Classification Of Human Balance Ability Based On Video Feature Description In VR Environment

The balance ability of human body plays an extremely important role in our daily life,which seriously affects ours normal study and life.Therefore,it is particularly necessary to screen out the people who has balance defects in advance,and let them enter the rehabilitation training and treatment as soon as possible.At present,the mainly method for the classification of balance ability is the physical test.In this paper,virtual reality technology and image processing algorithms are combine to study.The specific research contents are as follows:(1)The Multi-barycentric feature extraction method based on video description was studied.The first preliminary work is mainly to find the subjects,and collect training videos of walking in the virtual reality environment.Next,the image processing related algorithms is used to preprocess those videos.Then the edge detection image is divided from top to bottom by 30%,30%and 40%to get the multi-center of gravity of the human body by use the moment of the image.(2)The classification method of human balance ability based on multi-barycentric area model was studied.Firstly,the multi-barycentric area model(MBAM)based on the balance principle of tumbler was proposed by the characteristics of human walking.Secondly,those subjects were trained in the virtual reality environment.Finally,combining the four characteristics(MBAM,MBAM variance,walking roadmap and moving speed)for numerical analysis,it can be found that there are significant differences in the characteristics of human people with different balance ability,which verifies the effectiveness of the method in this paper.(3)The classification method of human balance ability based on multi-feature fusion of human posture was studied.People with poor balance ability will have an abnormal posture performance when walking in the virtual reality environment.Firstly,the volatility of center of mass,the SURF feature and the Hu invariant moment feature representing the posture shape feature of human body in different subjects during walking were extracted.Then the fusion feature is input to the support vector machine for classification.The experimental results show that the fusion feature can be effectively applied to the classification of human balance.