LCO–EGC: levy chaotic optimization-based enhanced graph convolutional network for monitoring health of sports athletes

• Published in: Wireless networks Vol. 30; no. 3; pp. 1401 - 1422
• Main Authors: Paul, N. R. Rejin, Arunkumar, G., Chaturvedi, Abhay, Singh, Upendra
• Format: Journal Article
• Language: English
• Published: New York Springer US 01.04.2024; Springer Nature B.V
• Subjects: Artificial neural networks; Communications Engineering; Computer Communication Networks; Datasets; Electrical Engineering; Engineering; Gait; Health care; Information management; Internet of Things; IT in Business; Kinematics; Networks; Optimization; Original Paper; Performance evaluation; Performance measurement; Search algorithms; Telemedicine; Health monitoring; Levy flight-based chaotic gravitational search algorithm; Internet of things; Weighted average ensemble; Sensor network; Attention enhanced spatial–temporal graph convolutional long short-term memory network
• ISSN: 1022-0038; 1572-8196
• DOI: 10.1007/s11276-023-03574-4

In this modern world, healthcare monitoring is essential to save human lives. The Internet of Things (IoT) plays a vital role in the monitoring of healthcare and also in improving healthcare diagnostics. The IoT is utilized to manage patients’ information as well as to detect diseases early. Thus, we proposed Levy Chaotic Optimization-based Enhanced Graph Convolutional (LCO–ECG) Network for health monitoring in sports athletic. The Hyperparameter of enhanced graphical convolutional neural network is optimized using Levy Chaotic gravitational search algorithm (LCGSA). Also, using LCGSA, the weights are tuned to enhance the efficiency of the resulting ensemble model. The ECG model is applied to get more significant feature information. An action detection system’s accuracy will increase as a result. Here, to validate the proposed method we utilized two datasets including Kinematic gait data using a Microsoft Kinect v2 sensor during gait sequences over a treadmill and Comprehensive Kinetic and EMG datasets. Also, accuracy, precision, recall, F1-score, and AUC are the performance metrics utilized to evaluate classification performance more effectively. The proposed method attained the measures of 95.32%, 94.25%, 95.41%, and 95.97%.