The Research On Recognition Of Finger-movements Based On MMG

In recent years, the majority of disabled people desire a life with high quality as the improvement of living standards, so the prosthetic products that can help the disabled to make up the limb defects have a huge market demand. Thus the control of Bionic Intelligent Prosthetic attracted more and more attention and has made some progress, among them the most mature technique is the EMGs Prosthetic and it occupied the major market of this field. Due to the high price of EMGs Prosthetic, MMGs Prosthetic came into being and has considerable application prospects.In addition, the design and development of embedded real-time control system is the key factor for the Bionic Intelligent Prosthetic technology transformed to products.Nowadays, the study on MMGs Prosthetic mainly focused on the movement patterns of entire hand, so in order to improve the dexterity of prosthetic, further research is needed on the movement patterns of the fingers, an approach to the identification of six finger movements is proposed. The method adopts four channels MMG signal from forearm, and fifty time-domain and frequency-domain features are extracted from each MMG signal channel. In the feature optimization, Isomap and diffusion maping are used for dimensionality reduction and then optimized feature vector representing different single and combined finger movements are analyzed via k-Nearest Neighbor (KNN) classifier for motion recognition. Finally, the study transplants the pattern recognition algorithm to embedded real-time control platform. The paper transplants the pattern recognition algorithm in Matlab environment to the DSP core and run in it via the transplant technique based on Simulink and CCS.The result showed that the average recognition accuracy of six finger movements has reached94.25%±1.6%and95.48%±2.47%, and there exists a positive and high correlation between the energy of finger movements and the accuracy. In order to verify the effectiveness of the algorithm, the paper applies Virtual Reality technique to build the prosthetic model of multi-finger and multi-degree and sends the control signal to the virtual reality environment interface through asynchronous serial ports to implement the regular movements. The study shows that the program of using MMGs to recognize the finger-movements is feasible and has broad application prospects.