In this paper, discrete orthogonal moment-based shape features up to 5th order are proposed for Indian sign language (ISL) recognition system. The shape recognition capability of discrete orthogonal moment-based local features is verified on two databases. These include the standard Jochen-Triesch’s database and 26 ISL alphabets. The ISL alphabets are collected on both uniform and complex backgrounds, with variations in position, scale and rotation. The feature-set is increased for 26 ISL alphabets by varying Region of Interest (ROI) and extracting features from each ROI. A minimum possible feature-set with least redundancy is selected that gives the best recognition accuracy. The effect of order and feature dimensionality for different classifiers is studied. Results show that both Dual-Hahn and Krawtchouk moments are found to exhibit user, scale, rotation and translation invariance. Moreover, they have shape identification capability, thus achieving good recognition accuracy.