Research On Activity Recognition Method Based On Plantar Pressure And Acceleration Signals

Human activity recognition is an extremely critical technology in the field of bio-behavioral research and its main purpose is to recognize human activity by using some mathematical methods and models,and then provide some constructive advice and guidance for clinical rehabilitation and sports training.Although there are a large number of methods that can effectively accomplish the analysis of time series,the research on plantar pressure and acceleration signals is still an urgent problem to be solved in the field of human activity analysis.The current data on plantar pressure and acceleration signals are basically about time series,while complex networks and hidden Markov model are widely used for time series data because of their good theoretical basis and high flexibility as well as accurate experimental results.In this paper,we improve these two methods and propose two different processing methods about time series of human activities,at the same time,the performance and accuracy of the model are verified.The work of this paper is shown below.1.A human activity recognition method based on the multiplex limited penetrable visibility graph is proposed for the related properties of complex networks that can inherit time series effectively.This method converts the time series characteristics of human activity into complex network features and combines the characteristics of human activity to construct a two-layer limited penetrable visibility graph.The experimental results show that the method can effectively cluster different activities and then realize the recognition of different activities.2.Aiming at the characteristics of the hidden Markov model that can handle sensor time series,a hidden Markov model based on multiple Taylor regression is proposed.Using raw accelerometer data obtained from human inertial sensors in a health monitoring environment,the method is based on the joint segmentation of multidimensional time series using Hidden Markov Models(HMM)in a multiple regression context.The expectation maximization(EM)algorithm is used to learn the model in an unsupervised framework without the need for activity labels.The experimental results show that the proposed method is applicable to time-accelerated data.It is possible to allow for the segmentation and classification of human activities.3.For Matlab and Python,two powerful data analysis and processing languages,this chapter introduces the human activity analysis system designed based on these two programming languages,and uses Matlab GUI and Python Django to realize the development of the analysis system for two different research objects.At the same time,the framework foundation built by these two types of systems is briefly described,which provides feasible solutions for subsequent research and makes the application of human activity recognition technology possible.