An Improved Activity Recognition Method Based on Smart Watch Data

Activity recognition is mainly used in health care, authentication and many other fields. Activity recognition based on a smart watch usually performs much better than any other means of activity recognition in these areas. With its small size, high integration and multi-function, smart watch holds more advantages over the devices used in image based activity recognition, for example, a monitor, which is usually location fixed and view limited. In this paper, we collected data of five behaviors, i.e. stand, walk, run, upstairs, downstairs from the accelerometer on a smart watch. These data are classified with CART decision tree. Experimental results prove that the proposed method improves the accuracy of activity recognition.