Due to availability of reliable and low cost devices, range maps (depth maps) are extensively used in many applications. Recent advances in human-computer interaction enabled us to interact with computers in intuitive and friendly way. In this paper, we propose a novel approach for recognizing static hand gestures using depth information captured from Photon Mixing Device (PMD) cameras. We segment hand from background based on received signal amplitude and pixel depth values. The segmentation is robust and works well even with cluttered backgrounds. Shape of the hand is captured with gradient magnitude features. We use Random Projection (RP) and Kernel Principal Component Analysis (KPCA) for dimensionality reduction and then perform subsequent classification in the lower dimension space. We also propose a strategy to reduce the training time required in the process. To validate performance of our approach, we experimented on American Sign Language (ASL) gestures. Experimental results show that our approach is efficient and quite effective in recognizing static gestures. A five-fold cross validation accuracy for static ASL gestures was 99.8%.