Human Motion Analysis is receiving a growing attention in the field of assistive technologies. Portable systems, able to be carried home or mounted on socially assistive robots, can help in monitoring and evaluating the autonomy level of elderly people in the upcoming silver society. This paper presents a new paradigm to describe and evaluate human motion that can be used in these scenarios. The proposal is based on parametric segmentation and evaluation of action primitives. These actions can be combined in different sequences or even evaluated in parallel, providing a modular solution that can easily adapt to the analysis of new behaviours or motion tests. The particular use case of the Get Up & Go test has been used to study the validity of the proposal. Autonomous evaluation of the gaits of different performers have been achieved using data captured by a Kinect 2.0 device mounted on a social robot. Experiments have also involved gait data captured with a precise Vicon Nexus system based on markers, to compare with previous results and characterize capture errors of the Kinect device. Results show that the proposed system is adequate to be used in these scenarios.
