Design And Implementation Of Human Lower Limb Behavior Recognition System Based On Embedded Technology

Human behavior recognition is an important part of the research of dynamic exoskeleton robot system technology.Human behavior recognition has important significance for humancomputer interaction system research.The research content of this thesis focuses on the purpose of human lower limb behavior recognition.A human body lower limb behavior recognition system scheme combining multi-channel surface EMG signals and multi-channel trajectory(acceleration)sensor signals is proposed.At the same time,based on embedded technology,design and establishment is carried out.A portable human behavior recognition system based on multi-sensor data fusion.Based on the system,through the simultaneous acquisition and processing analysis of 2-channel surface EMG signals and 3-channel acceleration signals,the recognition accuracy of walking,jumping and up and down stairs is 95.08%±0.44,and the action decoding time is less than 300ms.The specific research content of this paper is summarized as follows:(1)Human body behavior recognition system implementation plan design.This paper designs software and hardware systems based on embedded technology to realize synchronous data acquisition,real-time preprocessing,active segment detection and data segmentation,feature extraction and classifier motion recognition classification of multi-channel surface EMG signals and multi-channel acceleration signals.(2)Algorithm design of behavior recognition system.For each link in the system implementation,the relevant algorithms are designed and implemented to realize the functions.In the data acquisition part,the dual-storage mode is used to realize the synchronization of different sensor data acquisition;in the data pre-processing stage,the filter is designed and filtered for the surface EMG signal;the active segment detection and data segmentation are performed based on the sliding window sample entropy algorithm.Active segment endpoint detection,applying a moving window with a certain degree of overlap to frame the active segment data;feature extraction,time domain feature extraction and downsampling for surface EMG signals and acceleration signals to obtain feature vectors;behavior classification,this paper The classifier design is based on artificial neural network to realize the recognition and classification of actions.(3)Sensor system design and selection.In this paper,the design and selection of the surface EMG signal and the acceleration signal are carried out respectively.The surface electromyography signal picking and signal conditioning circuit design is based on the AD8221 instrumentation amplifier.The WT901C485 module is used to realize the data acquisition of the acceleration signal.(4)Multi-sensor synchronous data acquisition system software and hardware construction.This paper designs a multi-sensor synchronous data acquisition system based on STM32F407 microprocessor to realize synchronous data acquisition of multi-channel surface EMG signals and multi-channel acceleration signals.At the same time,based on AD7606 design multichannel synchronous analog-to-digital conversion interface module for surface EMG signals;based on MAX485 design TTL-485 interface level conversion module,realize the communication connection between data acquisition system main control and acceleration sensor WT901C485 module,and then Realize data acquisition of acceleration signals.(5)Software and hardware construction of real-time data processing system based on embedded Linux.This paper is based on S5P4418+Linux embedded platform for data processing system hardware and system implementation.In the design of data processing software,multi-task concurrent implementation of data processing process based on multiprocess mode.(6)Lower limb behavior recognition system experiment.Based on the embedded software and hardware implementation of the designed behavior recognition system scheme algorithm,this paper designs the behavior recognition experiment to realize the functional test of the system.Based on the system,real-time identification of walking,jumping,stair climbing and stair behavior is obtained.A better result.