A Method For Calculating The Body Weight Of A Person Based On A Somatosensory Interaction Device

The human body center of gravity (COG) is a basic parameter that reflects the characteristics of human body structure and quality distribution. It plays an important role in physical education and sports in primary and middle schools. Especially, COG is a very important biomechanical parameter that measures the balance ability of human body. So, the measurement of COG has important academic and application value. At present, somatosensory interaction technology has been widely used in many areas,such as teaching, physical exercise and medical rehabilitation. But, there are some issues to obtain COG in the environment of somatosensory interaction, such as big error,lack of individuality and so on.In this paper, we obtained the coordinates of 25 joints of the human body by the somatosensory interaction device (Kinect),combined with the Segmental Method(SM),constructed a model with 19 joints. Then, basic on the joint model,we proposed a real-time linear calculation method (RLCM) based on Kinect for obtaining the human body COG position. In addition, we used the center of force (COF) data as the standard COG data, and the COF data obtained by the tablet pressure test system Footscan. We compared the COG data calculated by RLCM with standard COG data (17 groups), the results showed that the COG data calculated by RLCM had a long-term common random trend with the standard COG data, and the P value is less than 0.03 (p< 0.03);With the decrease of the action range, the trend was increasing, and the relative error was increasing. So, that means RLCM can be used to measure the COG of the human body, especially in the measurement of the larger range of action, the relative error is smaller.Because the RLCM is overly dependent on the human segment model, in order to improve the precision of COG calculation in somatosensory interaction environment and to adapt to the movement characteristics of each individual participating in the test,we used three regression methods to rebuild the joint model, and the three regression methods is Multiple Linear Regression method, Artificial Neural Network Regression method, Support Vector Regression method. The Mean Absolute Error (MAE) of the three methods are 2.49mm, 2.20mm,0.50mm, and all the MAE are lower than the MAE of RLCM's 7.12mm, and the Support Vector Regression method can achieve better effect.This study has made an attempt in real-time calculation of COG, hoping to provide more effective evidence for the assessment and training of the human body's balance ability,and at the same time,to provide some reference to the physical education, sports and other fields.