Real-time eye blink suppression using neural adaptive filters for EEG-based brain computer interface

This paper presents an adaptive noise canceller (ANC) using neural network for real-time removing the eye blinks interference from the EEG signals. Conventional ANC filter is based on the linear model of the interference. Such a naive linear model provides poorer predictions for biomedical signals. In this work, we employ artificial neural network for modeling the interference. The reference signal was collected by using an electrode placed on the forehead above the eye compared with the conventional eye blink recording which uses two electrodes, one placed above and the other below the eye. The reference signal is also contaminated by the EEG. To reduce the EEG interference, the reference signal is first lowpass filtered by a moving average filter and then applied to the ANC. The results show that single electrode recording provides satisfactory reference input correlated with the noise in the primary signal. The neural ANC is trained with one segment of data and evaluated with the rest of data. It is found that neural ANC provides a perfect cancellation of eye blink interference. Moreover, the base line wander caused by eye movement has been reduced.