Machine Learning for Human Motion Understanding

Due to the advancement in human motion capturing technology and the availability of public human motion database, human motion understanding has become a core component of many research problems in multiple research domains. In computer graphics, a good representation of human motion facilitates characters with realistic movement, intuitive ways of crowd controls, and the real-time processing of captured motion data. In computer vision, modelling human motion is a key process for effective action classification and gait analysis. In order to understand human motion represented in different format, machine learning algorithms are adapted to learn a representation of motions, or to extract features of them. In this open lecture, I will discuss the importance of human motion understand in computer science. With the support of my research projects, I will introduce different machine learning algorithms such as lazy learning, dictionary learning and deep learning. I will explain how these algorithms can be used to model human motion in order to solve real-world problem in the computer graphics and computer vision areas as mentioned above. I will show how my projects achieve impact in research and the society, and conclude my presentation with future opportunities and potential research directions. This lecture is suitable for all audiences including those with limited experience in machine learning and visual computing, as well as the professionals in the fields.