Dynamic hand motion recognition based on transient and steady-state emg signals

Traditional motion recognition methods developed on the basis of steady-state electromyography (EMG) signals cannot well deal with the transient EMG signals. Thus, a large quantity of incorrect classification outputs would be introduced during the dynamic stages of the motions. In order to achieve a high-accuracy recognition system especially for dynamic motion control, this paper combines the transient EMG and the steady-state EMG signals together for training the recognition system. A threshold decision method is utilized in the time-domain feature space to collect the combined EMG signals. Besides, a statistical classifier named support vector machine is adopted in the online recognition procedure to distinguish the motion types. Experiments are conducted to quantify the classification accuracy and response delay of the recognition systems, and compare these with traditional steady-state EMG-based methods. The results indicate that the recognition accuracy can be greatly improved and the detection delay of the motions can be significantly compressed in the transient stages of the hand motions. The method shows a promising application in the dynamic motion control of dexterous prosthetic hands in the future.