Acquisition Of Human Pose Information And Its Application On Balance Evaluation

As an important physiological function of the human body, balance ability playsan irreplaceable role in maintaining the body s normal posture and activity. Evaluationof balance ability is an important basis for diagnosis and treatment and rehabilitationtraining on balance disorders. Methods for the evaluation of balance ability such asobservation, scale of assessment and other traditional methods ask high requirementson people, and the results of evaluation are easily affected by the subjective effects ofpersons. With the use of modern sensor technology and computer technology, it canuse a variety of measurement and analysis methods to deal with the relevant data ofbalance ability. Providing more accurate and objective assessment results has asignificance meaning in raising the level of rehabilitation medicine about balanceability.The paper has a further researchin attitude information fusion, evaluation indexand evaluation methods and designs two experiments including static balance test anddynamic balance test based on a low cost balance-detecting system constructed byforce plate and attitude module. The main achievements and innovations in this paperare summarized as follows:(1)Currently, the balance tests mainly use a single measurement tool such as forceplate. This paper builds a balance-measuring system which combines with force plateand attitude module and developesa software to collect body s motion of mechanicalsignal. This balance-measuring system can greatly increase the balance indexinformation with the combine of a variety of balance test paradigm. You can use theattitude module to detect the posture change that force plate can not capture.(2)The Kalman filter algorithm is commonly used in attitude sensor data fusion, itoften have good result and accuracy. But Kalman filter algorithm is complicated, thesampling frequency of linear regression iterations in Kalman filter is usually very highand its calculation is relatively large. This paper uses gradient descent method to fuseattitude information. The calculation of gradient descent algorithm is less thanKalman filter algorithm, so it s more suitable to be used in a MEMS sensor.(3)In this paper, balance indexs such as length of gravity, shaking diameter, Eulerangle, hemiplegia index, standing ratio, angular velocity are used to analyze the balance data of subjects. This paper proposes to use the principal component analysisto extract the comprehensive information of indexs and reduce the dimension ofindexs because of too many indexs. Finally the paper uses cluster analysis based onprincipal component analysis to classify balance ability of subjects and achieves thepurpose of evaluation.(4)Based on the evaluation indexes and evaluation methods described before,this paper designs two experimental paradigms including static balance test anddynamic balance test to analyze and evaluate the balance ability of subjects. Theexperimental results show that length of gravity and angle in the Euler angles have asignificantly effect on static balance ability. While in the dynamic balance test, bothhemiplegia index and standing radio show a good result, but the calculation ofhemiplegia index and its low processing difficulty make it more suitable to evaluatebalance ability.