sEMG Onset Detection via Bidirectional Recurrent Neural Networks With Applications to Sports Science

Ergeneci, Mert; Carter, Daryl; Kosmas, Panagiotis

doi:10.1109/jsen.2022.3198882

“…The tasks in our dataset include sEMG signal with sequential patterns and temporal dependencies, and we choose bidirectional long short-term memory (BiLSTM) recurrent neural network (RNN) to learn those temporal patterns [4], [35]. Conventional LSTM RNNs are used in sequence classification due to their ability to extract temporal attributes [36]- [38].…”

Section: Base Deep Learning Modelmentioning

confidence: 99%

“…To mitigate the risks of injury, the trainers should observe the physiological changes of the athletes regularly [3]. Surface electromyography (sEMG), a non-invasive medical technique for measuring the electrical activity on muscle tissue, can be utilised to monitor such physiological shifts of athletes during training [4]. The sEMG signal maintains information required to detect the movements and monitor the musclerelated physiological risk factors [5]- [8].…”

Section: Introductionmentioning

confidence: 99%

sEMG Motion Classification Via Few-Shot Learning With Applications To Sports Science

Ergeneci¹,

Bayram²,

Yildiz³

et al. 2023

Preprint

0

View full text Add to dashboard Cite

Motion classification with surface electromyog- raphy (sEMG) has been studied for practical applications in prosthesis limb control and human-machine interaction. Recent studies have shown that feature learning with deep neural networks (DNN) reaches considerable accuracy in motion classification tasks. However, DNNs require large datasets for acceptable performance and fail for tasks with few data samples available for training. Professional athlete training includes hundreds of exercises, and coupled with privacy and confidentiality issues acquiring a large dataset for all the exercises is not feasible. As a result, state-of-the-art DNN architectures are unsuitable for real-life sports applications. We utilise few-shot learning (FSL) techniques to overcome the small dataset problem of sports-related motion classification tasks. The employed methodology uses the knowledge gathered from a large set of tasks to classify unseen tasks with a few data samples. The FSL approach with a siamese network and triplet loss reached the best performance with a median F1-score of 72.01%, 76%, and 79% for 1, 5 and 10 shot datasets that include an unseen set of tasks, respectively. In contrast, DNN with transfer learning (TF) reached 49.27%, 51.58%, and 67.66% for the same set of tasks, respectively.

show abstract

“…, where s = 2, r is the 0.25 times the standard deviation of x i f , and d is the Chebyshev distance [33]. We determine the parameters r and s according to the work [4].The output SampEn vector of the ith observation S i [n] is finally calculated as…”

Section: B Preprocessingmentioning

confidence: 99%

sEMG Motion Classification Via Few-Shot Learning With Applications To Sports Science

Ergeneci¹,

Bayram²,

Yildiz³

et al. 2023

Preprint

1

0

View full text Add to dashboard Cite

Motion classification with surface electromyog- raphy (sEMG) has been studied for practical applications in prosthesis limb control and human-machine interaction. Recent studies have shown that feature learning with deep neural networks (DNN) reaches considerable accuracy in motion classification tasks. However, DNNs require large datasets for acceptable performance and fail for tasks with few data samples available for training. Professional athlete training includes hundreds of exercises, and coupled with privacy and confidentiality issues acquiring a large dataset for all the exercises is not feasible. As a result, state-of-the-art DNN architectures are unsuitable for real-life sports applications. We utilise few-shot learning (FSL) techniques to overcome the small dataset problem of sports-related motion classification tasks. The employed methodology uses the knowledge gathered from a large set of tasks to classify unseen tasks with a few data samples. The FSL approach with a siamese network and triplet loss reached the best performance with a median F1-score of 72.01%, 76%, and 79% for 1, 5 and 10 shot datasets that include an unseen set of tasks, respectively. In contrast, DNN with transfer learning (TF) reached 49.27%, 51.58%, and 67.66% for the same set of tasks, respectively.

show abstract

“…[13][14][15] And others are investigating how this technology and the data analysis it collects can be useful for improving players' physical and tactical performance, [16][17][18][19][20][21][22] the fan, athlete or student experience, 7,[23][24][25][26] or injury prevention and reduction. [27][28][29][30][31][32] Many of these analyses focus on showing the benefits of technology in competitive sports, in large and complex organisations and in large stadiums (arenas). However, in the extensive and varied taxonomy of the sports industry, not all organisations need or have access to the same technological tools for the development of their activity, nor do they have the digital skills to leverage their functionalities.…”

Section: Introductionmentioning

confidence: 99%

“…13–15 And others are investigating how this technology and the data analysis it collects can be useful for improving players’ physical and tactical performance, 16–22 the fan, athlete or student experience, 7,23–26 or injury prevention and reduction. 27–32…”

Section: Introductionmentioning

confidence: 99%

Technology and digital transformation for the structural reform of the sports industry: Building the roadmap

Magaz-González,

García-Tascón,

Sahelices-Pinto

et al. 2023

Proceedings of the Institution of Mechanical Engineers, Part P:

View full text Add to dashboard Cite

The COVID-19 pandemic in 2020, has accelerated technological immersion into society, economy and public administrations. The sports organisations are no strangers to this digitisation and must carry out their own digital transformation. However, investment in digital technology must be preceded by a diagnosis of the technological needs of each sports entity. Said mapping will help organisations know the most appropriate technological tools for their core businesses so they can properly design their digital transformation strategies. The objectives of this study were to design and create a tool to understand the digital structure of Spanish sports. The result has been the configuration of an instrument that includes the description of different technologies and different digital competencies specific to the sports industry, and which allows individuals to know the use, importance, perceived difficulty of use and economic accessibility of available technologies, as well as the degree of developed competency in the different sports organisations. It is concluded that the creation and application of a consultation instrument on digitalisation is the first and necessary step to carry out relevant, in-depth, valid and replicable research, which allows information to be gathered on the digitalisation needs of sports organisations to design their digital transformation roadmap and that the aids for digital transformation is distributed efficiently.

show abstract

sEMG Onset Detection via Bidirectional Recurrent Neural Networks With Applications to Sports Science

Cited by 11 publications

References 47 publications

sEMG Motion Classification Via Few-Shot Learning With Applications To Sports Science

sEMG Motion Classification Via Few-Shot Learning With Applications To Sports Science

sEMG Motion Classification Via Few-Shot Learning With Applications To Sports Science

Technology and digital transformation for the structural reform of the sports industry: Building the roadmap

Contact Info

Product

Resources

About