Research And Application Of Gesture Recognition Based On Surface EMG Signal

Surface EMG signal is a weak biological signal produced by nervous system controlling muscle contraction.Surface EMG can represent the movement of human body and is a way to study the biomechanics of human muscle.In recent years,surface EMG has been widely used in biomedicine,rehabilitation engineering and artificial intelligence.Its research has led to the development of human-computer interaction and hardware devices.Surface EMG signal has important research value in gesture recognition.By the study on Surface EMG,patients with physical disabilities perform limb movements effectively through prostheses,and hopefully return to a normal state of life.The study can make the communication between deaf and healthy people easier,so that people can understand the communication intention between the deaf and the healthy easily.It can also be used in somatosensory games and intelligent devices to make people's lives more convenient and intelligent,and it can improve the quality of life.In this paper,in order to recognize different gestures effecti-vely,two methods are proposed for SEMG gesture recognition.One is based on variational mode decomposition(VMD)method,the other is based on variational mode decomposition and supervised local linear embedding algorithm(SLLE).These two methods are used to process the gesture signal.The results show that the methods in this paper have a good recognition effect on the gestures.The research contents of this paper are as follows:(1)In this paper,a method of SEMG gesture recognition based on variational mode decomposition is studied.The intrinsic mode function(IMF)is obtained by decomposing the surface EMG signal by VMD method.New effective discriminant features are extracted by multivariate multi-scale entropy and fuzzy entropy.Finally,the kernel parameters of the radial basis function and the optimal parameters of the penalty factor of the support vector machine are found by using the Grid-Search algorithm.Through the experiment,the classification accuracy of hand holding cylinder,hand heavy object,hand holding ball body and hand holding small object are 87.5%,81.3%,95.0%and 92.5%respectively.(2)In this paper,a method of gesture recognition based on VMD and SLLE is proposed.VMD and fast independent component analysis(FastICA)are used to preprocess and denoise the surface EMG signal.The supervisory locally linear embedding algorithm is used to reduce the dimension of the surface EMG signal to obtain the low-dimensional features of the data.Finally,support vector machine is used to classify and recognize gestures.From the experiment,the recognition rate of the six gestures of holding a cylinder,carrying a heavy object,holding a flat object,holding an object in the palm,holding a sphere in the hand and holding a small object in the hand are 100.00%,97.33%,92.00%,81.33%,96.00%and 96.00%respectively.The results show that the proposed algorithm has good recognition effect and good noise robustness.