Gait Analysis Based On Continuous Plantar Distributed Pressure Measurements

Walking is the most common form of human movement,and it is also a relatively complex and periodic movement,while gait analysis is a method to measure various signals of human walking and analyze spatiotemporal characteristics and kinematic characteristics according to the changes of signals.As an important physiological and mechanical information of limb movement,plantar pressure signals are closely related to limb movement patterns.Different gait movements generate pressure signals with different characteristics.Different gait motions can be identified through the related relationship between pressure signals and gait patterns.Therefore,this paper carries out a research topic of gait analysis based on continuous distributed plantar pressure measurements.However,because the single pressure sensor cannot describe all the information of human gait,the MEMS inertial sensor is added to the pressure sensor to realize and analysis human gait.This paper first builds a human gait data acquisition system.The gait data acquisition system consists of gait information perception module,stack-based data acquisition slave module,main module which is including wireless data transmission unit and computer processing module.In this paper,the IMU inertial sensor is fixed to the heel to sense the posture information of the human body during normal movements,and the four membrane pressure sensors are placed on the thumb,the first metatarsal bone,the fourth metatarsal bone and the heel to sense the human body travel Pressure information.First,this paper designs a gait data filtering algorithm,blank data segment removal and gait segmentation algorithm and analyzes the characteristics of sensor data such as acceleration,angular velocity,pressure,attitude,etc.in an asynchronous mode during a gait period.Based on the extraction of 75 gait statistical features and 6 gait physical features in a gait period,these features are normalized.Before the recognition,it is considered that there are many initial features,and the training result of the SVM algorithm model depends on the selection of features.Therefore,the relief F algorithm is used here to sort all the gait features according to the feature weights,and then form different sorted features.Feature subsets,and use the KNN algorithm to evaluate the pros and cons of different feature subsets,and finally select the optimal feature subset to train the SVM algorithm model.When using the SVM algorithm to train the model,the appropriate kernel functionparameters and The penalty factor plays a major role in the accuracy of the model,so this paper uses genetic algorithms to optimize the parameters.In this paper,8 kind of gait patterns are defined and recognized in a expriment.In this experiment,the genetic algorithm is used to optimize the SVM algorithm.After parameter optimization,the feature set composed of 33-dimensional features after feature picking is better recognized,and the recognition rate can be achieved99.41%.