Application of video interpolation to markerless movement analysis

Markerless video analysis techniques, such as human posture extraction, could address technology complexity limitations of clinic-based movement analysis. Commercial markerless systems often require multiple, precisely calibrated and synchronised video streams. These systems are high cost, require specialised equipment, dedicated spaces, and technical expertise. Single-camera posture extraction[1] has quantified spatiotemporal parameters of gait[2]. However, the identification of events, to determine informative measures such as step time variation, is not precise enough for movement health monitoring. Video quality (e.g., resolution, frame rate) can affect posture extraction accuracy. Video frame interpolation (VFI) artificially increases frame rate by estimating flow between intermediate frames[3]. Whilst VFI does not provide new information, VFI might mitigate factors affecting trajectory post-processing. This study assessed whether VFI can improve markerless step time estimation using a single camera.