Eeg-based Emotion Recognition And Depression Diagnosis

Emotion plays an important role in people's daily life.Since the concept of "affective computing" emerged,researchers in related fields have been working to integrate emotion into human computer interaction.To achieve this goal,the accurate recognition of emotion is the foundation and important stage.Electroencephalography(EEG)is a commonly used signal for recording brain activity.Because of the advantages of objectivity,noninvasive,low cost and high temporal resolution,EEG is suitable for the study of emotion recognition.This thesis proposes a method for EEG emotion recognition based on convolutional neural network(CNN),which converts EEG signals into multiple feature maps extracting frequency features of EEG and preserving the relative position and distance information between electrodes to a certain extent.Then,a CNN model is constructed to finish emotion recognition task.The performance of the method is evaluated on DEAP dataset.The result shows that the classification accuracy is 81.64%for valence and 80.25% for arousal.In today's society,where the life is fast-paced and stressful,people should pay attention to their not only emotional state,but also mental health.Nowadays,depression has become a major health burden worldwide.This thesis proposes a method for EEG depression diagnosis based on CNN and feature fusion.Our method extracts frequency and spatial features of EEG signals through CNN,and then carries out feature fusion and finish classification.The performance of our method is tested on MODMA dataset.Empirical results show that the accuracy of depression recognition is 92.22%.In addition,our model has few parameters and low computation.Thus,it can be used on wearable devices to enable people to timely and accurately realize their mental health states in daily life.