Human action recognition is an important computer vision task because of many possible applications such as surveillance, gaming and entertainment, content-based video retrieval, human-computer interaction, assisted environments etc. In this talk, I will present an approach for human action recognition from depth maps and skeleton data available from depth sensors such as Kinect. The approach is based on an amalgamation of key local and global features from depth maps and skeleton data. Some of the features include spatial pose, temporal variation in ‘joints’ position, spatio-temporal gradient orientation, motion induced shape information from depth maps. Feature selection is carried out to select a relevant subset of features for action recognition. The resultant features are evaluated using SVM classifier. The effectiveness of the method is demonstrated on publicly available MSR-Action3D and UT-Kinect action datasets.