The goal of this research involving a motor imagery brain-computer interface paradigm is to assess the possibility of enhancing the classification rate handling a feature vector based on the modulation of electrophysiological brain activity in specific bands. A new amplitude modulation energy index of the cerebral rhythms is proposed as feature vector concept. The method is proven on a public database and on a set of electroencephalographic data recorded in our own laboratory. In both cases, only eight electrodes are used in order to reach high performance classifying rates. The discrimination of motor tasks (imagination of right and left hand movements) is analyzed by means of five classifiers: support vector machine, k nearest neighbor, linear discriminant analysis, quadratic discriminant analysis and Mahalanobis distance based classifier. For our database, the medians of the classification rates for two of classifiers are very high (94.62 % -97.76 %) when some rhythms are modulated in theta and alpha bands. Significantly higher classification rates reported herein (greater than 90 % for both of the databases) compared with classifiers trained on the other features prove that the index may be very useful for highlighting the modulation found in certain bands of the EEG rhythms.

No abstract

No abstract

No abstract

No abstract