抄録
This paper presents motor imagery classification using Bayesian logistic regression and its application to sensor-based wheelchair control. First, to classify four kinds of motor imageries (that corresponds to four actions:go left, right, straight and stop in wheelchair control) in electroencephalogram (EEG) signals, we provide a new practical approach combining common spatial pattern (CSP), Bayesian logistic regression and automatic relevance determination (ARD). In addition, a well-balanced control scheme for considering both the EEG classification result (from the proposed practical approach) and obstacle information (from a two-dimensional laser scan sensor) is applied to wheelchair direction control. The experimental results demonstrate the utility of our approach.
