Knowledge gap: Gait evaluation tools are usually difficult when applied to children, for instance use of sensors or the lack of portability of heavy fix installations. On top of that, extracting meaningful signal from those records is also challenging.

Approach: In this paper, they propose to rely on Microsoft Kinect instrument for skeleton tracking. A complete methodology going from raw data to gait parameters extraction is provided. Two main points of their approach is: learning stride templates from training data and use them for automatic temporal segmentation and feature extraction, such as step time, swing time, stride length, step length,... They compare their approach with manual annotation of Kinect records and with another device GAITRite, which consists in a roll-up walkway.

Strengths:

Their approach is fully automatic, given the training of stride templates.

Weaknesses: