Emotion Recognition Based On Mouse Trajectory And EEG Signals

Emotion recognition refers to extracting relevant features of emotional data to obtain feature data that can express human emotions,and then modeling and analyzing the feature data to find out the correlation between the external physical manifestation of emotions and the true internal emotional state,and finally The process of identifying the emotional category people are currently in.The world is entering the era of artificial intelligence,and the fields of human-computer interaction and emotion recognition are of great research value and significance.Therefore,this article uses deep learning algorithms to propose a number of emotion recognition models based on mouse sliding trajectories.While collecting mouse data,collect participants' brainwave data,and build network training for brainwave signals to enhance the reliability and persuasiveness of mouse emotion classification.Finally,the experimental results of the two are compared and summarized respectively.The specific research content and results of the thesis are as follows:(1)This article first designed an emotion induction experiment,collecting the mouse trajectory and brain wave data of the participants when using the computer through a combination of software and hardware,so as to establish a unique database.After data cleaning,standardization and other preprocessing operations,The data is divided into equal lengths to obtain the final experimental sample data.(2)Based on mouse data,this paper uses deep learning algorithms to design three network models(1D-CNN-Mouse,1D-LSTM-Mouse and 1D-CNN-LSTM-Mouse)that can be used for emotion recognition and analysis.Then train these networks separately.The results show that the accuracy rates of the three models on the training set are 94%,93%,and 97%,respectively,and the accuracy rates of the test set are85.3%,86.4%,and 87.2%.Among them,1D-LSTM-Mouse compared with1D-CNN-Mouse network,the accuracy of the training set has decreased,but the accuracy of the test set has increased,and the generalization ability of the network is better.The CNN and LSTM network fusion design model(1D-CNN-LSTM-Mouse)has the highest accuracy on both the training set and the test set,which can better judge the emotional state of the participants.(3)Design two deep learning models(1D-CNN-EEG,1D-CNN-LSTM-EEG)for brain wave signals.The recognition accuracy of the two networks on the training set is close to 100%,and the accuracy on the test set is as high as 93% and 96%,respectively.At the same time,this article attempts to combine the reinforcement learning algorithm to classify the emotions of the EEG data,that is,the 1D-CNN-EEG and 1D-CNN-LSTM-EEG models are combined with the reinforcement learning to form two deep Q network models(1D-DQN-EEG and 1D-DRQN-EEG).The classification accuracy of the model on the test set is 91.1% and 93.2%.The experimental results show that the model based on EEG data can basically correctly identify the emotional state of the subjects,and at the same time increase the reliability of the mouse trajectory to realize emotion recognition.In summary,emotions can be reflected in different data.For the experimental results of this article,compared with other single models,the combination of CNN and LSTM networks has more obvious advantages,and the model classification is more accurate and deep Learning algorithms have better results than reinforcement learning algorithms.This article basically realizes the emotion classification experiment based on mouse trajectory and EEG signal.The experiment proves that the method of judging the subject's emotion by collecting mouse trajectory is feasible and effective.