Human computer interface using the recognized finger parts of hand depth silhouette via random forests

Hand gesture recognition provides an attractive option for Human Computer Interaction (HCI). In particular, vision-based recognition of finger and hand gestures can help humans to communicate with a computer more efficiently. In this paper, we present a novel approach of recognizing finger and hand parts from a hand depth silhouette using Random Forests (RFs), a multi-class classifier, and its use for a hand gesture HCI. We present how to train the RFs using our own database. Then, the trained RFs are used to recognize finger and hand parts, which are used to recognize hand gestures. We also present an HCI application of finger mouse in which the computer cursor is controlled with a recognized finger.