EEG Signal-Based Eye Blink Classifier Using Convolutional Neural Network For BCI Systems

Electroencephalogram (EEG) signals are recorded from the brain activity and it plays an important role in both of diagnosing brain diseases and investigating brain activity. This paper proposed a method to evaluate eye blinks based on EEG signals and a Convolutional Neural Network (CNN). In particular, the EEG signals with left and right blinks were collected using an Emotiv Epoc+ measuring system. A Savitzky-Golay filter is utilized to smooth the EEG signals for enhancing the performance of eye blink classification. From the filtered EEG signals, the CNN model is applied to classify the left or right blinks for evaluating the relationship between the eye blinks and brain activity. The experimental results showed that the proposed classifier is effective for classifying eye blinks based on the EEG signals with the its accuracy reached 92.9%. Moreover, the eye blink activities after the classification can be applied in the Brain-Computer Interface (BCI) system for controlling mobile platforms such as electric wheelchairs for disabled people.