Automatic Recognition of Concurrent and Coupled Human Motion Sequences

We developed methods and algorithms for all parts of a motion recognition system, i. e. Feature Extraction, Motion Segmentation and Labeling, Motion Primitive and Context Modeling as well as Decoding. We collected several datasets to compare our proposed methods with the state-of-the-art in human motion recognition. The main contributions of this thesis are a structured functional motion decomposition and a flexible and scalable motion recognition system suitable for a Humanoid Robot.