A multiclassifier system with dynamic ensemble selection applied to the recognition of EMG signals for the control of bio-prosthetic hand
The paper presents a concept of hand movements recognition on the basis of EMG signal analysis. Signal features are represented by coefficient of autoregressive (AR) model, and as classifier the original multiclassifier systems with dynamic ensemble selection are applied. The performance of the proposed methods was experimentally compared against three classifiers using real datasets. The systems developed achieved the highest overall classification accuracies demonstrating the potential of dynamic classifier selection for recognition of EMG signals.