Radar-Based Exercise Energy Expenditure Estimation with Deep Learning

• Published in: IEEE Symposium on Wireless Technology and Applications, ISWTA
• Main Author: Yusni A.N.; Mohd Shariff K.K.; Md Ali M.A.; Yassin I.M.; Sariman H.; Yazli A.Y.
• Format: Conference paper
• Language: English
• Published: IEEE Computer Society 2024
• Online Access: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85203816872&doi=10.1109%2fISWTA62130.2024.10651717&partnerID=40&md5=764c3fb6c2992dd81b33b27dffff9228

Exercise is crucial for maintaining a healthy weight and reducing the risk of chronic diseases, yet over half of Malaysia's adult population is overweight or obese due to lack of activity. Accurate monitoring of energy expenditure during exercise is therefore important. A deep learning approach using micro-Doppler radar data is presented to estimate human energy expenditure. Traditional measurement techniques are complex and expensive, while existing wearable sensors have limitations. The method captures micro-Doppler radar signatures from 11 participants performing treadmill walking and running exercises using a 24 GHz continuous-wave radar. The radar signals are preprocessed into time-frequency spectrograms and inputted to a convolutional neural network (CNN) model for training to predict energy expenditure values. The CNN's performance yielded a root mean squared error of 12 kcal/min, providing valuable insights into energy expenditure estimation. © 2024 IEEE.