To search, Click below search items.


All Published Papers Search Service


Mimicry Engine Using Kinect


Garima Goyal, Onkar Borgaonkar, Avinash More, Abhijeet Nirmal


Vol. 16  No. 2  pp. 112-116


This paper presents a system for matching of 3 Dimensional actions using DTW. This system uses a depth sensing device to generate a skeleton of the user. For the training part, any Expert can train the system for a specific action sequence. The User can perform a similar action and the system will match the actions to calculate the percent match. The intention is to make the system independent of the user size and position by matching the joint angles instead of joint coordinates. At the moment the system is time variant. To further increase the tolerance of error in the user action, we use a range of actions performed by the expert and calculate the similarity of the intended action and user action based on a user specified strictness level. This system will calculate the accuracy of the user’s actions and help improve the action and stance if used in tutoring software. This will be a generalized system which can be used as any action matching and action improving engine. The recognition system is able match the candidate pattern with that of the expert.


DTW (Dynamic Time Warping), depth sensing, joint angles, time variant, similarity, range of actions.