We present a novel approach that exploits shape context to recognize emotion from monocular dance image sequences. The method makes use of contour information as well as region-based shape information. The procedure of the method is as follows. First, we compute binary silhouette images and its bounding box from dance images. Next, we extract the quantitative features that represent the quality of the motion of a dance. Then, we find meaningful low-dimensional structures, removing redundant information but retaining essential information possessing high discrimination power, of the features using SVD (Singular Value Decomposition). Finally, we classify the low-dimensional features into predefined emotional categories using TDMLP (Time Delayed MultiLayer Perceptron). Experimental results demonstrate the validity of the proposed method.