Pattern Recognition Of Lower Extremities Based On SEMG

In recent years,with the rise of AI technology,human behavior pattern recognition technology has been widely concerned.Based on wearable devices,human lower limb behavior pattern recognition technology has broad application prospects in rehabilitation medicine,pedestrian navigation,virtual reality and other fields.Human muscle electrical signals contain a lot of information which is directly related to human motion,and the continuous acquisition of signals is convenient and fast,so it is widely applied.The recognition of human lower limbs behavior pattern based on surface electromyography is studied in this paper.(1)To build a human muscle electrical signal acquisition system.Combining the distribution and function of the muscles of the lower extremities and the parameters of the human gait,we select the muscle blocks that can identify the main movement patterns of the lower limbs,and collect the muscle electrical signals.(2)Band pass filtering and wavelet reconstruction are used to remove the noise of muscle electrical signals.The method based on the wavelet reconstruction is obtained by using the threshold limit and the weighted reconstruction after the wavelet decomposition of the signal.The muscle electrical signal processed by this method can more clearly reflect the human gait characteristics.The signal is smoothed and the active segment is extracted.(3)SEMG signal is extracted and dimensionality reduced.The characteristics of wavelet feature parameters and nonlinear dynamic parameters are extracted,and the dimensionality of the extracted high dimensional eigenvectors is reduced.On the basis of preserving the original characteristic variance,the operation efficiency and recognition accuracy of the subsequent algorithm are improved.(4)Design BP neural network,give the concrete calculation process and method of forward propagation and error reverse transfer,and explain the method and selection principle of gradient examination,parameter tuning.The structure design steps and training methods of error back propagation neural network are given.The algorithm of K mean algorithm for turning recognition is designed.(5)A comprehensive experiment involving multiple lower limb motion patterns is designed.The necessity of the dimension reduction for the recognition algorithm is verified.The selection of parameters such as regularization parameters and learning rates in the network design process is illustrated by the experimental data.The back propagation neural network is applied to the recognition of the body movement pattern of human lower limbs in direct,upstairs,downstairs,running and backward.The rate of recognition is higher than 80%,and the comprehensive recognition rate is higher than 75%.The K mean algorithm is applied to the motion pattern recognition of straight line,left turn and right turn,and the recognition effect is similar to that of neural network.