The Research of BDPCA Identifying Emotion by EEG

Abstract

Electroencephalogram (EEG) data contain wealthy information about the brain’s and body’s pathology and physiological state. It’s not easily to identify the truth that EEG contained. The unsupervised learning method don’t need to take label by human. Without subjective feeling, it greatly improve training accuracy. In this current paper, adopted improved Density Peak Clustering Algorithm (DPCA) to train EEG data. To solved the problem that difficult to determined cluster center number, Bayesian Information Criterion(BIC) was introduced. The algorithm was verified feasibility that in EEG processing by experiment which divided fatigue state level in lab. And used SJTU Emotion EEG Data set (SEED) identifying different emotions. Compared with other cluster algorithms, BDPCA accuracy totally raised about 5%. And BDPCA behavior was steadier in different emotion types.