Research On Emotion Recognition Of Physiological Signals Under Environmental Factors

With the development of Internet technology,the human-computer interaction time is getting longer and longer.In the process of human-computer interaction,human emotions often change.Therefore,the direction of affective computing has been paid more and more attention Nowadays,there are more and more wearable devices,which can obtain physiological signals without the user's awareness.The use of human physiological signals for emotion recognition is more objective and accurate than external features for emotion calculation.Therefore,the research of emotion recognition based on physiological signals has received much attention.In psychological research,scholars have noticed the influence of the weather on human emotions.At present,light therapy has become one of the main methods to treat patients with seasonal affective disorder in clinical medicine in China.Meteorological psychologists point out that all kinds of weather factors will affect people's body and mind,and then affect people's speech and behavior.This paper explores the influence of environmental factors on emotions based on physiological signals,and the main research work is as follows:1.Establish emotional database of physiological signals.Firstly,a video library of seven emotions(anger,disgust,fear,joy,calm,sadness and surprise)was established by combining the tag search and manual selection methods of video websites.Then,physiological signals,including Galvanic Skin Response signals and Blood Volume Pulse signals,were collected under seven emotional video stimuli under four environmental conditions of light and comfort,no light and comfort,light cold,no light and cold,respectively,and labeled with the simplified Chinese version of PAD emotion scale,forming 1120 samples.2.Propose the emotion recognition method of physiological signals based on machine learning,and explore the emotion recognition results under four environmental states.Firstly,feature extraction was carried out on the Galvanic Skin Response signal and Blood Volume Pulse signal.Then KNN(K-nearest Neighbors)and SVM(Support Vector Machine)are used to evaluate GSR(Galvanic Skin Response)and BVP(Blood Volume Pulse)for emotion recognition.The experimental results show that the SVM model with the mean characteristics of GSR signals has the best recognition effect.Second,the four conditions of GSR signal respectively the emotion recognition by using the SVM model,the experimental results show that in a comfortable environment,joy recognition effect is relatively good,under the condition of light,disgust and fear recognition effect is relatively good,and happy emotion recognition rate is relatively lower,in a cold environment,the recognition rate of sadness was the highest,while the recognition rate of joy and surprise decreased relatively.3.Propose the emotional recognition method of physiological signals based on LSTM(Long Short Term Memory)model,and explore the influence of different environmental states on emotional states.The GSR signals were segmenting and windowed,and the windowed GSR signals were calculated for statistical values to form characteristic parameters and construct the emotional recognition model of LSTM physiological signals.Experimental results show that in a comfortable environment,joy recognition effect is relatively good,under the condition of light,anger and fear emotion recognition effect is relatively good,and happy emotion recognition rate is relatively lower,in a cold environment,The recognition rate of sadness was the highest,while the recognition rate of anger,fear and happiness was relatively lower.Figure [23] Table [6] Reference [73]...