Research On Human Activity Recognition Technology In Wireless Network Environment Changes

Movement is one of the most basic characteristics of human life.These seemingly simple human activities contain a wealth of information.Analyze human activities,and then fully explore the hidden practical information that can serve humanity more humanely.In recent years,more and more researchers have devoted themselves to the field of human activity recognition.With the continuous deepening of research,bottlenecks in the development of some traditional human activity recognition methods have begun,and various irreparable defects have begun to appear.The rapid development of wireless network technology has brought dawn to human activity recognition.This kind of non-contact sensing of human activity technology has been favored by researchers in this field for its unique advantages.Through ordinary commercial wireless devices,fine-grained channel state information(CSI)containing human activity information can be easily obtained.This human activity recognition method does not need to wear special equipment,and can accomplish passive human activity recognition without invasiveness.Moreover,the cost is low,it is not easy to be affected by external conditions such as light,and no visual blind spot,high recognition accuracy and good robustness,so it has a wide application prospect.Thesis paper designs and completes a passive human activity recognition system based on CSI.In the indoor environment,it can complete the recognition of four types of human activities such as sitting,walking,crossing obstacles while walking,and tripping while walking,as well as the detection of Cerebellar Ataxia and head tremor.The research and analysis of the system are mainly embodied in the elaboration of the recognition mechanism of human activity,the extraction of human activity information,and the evaluation of human activity recognition results.In order to effectively complete human activity recognition and disease detection.Firstly,the principle of human activity recognition based on CSI is explained,and the feasibility of human activity recognition using this technology is analyzed.It also outlines the two types of diseases that the system will detect.By analyzing the abnormal performance of two types of diseases in human activities,it is found that using this technology to detect diseases is feasible.Then,a method for extracting non-contact human activity information is designed and implemented.By analyzing the CSI data format,CSI amplitude information and phase information are obtained.After comparing and analyzing the sensitivity of the two types of information of amplitude and phase to human activity,the amplitude information was selected to perceive human activity.At the same time,a series of processing was performed on the data,including using standard deviation to select subcarriers,using Hampel algorithm to remove outliers,using wavelet filtering to remove noise from the data,and using time-domain analysis to extract time-domain features.Finally,a large number of experiments were performed to verify the effectiveness of the designed system.KNN,decision tree and SVM classification algorithms were used to complete the classification and the experimental results were evaluated.The experimental results show that the proposed system achieves 97% accuracy in recognition of four types of human activities such as sitting,walking,crossing obstacles during walking and tripping while walking.The accuracy of the heel-knee-shin test and the rapid alternating movements test reached 98% and 99%,respectively,and the head tremor detection accuracy reached 95%.