RECOGNITION OF HUMAN HAND GESTURE FROM A MONOCULAR IMAGE SEQUENCE FOR HUMAN-MACHINE COMMUNICATION SYSTEM

摘要

The goal of this study is to construct a human-machine communication system. In the present study, we propose a technique to estimate the motion and position of human hand from a monocular camera, so as to extract motion parametors and recognize human gestures. Sequential images which are recorded with a monocular camera are transformed into log-polar mapped images to reduce the date size of the images without accuracy decreasing near the target. We carried out experiments to demonstrate the effectiveness of the method proposed. The experiments were carried out to actual sequential images. As the experimental results, the system performed the tracking and gesture recognition with high accuracy and with processing time of about 40 milliseconds per a frame, with an available personal computer without extra peripheral equipments.