Real-time myoelectric decoding of individual finger movements for a virtual target task.

This study presents the development of a myoelectric decoding algorithm capable of continuous online decoding of finger movements with the intended eventual application for use in prostheses for transradial amputees. The effectiveness of the algorithm was evaluated through controlling a multi-fingered hand in a virtual environment. Two intact limbed adult subjects were able to use myoelectric signals collected from 8 bipolar electrodes to control four fingers in real-time to touch and maintain contact with targets appearing at various points in the flexion space of the hand. In these tasks, subjects achieved accuracies of 94% when target regions extended +/- 11.5 degrees about a target angle and 81% when the target region extended only +/- 5.75 degrees about the target angle. The real-time virtual system provides a practical and economic way to develop and train algorithms and amputee subjects using dexterous prostheses.