3D hand gesture recognition based on Polar Rotation Feature and Linear Discriminant Analysis

Abstract

A new method based on Polar Rotation Feature and Linear Discriminant Analysis for hand gesture recognition is proposed in this paper. The gesture images in our system are derived from a 3D laser scanner which generates depth data. Hand area segmentation, hole-filling and normalization are done first, then a feature of the polar rotation distance is extracted via polar-coordinate transformation. Utilized PCA+LDA as the classifier. Experiences show our algorithm is robust and accurate. Finally we achieve 96.67% recognition rates under a set of six kinds of hand gestures.