Generalized and Efficient Skill Assessment from IMU Data with Applications in Gymnastics and Medical Training

Abstract

Human activity recognition is progressing from automatically determining what a person is doing and when, to additionally analyzing the quality of these activities—typically referred to as skill assessment. In this chapter, we propose a new framework for skill assessment that generalizes across application domains and can be deployed for near-real-time applications. It is based on the notion of repeatability of activities defining skill. The analysis is based on two subsequent classification steps that analyze (1) movements or activities and (2) their qualities, that is, the actual skills of a human performing them. The first classifier is trained in either a supervised or unsupervised manner and provides confidence scores, which are then used for assessing skills. We evaluate the proposed method in two scenarios: gymnastics and surgical skill training of medical students. We demonstrate both the overall effectiveness and efficiency of the generalized assessment method, especially compared to previous work.

Assessing Human Motion During Exercise Using Machine Learning: A Literature Review

“…However, additional classes can provide a finer distinction between the performed motions [50]. In addition, hybrid approaches have been studied in which ranking-based assessment is followed by regression-based algorithms that use the confidence scores of the classification to compute a motion quality score [3]. Ranking-based assessments have the advantage of offering a better distinction in terms of motion quality.…”

“…The cross-validation technique used depends on the characteristics of the dataset. For instance, smaller datasets may have higher accuracy with a leave-k-subjectsout approach [3] because the data of individual subjects are not mixed in both the training and validation subsets. The cross-validation technique can also be used during the training phase of the model to facilitate the selection of the optimal parameters of the algorithm.…”

“…These can include the duration of an exercise or a sub-phase of an exercise (e.g., during the loading and release phases of a serve in tennis [31]). Finally, different signal analysis algorithms were used in ten publications [3], [33], [49], [57], [68], [69], [76], [88], [97], [104]. Such algorithms can engineer features relating motion trajectories to predefined templates or extract frequency domain characteristics of the signals.…”

“…In other words, it analyzes and quantifies how someone performs an activity. HMQA should not be confused with Human Activity Recognition (HAR), which is another motion analysis technique that focuses on recognizing what activity is being performed [2], [3].…”

Applications of Haptic Technology, Virtual Reality, and Artificial Intelligence in Medical Training During the COVID-19 Pandemic

“…For a more advanced classifier fusion technique, one can refer to the proposed method in Kazemian et al (2010) where more advanced methods such as Dempster’s Rule of Combination (DCR) and Choquet integral are compared with more basic techniques. Activity recognition and movement classification is another efficient way to calculate metrics representing the surgical skill automatically Khan et al (2020) . Moreover, instrument detection in a video and drawing centroid based on the orientation and movement of the instruments can reflect the focus and ability to plan moves in a surgeon.…”