Advances in sensing and communications have made it easy to obtain large sets of measurements about an entity, activity or event. A new approach in machine learning is the recent emergence of the multi-view approach. This emerges from real applications where examples are described by different feature sets or different “views” and hence is concerned with the problem of machine learning of the underlying phenomena from multiple distinct data sets/sub-models. Nonlinear state estimation is a long-standing fundamental problem in engineering and computer science applications. For the identification of the dynamics of nonlinear systems with increased complexities, multi-sensor data fusion is widely researched, which are defined as the process of combining information from multiple sources to produce the most precise and complete picture, mostly based on multiple Kalman filters. It is typified by generating its internal states by a state estimator recursively for conditional monitoring and control, etc.
The project will be focus on the context of multi-view learning by generating combining multiple models (views) for human motion estimation in a dynamical environment. Note that while closely related to the problem of multi-sensor fusion, the current multi-view approaches are different since it focuses on the model fusion, yet not well developed in the domain of system with strong nonlinear dynamics. Alternatively, the multi-sensor data fusion approaches can be broadened by taking into the multiple modality of sensing data sources and advances in machine learning.
This project will focus on designing, developing algorithms then validate and applying to applications in robotics, video processing and human movement studies. In many cases movement data is measured from cameras, and inertial sensors (e.g. accelerometers), and the challenge is to reconstruct the underlying movements that was most likely to have caused this data. Students will have the opportunity to explore this concept in biomechanics, tele-robotics, movement tracking etc.
Eligibility requirements: UK honours equivalent in Computer Science, Maths, Engineering.
Enquiries for further details contact: Professor Xia Hong (email@example.com )