Fine-Grained Independent Approach for Workout Classification Using Integrated Metric Transfer Learning

Fine-Grained Independent Approach for Workout Classification Using Integrated Metric Transfer Learning

S. Rubin Bose, M. Abu Shahil Sirajudheen, G. Kirupanandan, S. Arunagiri, R. Regin, S. Suman Rajest
Copyright: © 2024 |Pages: 15
DOI: 10.4018/979-8-3693-0502-7.ch017
OnDemand:
(Individual Chapters)
Available
$37.50
Current Special Offers
No Current Special Offers
TOTAL SAVINGS: $37.50

Abstract

Physical activity helps manage weight and stay healthy. It becomes more critical during a pandemic since outside activities are restricted. Using tiny wearable sensors and cutting-edge machine intelligence to track physical activity can help fight obesity. This study introduces machine learning and wearable sensor methods to track physical activity. Daily physical activities are typically unstructured and unplanned, and sitting or standing may be more common than others (walking stairs upstairs down). No activity categorization system has examined how class imbalance affects machine learning classifier performance. Fitness can boost cardiovascular capacity, focus, obesity prevention, and life expectancy. Dumbbells, yoga mats, and horizontal bars are used for home fitness. Home gym-goers utilise social media to learn fitness, but its effectiveness is limited.
Chapter Preview
Top

2. Literature Review

Ding, & Ren, (2019) proposed algorithm aims to improve the accuracy of medical exercise rehabilitation image segmentation by using the HFCNN to learn and extract the features of the image and IoT technology to integrate real-time data from sensors and wearable devices, such as heart rate monitors and motion trackers, to provide additional information for the segmentation process. The authors demonstrate the effectiveness of the proposed algorithm through experiments on a dataset of medical exercise rehabilitation images. The results show that the HFCNN-based segmentation algorithm outperforms traditional methods and that the addition of IoT data further improves the accuracy of the segmentation.

A convolutional neural network (CNN) architecture is suggested by Gupta (2021) as a means to learn and extract features from the raw sensor data acquired by the wearable device. In order to categorise the activity, a SoftMax classifier is employed with the features that have been extracted. Research using a freely accessible dataset of wearable sporting activities proves the efficacy of the suggested approach. The results demonstrate that the suggested strategy accomplishes high rates of accuracy for many sorts of activities, and it outperforms conventional approaches to activity classification.

Complete Chapter List

Search this Book:
Reset