Real-time performance of hand motion recognition using kinematic signals for impaired hand function training

Re-gaining the fine motor skills of hand is the ultimate goal of the rehabilitation for the stroke survivors and traumatic brain injured patients with chronic hemiparesis. The clinical outcomes with the traditional passive rehabilitation approaches are often limited and slow for impaired hand-function recovery. It is well known that actively involving the conscious efforts of patients into hand-function training would be critical for improving cerebral functional reorganization according to the brain plasticity theory. In this study, a training system for the rehabilitation of fine hand functions was developed based on a flexible data glove system and validated in able-bodied subjects. The real-time performance of the training system was assessed with a measure, motion completion rate, in seven healthy subjects. The results of this study showed that using the kinematic signals a high average offline classification accuracy (98.67%±1.51%) and sound real-time completion rate (89.17%±5.49%) could be achieved in able-bodied subjects, which suggested a promise of applying kinematic signals in hand-function rehabilitation training. The future works will be conducted in stroke patients to further validate the performance of the proposed system.