Research on Classification and Recognition of Emotional EEG Signal Based on Extreme Learning Machine

Abstract

There are many research methods of emotion recognition, and the emotion recognition of EEG signals is one of the important researches in the field of emotion computing. Extreme learning machine can recognize more generalized features with fewer training parameters. Therefore, this paper constructs a power spectral density-extreme learning machine classification model and applies it to the classification and recognition of emotional EEG signals. The experimental paradigm was designed and the emotion-related EEG signals of 10 subjects were collected, and the PSD characteristics of the corresponding frequency bands of their EEG signals in the three emotions were extracted, and the model was used to achieve specific and cross-subject evaluation of emotional EEG signals. Positive, neutral, and negative three categories are distinguished. Among them, the accuracy of single body recognition is up to 96.17%, and the average accuracy is 81.72%; the cross-subject recognition accuracy is 76.82%, which verifies the combination of this feature and the classification method. Effectiveness in the field of emotional EEG signal recognition.