Abstract
In the present study kinematic data elicited via a body motion analysis system were used in order to accurately identify individuals throughout specific periods of time. Fifteen males participated in a series of running trials interspersed with an eight-week training period. Body motion analysis comprised data from video recordings during running. After video analysis, various kinematic parameters related to motion of specific body parts (trunk, hip, knee, calf) were compared in order to measure body motion analysis’ recognition efficiency. These kinematic parameters were used as inputs for a classical artificial neural network, in order to recognize each individual, whilst, the output represented the identity of the individual. The artificial neural network is optimized regarding the values of crucial parameters such as the number of neurons, the time parameter and the initial value of the learning rate, etc. using the evaluation set. Three identification indices were selected. The general identification index (Ig) which expressed the % of the correct positive and correct negative identifications to the total population. The false negative index (If-neg) which expressed the % of the incorrect identifications of a non-authentic individual and the false positive index (If-pos) which expressed the % of the incorrect identifications of an authentic individual. The statistics showed that even with the use of 16 additional kinematic parameters the efficiency of the identification process was not improved. Further analysis showed that separately some kinematic parameters provided either higher If-neg or If-pos values whilst others presented low values in both identification indices. It seems that the need for satisfying the biometric criterion of social acceptability resulted in the use of parameters derived from specific body parts which diminished the video analysis efficiency and consequently person identification ability of body motion analysis.