In this work, we present a human activity recognition (HAR) system for smart homes utilizing depth silhouettes and  transformation. Previous utilization of  transformation on binary silhouettes of human activities showed its usefulness in HAR. However the binary silhouettes only provide the shape information of activities. In this work, we utilize  transformation on depth silhouettes such that the depth information of human body parts can be used in HAR in addition to the shape information. In  transformation, a 2D directional shape feature map is computed first through Radon transform of each depth silhouette, and then a 1D feature profile, that is translation and scaling invariant, gets computed through  transform. Then to a set of the  transformed profiles of various activities, we apply Principle Component Analysis (PCA) and Linear Discriminant Analysis (LDA) to extract prominent activity features which are more compact and robust. Finally, Hidden Markov Models (HMMs) are used to train and recognize daily home activities. Our results show 96.55% in the mean recognition rate over ten typical home activities whereas the same system utilizing binary silhouettes achieves only 85.75%. Our system should be useful as a smart HAR system for smart homes.