Sitting Posture Sensing Technology Based On Neural Network

Sitting posture is one of the most common human postures,people spend most of the time in the state of sitting posture,especially for students and white-collar workers in the modern society.The bad thing is that most of the time people sit in the wrong posture.For a long time,it will not only bring discomfort to the body,but also lead to the occurrence of some diseases.Seriously reducing the quality of life.Whether sitting posture is correct or not determines people's health status to a certain extent.Bad sitting posture can bring great pressure on the spine,joints and muscles and the body will be led to a series of problems,diseases of cervical vertebra disease and lumbar disc herniation have the most high occurrence.However,in daily life,when most people at the time of work or study it is hard for them to notice the sitting state of their own,then it is difficult to be conscious to correct the wrong sitting posture,so we introduce the study of sitting posture judge system.The research of this system can guide people to adopt correct sitting posture and prevent diseases caused by incorrect sitting postures.At present,a video camera and image processing related content is mainly used to determine the classification of the sitting posture in domestic.The way of using the camera video can sometimes make people feel uncomfortable,involving some privacy issues.The body pressure distribution measurement and posture discrimination are closely related,but most of the existing domestic pressure distribution measurement system is to measure the plantar pressure distribution and there is little for sitting pressure distribution,so this project is designed to measure the pressure distribution of sitting posture discrimination system.The sitting posture discrimination system mainly consists of two parts: hardware part and software part.Pressure sensing element is a 5 × 10 array sensor based on Flexiforce,which is composed of a single point thin film pressure resistance sensor.The array sensor have many signal channels,in order to make the hardware circuit more convenient and make full use of resources,the project design a two stage multi-channel analog switch selection circuit to switch the signals of array sensor sensed in high speed.The A/D conversion part of signal acquisition uses the ALIENTEK Mini STM32 A/D conversion channel,then using the serial port RS232 to communicate with the computer.The main application of the software part is the principle of artificial neural network.Based on the wavel et packet decomposition theory to select seven characteristic value as the input of the network,constructing a 7×12×1 back propagation network,training and learning the collected sitting pressure distribution data,realizing the discriminant classificati on of sitting pressure distribution data.Matlab GUI is used to produce the visual interface,using interpolation to achieve the smooth display of pressure distribution and determine the sitting type of the test data.Finally,an example test was conducted on some volunteers,the results have obvious test results,conforming the facts.In addition,seeing from the analysis of all the results,the system's measurement accuracy is relatively high.The experiment results show that the reasonable feasibility of the sitting position sensing system based on neural network and it can accurately measure the pressure distribution of the human body sitting posture and classification of the sitting posture to a certain extent.To lay the foundation for the following research of the topic.