ステレオ画像列からの動作認識

概要

論文の詳細を見る
This paper presents a gesture recognition method extending Temporal Templates so that they can contain not only vertical and horizontal motion but also depth information obtained from a binocular stereopsis. The proposed method can discriminate gestures with depth motion. At first, a disparity image generated from stereo images is divided into several disparity stages and in each disparity stage, a grayscale feature image (Temporal Template) is created by assigning the intensity according to the frame number to the area where motion has been detected. Next, a gesture model is generated from learning feature images acquired in each disparity stage by SVM. A gesture is recognized by checking feature images generated from input stereo images against SVM model. Experimental results have shown the effectiveness of the proposed method for recognizing gestures with depth motion.
公益社団法人計測自動制御学会の論文