Study On The Real-time Classification Of Hand Motion For Stroke Survivor Based On Physiological Signals

Active rehabilitation put great emphasize on the conscious effort of stroke survivors that is an important engineering means to promote their brain plasticity and improve the rehabilitation effect.To date,several proposed studies have proved the feasibility of hand motion intention based on pattern recognition methods in the offline analysis.However,the offline classification performance of multi-class motion is difficult to intuitively reflect the real-time performance.In this study,a real-time motion intention recognition method was studied for upper limb movement of stroke survivors based on multiple physiological signals.And the effectiveness of passive and active rehabilitation training was discussed.Firstly,the game-guided real-time rehabilitation training system was developed in our study,which includes linear discriminant algorithm(LDA),support vector machine(SVM)and template matching(TM)algorithm.The custom-made rehabilitation environment provides three main modules,i.e,signals acquisition and display,offline pattern recognition analysis,and game-guided real-time rehabilitation training.Secondly,the 40 channels surface electromyography(sEMG)signals from subjects' forearm were collected.LDA and SVM algorithm pattern recognition method were used to identify 18 functional movements of the their offline motion intention.The performance of offline motion intention recognition between the two algorithms was compared,namely,offline classification accuracy and classification efficiency.A real-time motion intention method with high classification accuracy and efficiency was found.Thirdly,in the current investigate,the 40 channels surface EMG and 14 channels data glove from subjects' forearm and hand were collected.The real-time classification performance of 15 finger movements was investigated by LDA algorithm and the TM algorithm with three different distance measures.We investigated the relationship that exists between classification accuracy and completion rate using linear regression analysis.Finally,the game-guided real-time rehabilitation training system was applied to stroke survivors.The original sEMG signals and EMG-Potential diagram were compared when subjects do the active and passive wrist extension/wrist Flexion rehabilitation training respectively.The four parameters of response time,completion time,completion rate and dynamic efficiency were used to accurately,quantitatively and objectively describe the survivors' limb control ability from the two levels of "reliability" and "validity".