognition Based On Deep Learning

With the development of science and technology,people have higher and higher demands on intelligent technology.Emotional computing is also taking place.Emotional computing is to give computers the ability to detect,classify,organize,and respond to human emotions,so that users can get efficient and friendly feelings.Human emotions are expressed through expressions,voices,gestures,and physiological signals.Among them,expressions,voices,and gestures are physically expressed and may not be able to accurately express human emotions.Physiological signals are signals that are naturally produced by the human body and obtained by easy-to-use sensors,and contain rich human emotional information.These changes in information can directly reflect human emotional states.Physiological emotion recognition can recognize people's emotions,fully understand people's emotions and mental states,and help people live a better life.This article made some explorations from the physiological signal emotion feature extraction method and physiological emotion recognition.In order to improved the traditional feature extraction and related disadvantages of emotion recognition,traditional feature extraction algorithm and deep learning algorithm was used to extract the physiological signal for emotion Identify.The main research work and results are summarized as follows:First,in order to explore the recognition effect of traditional feature extraction methods in the DataSet I physiological library and DEAP public physiological library,SVM and KNN classifiers were used to perform emotion recognition on the extracted physiological features.This paper mainly adopted time domain feature extraction method and frequency domain feature extraction.Time domain feature extraction mainly extracts the average value,standard deviation,average value of absolute value of first-order difference,average value of absolute value of first-order difference of normalized signal in DataSet I physiological library and SetDEAP public physiological library There are six types of statistical characteristics: the average value of the absolute value of the second-order difference and the average value of the absolute value of the second-order difference of the normalized signal.The frequency domain feature extraction methods are PSD and Fourier transform.Second,this paper proposed the Fourier coefficient model what was extracted physiological signal feature.Based on the Fourier coefficient model,this paper extracted the Fourier coefficient characteristics of the physiological signal,the first-order difference and the second-order difference of the Fourier coefficient characteristics,and the global characteristics(maximum,minimum,variance,median,and average)A total of 450 parameter features were used to support emotionvector recognition of physiological signals in the DEAP library.By comparing with statistical values and PSD,the experiment proved that the Fourier coefficient characteristics are effective for physiological signal emotional recognition.Third,the physiological signal recognition methods of A-LSTM and C-LSTM models are proposed.In this paper,the Attention mechanism is combined with LSTM and a convolutional layer is added to the LSTM model to perform feature extraction and emotion recognition on physiological signal signals.The Attention mechanism focuses the model's attention on the physiological signal segments with obvious features through weights,which can make the model better recognize emotions of physiological signals.The convolutional layer can quickly capture the local information of the signal.At the same time,the convolution of the one-dimensional vector will reduce the feature dimension.And through experiments,it is found that the recognition effect on the label of love is very good.Figure [21] table [12] reference [79]...