Visual Gesture Recognition of Fingers Based on Hypothesis Tracking

摘要

The analysis of the physiological mechanism of the hand fingers and sampling the hand model with constraints helps to reduce the number of parameters required to describe individual postures. 2D image sequences are then considered informative enough to describe the positions and shape of fingers. Following the Bayes Rule, a hypotheses projection from the ROI extracted from the image sequences to the feature points was proposed to locate the shape and orientation of each finger to show which and how the fingers should move.