Human action recognition, as one of the most important topics in computer vision, has been extensively researched during the last decades due to its potential diverse applications. However, it is still regarded as a challenging task especially in realistic scenarios. The main challenge lies in how to design an effective human action representation that is sufficiently descriptive while computationally efficient. In the past decades, local and holistic representations are extensively studied for human action recognition and both achieve state-of-the-art performance on commonly used benchmarks. In this article, we provide an introduction to human action recognition and a comprehensive review on recent progress in both local and holistic representations of actions. In addition, we also describe the widely used benchmark human action datasets on which action recognition methods are evaluated and compared.