Aerobic Fitness Level Estimation Using Wearables

• Published in: 2022 Computing in Cardiology (CinC) Vol. 498; pp. 1 - 4
• Main Authors: Smisek, Radovan, Nemcova, Andrea, Smital, Lukas, Chlibkova, Daniela, Kralik, Martin, Kolarova, Jana, Myska, Vojtech, Kolarik, Martin, Harvanek, Michal, Arm, Jakub, Bastan, Ondrej, Pospisil, Martin, Sima, Jan, Hubalek, Jaromir
• Format: Conference Proceeding
• Language: English
• Published: Creative Commons 04.09.2022
• Subjects: Deep learning; Electrocardiography; Estimation; Heart rate; Neural networks; Sociology; Wearable computers
• ISSN: 2325-887X
• DOI: 10.22489/CinC.2022.302

Background: Aerobic fitness level (AFL) is a parameter closely related to a person's overall health. The gold standard of measurement is currently using expensive laboratory equipment. Aims: This study aimed to estimate AFL automatically using data measured with wearables. Methods: AFL was estimated in 2D space. The first dimension is the exertion level, and the second is the body's response to the exertion. Exertion level was determined based on metabolic equivalent calculated for each classified activity using the data of speed and elevation. The activity classification is based on deep neural networks. The body's response estimation is based on heart rate calculated from ECG or PPG. The test set contained 27 subjects. The reference was measured under laboratory conditions using the gold standard method. AFL classification by ACSM guidelines was used. Results: AFL determined by our algorithm were 0.44\pm 0.09,0.50\pm 0.10,0.53\pm 0.09 , 0.58\pm 0.15 , and 0.70\pm 0.07 for the reference classes very poor, poor, fair, good, and excellent, respectively. The correlation between the reference and determined values is 0.76. Conclusion: Our method showed promising results and will be further developed.