Accurate motor imagery based dry electrode brain-computer interface system for consumer applications

Abstract

The most common brain-computer interface (BCI) systems use electroencephalographic (EEG) signals to communicate human cognitive or sensory-motor brain activities. Those non-invasive BCI systems rely on large number (up to 128) of wet (using conductive gel) electrodes for higher detection accuracy and good signal to noise ratio (SNR). They are studied and designed primarily with focus on medical applications. The electrodes are usually mounted on a special cap and connected through multiple wires. The proper positioning of the cap requires assistance and takes significant amount of time. In this work we review the principles for EEG signal processing and feature extraction most suitable for applications in consumer electronics. Further, we propose a motor imagery brain-computer interface (BCI) based system, using only two active easy to set dry electrodes connected wirelessly with a consumer electronic device. The proposed system relies on the optimal use of event related synchronization (ERS) and desynchronization (DRS) across three distinct EEG frequency bands in order to improve the detection and reduce the training time to only 10 sec. We present our ongoing research investigating the detection accuracy with different signal preprocessing techniques and feature extraction methods. The proposed system aims at making brain-computer interfaces popular with consumer products, providing a more natural human computer interaction (HCI).