Brain machine interface system automation considering user preferences and error perception feedback

This paper addresses the problem of mental fatigue caused by prolonged use of Brain Machine Interface (BMI) Systems. We propose a system that gradually becomes autonomous by learning user preferences and by considering error perception feedback. As a particular application, we show that our system allows patients to control electronic appliances in a hospital room, and learns the correlation of room sensor data, brain states, and user control commands. Moreover, error perception feedback based on a brain potential called error related negativity (ERN)-that spontaneously occurs when the user perceives an error made by the system-was used to correct system's mistakes and improve its learning performance. Experimental results with volunteers demonstrate that our system reduces the level of mental fatigue, and achieves over 90% overall learning performance when error perception feedback is considered. Note to Practitioners-This paper suggests a new approach for designing BMI systems that incorporate learning capabilities and error perception feedback in order to gradually become autonomous. This approach consists in learning the relationship between sensing data from the environment-brain and user actions when controlling robotic devices. After the system is trained, it can predict control commands on behalf of the user under similar conditions. If the system makes a mistake, user's error perception feedback is considered to improve the learning performance of the system. In this paper, we describe the methodologies to design and build hardware and software interfaces, acquire and process brain signals, and train the system using machine learning techniques. We then provide experimental evidence that demonstrates the effectiveness of this approach to design BMI systems that gradually become autonomous.