Research On Human Body Movement Intention Recognition Based On Surface Electromyography

The surface electromyography(s EMG)of the human body contains a wealth of information about the state of human motion.It has broad application prospects in human-computer interaction systems such as flexible power-assisted clothing and wearable exoskeleton.In this type of human-computer interaction system,perceptual control is realized based on the recognition of human movement intention.The perception and recognition of human movement intention are mainly divided into two categories: biomechanical perception based on joint angle,angular velocity,three-axis acceleration,and plantar information,and bioelectric signal perception based on electromyography,brain electricity and other bioelectrical signals.The goal of this research is to use myoelectricity to realize the perception and recognition of human movement intentions,and to provide multiple input information for the control of flexible power-assisted clothing.The s EMG is used to complete the pattern recognition of human movement intention,and the mapping relationship between the s EMG and the movement posture of the limbs is established.The ultimate goal is to realize that the continuous motion estimated by s EMG can be used as a reference control signal.Ensure the safety of the human body when wearing the flexible power suit and the natural coordination of the human-computer interaction process,and realize the suppleness of the entire control.The main work of this paper includes the following parts:First,determine the layout of lower limb sensing.Study the internal relationships among the lower limb muscles,movement,and joints of the human body.Through research,analysis and modeling,the biceps femoris,lateral femoris,medial femoris,gastrocnemius,and tibial anterior muscles were selected as the source of electromyographic signals,and the lower limb sensing layout plan was determined.Design the experimental scene to collect the surface EMG signals of the lower limbs of the human body walking on the ground,going up/down stairs,going up/downhill,and squatting.Second,the preprocessing of the surface EMG signal is performed.The surface EMG signal itself has the characteristics of weak signal,low frequency,and susceptibility to interference.In the process of signal acquisition,it is susceptible to interference from various bioelectric signals generated by human tissues and inherent noise of instruments and equipment.In this paper,the Butterworth and wavelet threshold denoising methods are designed to preprocess the EMG signal,and the signal is filtered and denoised.The EMG signal obtained by this preprocessing has a high signal-to-noise ratio and a small root mean square error.Finally,the human body movement intention identification is carried out.In terms of human movement intention pattern recognition,the EMG feature value is extracted,and the BP neural network is used for movement intention pattern recognition.The average recognition rate is 85%.After the principal component analysis method is used to reduce the dimensionality of the eigenvalues,on the one hand,the complexity of the network is reduced,and on the other hand,the average recognition rate of the BP neural network is 90%.In the aspect of human EMG motion posture mapping,two methods,support vector regression and Gaussian process regression,are used to perform regression mapping on the posture of the upper and lower limbs.In the process of algorithm comparison,the support vector regression algorithm is better than Gaussian process regression in human motion regression mapping.For walking on flat ground,up/down stairs,and up/down slope,the correlations of the five motion mappings are all greater than 80%.