Graph-based analysis of physical exercise actions

Abstract

In this paper, we develop a graph-based method to align two dynamic skeleton sequences, and apply it to both action recognition tasks as well as to the objective quantification of the goodness of the action performance. The automated measurement of "action quality" has potential to be used to monitor action imitations, for example, during a physical therapy. We seek matches between a query sequence and model sequences selected with graph mining. The best matches are obtained through minimizing an energy function that jointly measures space and time domain deformations. This measure has been used for recognizing actions, for separating acceptable and unacceptable action performances, or as a continuous quantification of the action performance goodness. Experimental evaluation demonstrates the improved results of our scheme vis-à-vis its nearest competitors. Furthermore, a plausible relationship has been obtained between action perturbation, given by the joint noise variances, and quality measure, given by matching energies averaged over a sequence.

Manipulation-Skill Assessment from Videos with Spatial Attention Network

“…Jin et al [22] proposed to automatically assess surgeon performance by tracking and analyzing tool movements in surgical videos, which hints us that the model for skill assessment should pay attention to the task-related regions in video rather than treat visual information in every region equally. For the purpose of automatic rehabilitation or sports skill training, there is another class of works focusing on the assessment of motion quality or sports performance through computer vision approaches [2,3,23,32,34,36,44,48,49]. Most of these methods are designed for specific tasks, and thus the generalizability is restricted.…”

Averaging of motion capture recordings for movements’ templates generation

“…In cases when hardware returns directly spatial coordinates of the body joints authors prefer to use feature that are derived from those coordinates. Among those features are various configurations of angle-based [6,10,46] and coordinate-based features [2,12]. Basing on our previous researches [26] the angle based-features gives better recognition results than coordinate -based.…”

Application of Assistive Computer Vision Methods to Oyama Karate Techniques Recognition

“…To classify actions represented by HOVV descriptor k-nearest neighbor classifier, Support Vector Machines classifier, and Extreme Learning Machines are used. In paper [4] authors develop a graph-based method to align two dynamic skeleton sequences. Authors quantize the skeleton pose space in order to decrease redundancy in the temporal domain.…”

“…However the joints-based representation is dependent on the relative position of the user to the sensor and body proportion. Papers [4,41,43] propose to use features based on angles between vectors derived from selected body joints. The angle-based features are invariant to relative position to camera, body proportion, and are normalized to range [0,π).…”