Research On Gait Phase Recognition Method Based On MEMS Inertial Sensor

Human walking process has the characteristics of periodicity and regularity,and gait phase refers to the time of typical posture change in the process of walking.It is an important indicator reflecting gait habits,age,health status and other factors affecting body coordination.Gait phase analysis is an important index to evaluate the skeletal muscle and nervous system of human body by combining the disciplines of kinematics and physiology.With the development of pattern recognition and artificial intelligence,gait phase recognition(also known as gait recognition)is gradually applied to various fields of social life,such as motion detection,medical rehabilitation,identity recognition,bionic robot,etc.Aiming at the problems of single acquisition signal,poor portability and real-time in the current gait recognition system,this paper designs a set of embedded wearable and real-time computing gait recognition system.The system is based on MEMS(Micro Electro Mechanical Systems,MEMS)inertial sensor to obtain and analyze the gait information,calculate the gait time parameters and position parameters,recognize the gait phase,and display the motion visualization.The main research contents are as follows:1)In view of the limitations and poor real-time performance of traditional gait data acquisition system based on video image,EMG signal and single sensor,a multi-sensor,wearable and portable gait data acquisition platform is designed and implemented to collect gait information in real time and synchronously.The acquisition platform is based on the MEMS inertial signal acquisition module to achieve the synchronous acquisition of acceleration,angular velocity and other multiple signals;secondly,in order to improve the real-time identification,this topic selects the raspberry pie 4B as the main control system,which can meet the embedded gait signal high-speed,accurate communication and data high-speed operation.2)Based on the adaptive threshold method,the gait phase is detected and the gait time and position parameters are calculated.Aiming at the error of gait phase detection caused by the change of motion form and random noise,this paper studies the adaptive threshold gait phase detection algorithm,improves the accuracy of gait phase detection,calculates the gait time parameters;secondly,aiming at the problem of error accumulation after acceleration information integration,the zero speed update detection algorithm is designed to correct Speed information,calculate gait position parameters.3)The feature values of five gait phases of multi-sensor are extracted,and thefive gait phases are classified by support vector machine classifier,and the visual display software is designed to show the recognition results.In view of the poor generalization ability of visual classification of gait phase,this paper proposes a gait phase classification and recognition method based on support vector machine algorithm;aiming at the problem of motion state visualization,the 3D model of human body is developed and designed based on Open GL to realize the visualization of human gait data.In this paper,gait data acquisition system is used to collect gait data.The mean values of five gait phases of multi-sensor heel landing,toe landing,heel off ground,toe off ground and swing midpoint are extracted as phase features for classification and recognition.The classification accuracy is 90.0%,83.3%,81.7%,90.0% and93.3%,respectively.The effective recognition of gait phase is realized in depth study of rehabilitation treatment and application of patients with movement defects laid the foundation.