Research On The Motion Pattern Recognition System Of Three-degree-of-freedom Electromyography Prosthesis Based On SVM

In recent years,the number of people with residual limbs has increased dramatically due to natural disaster,traffic accidents and other factors.For the upper limbs,the loss of the upper limbs will lead to a sharp decline in the physical and labor capacity of the disabled,making them unable to take care of themselves,causing serious trauma to their psychology and even endangering the whole family.However,with the advancement of technology,the maturity of information technology,and the development of rehabilitation medical services,there are a large number of myoelectric prostheses in the market.At present,most of these commercial myoelectric prostheses use threshold control,which has a high price,but there is no good control flexibility.Therefore,it is especially important to establish a mature motion pattern recognition system with the three-degree-of-freedom myoelectric prosthesis as the research object,combined with the new wavelet threshold algorithm and efficient pattern recognition method.According to the characteristics of surface electromyography signal,the signal conditioning circuit was designed.The four-channel sEMG signal acquisition instrument was used to complete the sEMG signal collection of the upper extremity tweezer,triceps,carpal flexor and ulnar wrist extensors.After digital filtering,a single function wavelet denoising method based on global wavelet domain is used to denoise the sEMG signal.Compared with the traditional wavelet threshold algorithm,the results show that the improved new threshold algorithm performs better in both quantitative and qualitative aspects.After pretreatment,the sEMG signal was analyzed from three aspects:time domain,frequency domain and time-frequency domain analysis.By comparing the effects of the three methods,the energy and variance of the wavelet packet coefficients are used as the characteristic parameters to construct the eight-dimensional feature vector.Based on the standard support vector machine(SVM)algorithm,in order to shorten the recognition time and improve the accuracy under the premise of ensuring its sparsity and applying the radial basis kernel function,a homogeneous second-order SVM algorithm is constructed by changing its slack variable and decision function,and the optimal surface is solved under the least constraint condition.At the same time,the particle swarm optimization algorithm is introduced to optimize the model.Finally,the actual sample data is used to compare the standard SVM and its improved model.The results show that the improved SVM based on particle swarm optimization has higher recognition accuracy and shorter recognition time for the six kinds of upper limbs.Finally,a three-degree-of-freedom electromyography prosthetic motion pattern recognition system was designed,including a four-channel sEMG signal acquisition instrument and a PC recognition software.The four-way EMG signal is collected by the collecting instrument.Then the signal is analyzed on the host computer to realize off-line training and recognition of the sample data,and the corresponding action information is displayed with a good man-machine interface.The system is designed according to the modular concept,which facilitates the maintenance and development of the system.