Kinect-based action recognition in a meeting room environment

Abstract

In recent years, rapid development of biometric technologies opened the door to a new class of fast and reliable identity management solutions. Gait is one of the few biometrics that can be recognized unobtrusively from a distance. This paper presents a new gait and action recognition approach based on the Kinect sensor. The proposed method utilizes the joint angles formed by different body parts during walking in order to construct a scale and view-invariant feature set for gait recognition. Next, we develop a new Kinect-based “walking” and “sitting” action recognition method in a meeting room environment. The method is the first step towards the biometric-based interactive meeting room system, which is capable not only to perform user authentication, but also to conduct action recognition during collaborative activities. The proposed method achieves promising results in empirical evaluation compared to some other Kinect-based approaches.