Patent attributes
Embodiments herein generally relate to methods and systems for human motion capture. In at least one embodiment, there is disclosed a method comprising: accessing one or more input 2D image frames of a subject performing a physical activity, wherein for each input 2D image frame, automatically: detecting one or more image portions corresponding to the subject's joints; identifying one or more pre-defined anchors in the image frame; processing the input 2D image frame to generate a modified input 2D image frame by: overlaying a bounding box and one or more reference axis lines over the input 2D image frame to generate the modified 2D image frame; resolving values for the one or more joint-specific angles generated by the intersection of the one or more reference axis lines; applying a joint coupling model to the determine relative joint positions, of the subject's joints, in a transverse and sagittal planes.