Detecting indicated object from head-mounted omnidirectional images by pointing gesture

We propose a method and its application for detecting an object indicated by pointing gestures. The method uses a linear model and the concepts of a cognitive origin and a reference plane. We use a wearable camera system to recognize pointing gestures. Our system consists of an omnidirectional head-mounted camera and methods called random sampling and importance sampling, and the advantages of our system are that the system runs in real-time and that the camera can shoot the arm and an indicated object simultaneously.