Compensatory Movement Detection in Upper Limb Rehabilitation With Deep Learning Methods

Khoramdel, Javad; Moori, A.; Moghaddam, Majid M.; Najafi, Esmaeil

doi:10.1109/icrom54204.2021.9663458

Cited by 6 publications

(10 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The first attempts to transform the positions of the joints so that their relative coordinates use a common reference point, which was performed by 12/49 publications of those that included such data [30], [33], [44], [47], [49], [81], [82], [86], [88], [107], [108], [109]. To perform this normalization, a body segments' property (e.g., pelvis height) is selected as reference, and its transformation properties (e.g., position, scale, magnitude) are transformed to match a specific value.…”

Section: ) Preprocessingmentioning

confidence: 99%

“…Then, the transformation matrix used for the reference body segment is applied to the rest of the joints. The most common reference body segments used in the publications were the head (i.e., based on the subjects' height) [44], [107], [108], [109], the center point between the two shoulders [49], [81], [82], [86], and the pelvis [33], [47], [88].…”

Section: ) Preprocessingmentioning

confidence: 99%

“…The other spatial normalization relates to the reference orientation of the values, that is, the direction of the human skeleton. This type of realignment was applied by 17/49 publications that included applicable data representations (i.e., a human skeleton) [33], [45], [46], [47], [48], [49], [60], [61], [81], [82], [85], [86], [88], [107], [108], [109], [110]. Realignment is usually performed by transforming the human skeleton such that the center point between either the shoulders or hips is facing forward.…”

Section: ) Preprocessingmentioning

confidence: 99%

“…Finally, four publications used generative DL algorithms. The algorithms used were Generative Adversarial Network (GAN) [70], Self-Organizing Map (SOM) [65], Transformer [82], and a combination of a Feed-forward neural network, SOM, and Nonlinear AutoRegressive Neural Network (NAR) [50]. Moreover, two publications used RL algorithms for their assessment, using a combination of Deep Q-network with Double Q-Learning (DQ), Markov Decision Process (MDP), and Artificial Neural Network [93], [94].…”

Section: ) Model Trainingmentioning

confidence: 99%

“…Other often-used types of assessments were using 3-point ordinal scale values evaluated from rank-based classification approaches [35], [48], [54], [62], [63], [64], [68], [77], [83], [86], [87], [97], [98], [99], [103], [107], [108], [109], and score-based assessments evaluated using regression methods [31], [40], [44], [45], [46], [47], [56], [58], [67], [70], [74], [75], [78], [80], [85], [96], [104]. Of the parametric approaches, the most commonly used was the classification to a set of nominallylabeled variables, where each of the labels described a different error that can occur during a motion execution [34], [52], [73], [81], [82], [91], [100], [112], [113], [115], [1...…”

Section: ) Model Trainingmentioning

confidence: 99%

See 4 more Smart Citations

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

et al. 2022

View full text Add to dashboard Cite

The World Health Organization promotes healthy living through regular physical activities, such as exercise and sports, as well as access to healthcare and rehabilitation services for people with motor dysfunctions. However, there is a lack of specialized personnel and increased costs associated with such activities. These have led to the increased use of machine learning for the analysis and evaluation of human motion during exercise. To study the latest advancements in this area, a systematic literature review focusing on publications from 2017 to 2021 was performed. As a result, 88 relevant publications were identified, which developed both shallow machine learning and deep learning algorithms. The results indicated that algorithms for human motion assessment should provide personalized and informative assessments, with explainable and interpretable outcomes, that can be computed in real-time or concurrently with the execution of an exercise. Furthermore, they should be easy to adapt based on the needs of applications and should be able to perform with different motion capture systems. This has been challenging because of the usually small amount of collected data, the lack of large open datasets, and the unique characteristics of exercise motions. Based on the above findings, guidelines for the development of such algorithms are proposed and discussed. They relate to the selection of the type of assessment, handling data imbalances, selecting of motion capture technologies, balancing between accuracy and speed, selecting the right algorithm, performing concurrent assessment during an exercise, personalization and scalability, and evaluation.

show abstract