Automatic human activity segmentation and labeling in RGBD videos

Abstract

Human activity recognition has become one of the most active research topics in image processing and pattern recognition. Manual analysis of video is labour intensive, fatiguing, and error prone. Solving the problem of recognizing human activities from video can lead to improvements in several application fields like surveillance systems, human computer interfaces, sports video analysis, digital shopping assistants, video retrieval, gaming and health-care. This paper aims to recognize an action performed in a sequence of continuous actions recorded with a Kinect sensor based on the information about the position of the main skeleton joints. The typical approach is to use manually labeled data to perform supervised training. In this paper we propose a method to perform automatic temporal segmentation in order to separate the sequence in a set of actions. By measuring the amount of movement that occurs in each joint of the skeleton we are able to find temporal segments that represent the singular actions.We also proposed an automatic labeling method of human actions using a clustering algorithm on a subset of the available features.