Decoding of wrist and finger movement from electroencephalography signal

The emergence of brain-computer interfacing has made the control of robots through thought a reality. Such real-time application calls for fast processing and accurate classification of brain signals. In this paper, we address the two-level classification of motor imagery signals, where the user differentiates between clockwise/ counter-clockwise movement of wrist and the opening/closing of the fingers. For this purpose, parameters of adaptive autoregressive (AAR) models and Extreme Energy Ratio criterion (EER) are employed as features, which are fed to standard classifiers for comparison. It concludes the features extracted based on EER, selected by sequential forward search and classified using radial basis function kernelized support vector machine, provides optimum performance of the classification process for implementation in real-time scenario, with an average accuracy 90.24% and a time complexity of 8.2449 seconds.