Research On Human Sitting Posture Recognition Method Based On Membrane Pressure Sensor Group

With the rapid development of the Internet of Things industry,portable medical rehabilitation equipment and daily health monitoring equipment also have a general trend of developing through the Internet of Things.Daily health monitoring equipment such as smart bracelets and body fat scales have slowly begun to integrate into people’s daily lives.With the vigorous development of the Internet,the Internet of Things,ICT,big data,and artificial intelligence industries,due to the nature of the Internet industry,more and more people are beginning to work in the office for long periods of time.The sitting health problems caused by things like this have slowly begun to attract attention.In the field of human sitting posture recognition and monitoring,researchers have always been the focus of research.Therefore,this paper designs a human sitting pressure data collection terminal based on a membrane pressure sensor,and studies the human sitting posture recognition method based on the data obtained by the human sitting pressure data collection terminal.The main research contents of this article are:This paper first builds a set of data collection terminals for human sitting posture pressure.The terminal system uses thin-film pressure sensors to collect data and uses the STM32F103RCT6 controller as the main control chip.The controller packs the collected pressure data into a fixed protocol through analog-to-digital conversion and composite filtering,and finally uploads the data to the upper computer UI interface via Bluetooth.This paper develops the upper computer UI interface based on PyQt5 and python.The functions implemented include data receiving,storage,calling,time series drawing and human sitting posture recognition.This paper designs an algorithm for the selection of the arrangement of the film pressure sensor group.Through the similarity of time series,the arrangement and selection from array pressure sensors to non-array pressure sensors are solved.Conduct a consistency experiment on the pressure sensor and calibrate the sensor value.Whenarranging the membrane pressure sensor,firstly design experiments and calibrate the sitting posture,and then collect the experimental data of five sitting postures including forward,backward,left,right and normal sitting.Finally,the pressure sensor arrangement selection algorithm based on time similarity is used to eliminate the sensors with the most similar data in each area.Finally,the sensor group with the determined arrangement is collected again through experiments to collect pressure data,and the classification and recognition method based on time series is adopted.First,convert the collected pressure data into time series data and extract the shapelet in the time series,and then use the traditional machine learning classification algorithm(SVM,KNN,Random Forest)to classify the extracted shapelet.Design a comparative experiment,perform PCA dimensionality reduction on the collected pressure data of 12 sensors,and then directly use SVM,KNN,random forest and other algorithms to classify the dimensionality reduction data to obtain the accuracy of the test set under this method.Comparative experiments show that after the collected pressure data is converted into time series and shapelet is extracted,then using SVM,random forest and other methods can significantly improve the accuracy of the test set classification.In particular,the random forest algorithm has the highest recognition accuracy on the test set among all methods.