Facial Expression Classification Using EEG and Gyroscope Signals

In this paper muscle and gyroscope signals provided by a low cost EEG headset were used to classify six different facial expressions. Muscle activities generated by facial expressions are seen in EEG data recorded from scalp. Using the already present EEG device to classify facial expressions allows for a new hybrid brain-computer interface (BCI) system without introducing new hardware such as separate electromyography (EMG) electrodes. To classify facial expressions, time domain and frequency domain EEG data with different sampling rates were used as inputs of the classifiers. The experimental results showed that with sampling rates and classification methods optimized for each participant and feature set, high accuracy classification of facial expressions was achieved. Moreover, adding information extracted from a gyroscope embedded into the used EEG headset increased the performance by an average of 9 to 16%.