A framework for whole-body gesture recognition from video feeds

Abstract

The growth of technology continues to make both hardware and software affordable and accessible creating space for the emergence of new applications. Rapid growth in computer vision and image processing applications have been evident in recent years. One area of interest in vision and image processing is automated identification of objects in real-time or recorded video streams and analysis of these identified objects. An important topic of research in this context is identification of humans and interpreting their actions. Human motion identification and video processing have been used in critical crime investigations and highly technical applications usually involving skilled human experts. Although the technology has many uses that can be applied in every day activities, it has not been put into such use due to requirements in sophisticated technology, human skill and high implementation costs. This paper presents a system, which is a major part of a project called movelt (movements interpreted), that receives video as input to process and recognize gestures of the objects of interest (the human whole body). Basic functionality of this system is to receive video stream as input and produce outputs gesture analysis of each object through a staged process of object detection, tracking, modelling and recognition of gestures as intermediate steps.