Research On Sitting Position Recognition Method Based On Array Sensing Technology

With the development of society and the progress of science and technology,the proportion of people working in the office environment is increasing,so the time of social people in sitting position is also increasing.However,most people's sitting posture belong to bad sitting posture,and long time bad sitting posture is one of the most important reasons leading to a series of health problems.One of the ways to solve this problem is to design a system that can monitor the sitting posture of human body in real time,so that users can correct their bad sitting posture in time.The core function of this system is real-time and accurate sitting posture recognition.Now,there are two methods for sitting posture recognition,one is based on machine vision or the other is based on simple mechanical characteristics.The former is to learn from the image captured by the camera,and then to recognize the sitting posture by image processing or machine learning.However,this method depends on the lighting conditions and shooting angle of the photograph,its robustness is difficult to guarantee,and it also involves some privacy issues.The latter is to arrange mechanical sensors at the key points of the human body's stress surface,and use the characteristic information collected by these sensors to recognize the sitting posture.This method has the characteristics of poor comfort,low accuracy and weak robustness.As an exploration and improvement,we study the method of sitting posture recognition based on array pressure sensor technology for real-time accurate sitting posture recognition.In our research,we first design and implement the corresponding data acquisition system.Its main work includes two aspects,hardware system design and software system design.The hardware system includes 32*32 array pressure sensor,processing circuit and PC.The software system includes the lower computer acquisition program,communication program and upper computer processing program.Then,we use the data acquisition system to collect a large number of data of the seating pressure distribution on the human-chair interface for many participants.Finally,three sitting posture recognition algorithms are constructed based on these data,sitting posture recognition algorithm based on simple pressure distribution characteristics,sitting posture recognition algorithm based on BP neural network and sitting posture recognition algorithm based on convolutional neural network(CNN).After testing and analyzing these three algorithms,we find that the sitting position recognition algorithm based on simple pressure distribution features has fast operation speed but poor stability and robustness,the sitting position recognition algorithm based on BP neural network has certain accuracy and robustness,the sitting position recognition algorithm based on CNN has strong robustness and high accuracy,and it is an excellent sitting position recognition algorithm.This paper combines the array pressure sensor with CNN algorithm to recognize the sitting posture,overcomes the dependence of the sitting posture recognition method based on image machine vision on the environment,and solves the problem of low recognition rate of the traditional method based on simple mechanical characteristics.The method has high accuracy,strong robustness and good stability.