Output details

Real-Time Posture Reconstruction for Microsoft Kinect

<26> In this project, we proposed a new data-driven algorithm to reconstruct captured human postures from noisy source data in real-time. Unlike traditional algorithms that either assume high accuracy or discard inaccurate data, our method evaluates the reliability value of the tracked body parts, and integrates the value into a machine learning framework to enhance the reconstruction process. The journal was formerly known as the IEEE Transactions on Systems, Man, and Cybernetics - Part B Cybernetics. This research was an international collaborative project involving the UK (Northumbria University), Hong Kong (Hong Kong Baptist University) and Japan (University of Tokyo).