키넥트를 이용한 인간 동작 시뮬레이션 인식 오류 개선

Reducing Recognition Error in Human Movement Simulation Using KINECT

Kinect is widely used in many different fields such as manufacturing, health care, ergonomics and contents due to easiness of human body modeling in virtual reality with a low-priced depth sensor. In these fields, a higher standard of accuracy in motion recreation is required for more objective measurement. However, there are some errors such as abnormal output, inaccurate motion, and data losses owing to the performance limit of the equipped sensor. These errors have negative effects on data analysis and application as reducing similarity between real movements and simulation. This paper proposes a correction algorithm that calculates a new predicted coordinate value using previous coordinate values collected by Kinect when an error occurs in order to improve recognition accuracy in human motion simulation.