Emotion recognition is an important research topic in human-machine interaction field, which can be applied to medicine, education, psychology, military, and other areas. EEG (electroencephalogram) signal is mostly used among various indices of emotion recognition. Higher accuracy of emotion classifiers benefits from extracting the most relevant and discriminant features of emotion states. The paper surveys EEG features that are widely used in current emotion recognition study, introducing EEG features from four viewpoints, i.e. time domain, frequency domain, time-frequency domain and space domain. SLDA algorithm is imported to three public EEG-emotion datasets — SEED, DREAMER and CAS-THU — in order to evaluate feature capabilities distinguishing emotion valence. Existing problems and future investigations are also discussed in this paper.
Copyright 2019 Science China Press Co., Ltd. 《中国科学》杂志社有限责任公司 版权所有