Posture Recognition Based On Array Strain Sensor

Improving the degree of economic and social development as well as the structural transformation of normal integral development,in recent years,science and technology,the first propeller to promote the progress of society,have made outstanding contributions to the package development of China.In the innovative period of science and technology,the number of science and technology practitioners continues to increase.A growing number of programmers have joined the sedentary groups that used to be dominated by clerical workers and motormen,causing the incidence of "diseases of the senior" such as cervical spondylosis and lumbar spondylosis to increase among the young.Therefore,sitting health is getting significant attention from the public.This paper's background,which is based on the situation that people pay attention to the health of sitting posture,makes my study carry out the recognition and classification of human sitting posture.First of all,looking up the existing domestic and foreign literature,I understand the sitting status and development trends of domestic and foreign researches in posture recognition.After that,the selection of a single-chip microcomputer and the placement of sensors are explained.In this paper,a measurement device with the AVR microprocessor and the array strain pressure sensor,used as the detection unit,is built.The experiment includes six sitting posture measurements,including sitting posture,forward tilt,left tilt,right tilt,left leg lift,and right leg lift.Then,the algorithm of CEEMD is used to process the obtained test data,and the interference and abnormal data in the data are removed to realize the preprocessing of the data.Then,the data's time-domain features are extracted by normalization and the frequency domain features are obtained through Fast Fourier Transform.Before the trained classifiers are applied to posture recognition classification,the time domain features and frequency domain features of these postures are used to train the support vector machine classifier and neural network classifier,respectively.After that,the total accuracy of recognition classification and individual accuracy of each movement are compared.Finally,the causes of misclassification of the results drawn from the classifier with high accuracy are analyzed,and then the possible causes of misclassification are given.