Natural muscular recruitment during reaching tasks to control hand prostheses

Several efforts have been carried out in the past to develop hand prostheses controllable by the voluntary activities of amputees and able to restore lost hand functions. Myoelectric prostheses represent a viable clinical solution thanks to non invasiveness and recording easiness of electromyographic signals (EMGs). The control of multi-degree of freedom (DoFs) prostheses in an effective and natural way is currently limited by the need of a complex pattern recognition approach and the use of "non homologous" muscles. Beside solutions based on pattern recognition in the central nervous system, the use of electrodes implanted into muscles or peripheral nerves or targeted muscle reinnervation, a possible solution for the development of a more "natural" EMG-based control strategy could be the discrimination of grasping tasks during the reaching phase. In this pilot study, experiments with three able-bodied subjects have been carried out in order to verify whether this strategy can be implemented. A support vector machine algorithm has been used for the prediction of different grip types during reach to grasp movements using EMG activity of distal and proximal upper limb muscles. The information coming from proximal muscles helped to increase robustness in the classification tasks.