EEG processing in emotion recognition: inspired from a musical staff

Common electroencephalograph (EEG) features have problems such as poor intrinsic correlation between characteristic quantities, low signal reproducibility and large data storage capacity, which lead to poor emotion recognition. To solve this problem, this paper proposes an EEG music model based on a musical staff. Firstly, this paper constructs a multi-channel EEG sensor network to measure the EEG of an individual under different emotional states, and establishes an EEG-Emotion mapping library for the individual. Then, the EEG is transformed by adaptive segmentation of the time-domain EEG signal using a musical staff model. The time-frequency characteristics of EEG, such as amplitude, contour and signal frequency, are expressed quantitatively in a standardized musical space. The results show that, while retaining the time-frequency features of EEG, the model has an average similarity of 0.9769 before and after reconstruction, a compression rate of 57.18%, and an emotional state recognition rate that is 10.1% higher than traditional features. The brain wave music generated by the model, as a media, provides reference for people to understand the change of emotional state, and also provides a new technical idea for the subsequent use of EEG music for emotional induction.