High-intensity interval training exercise recognition using smartwatch

The use of a smartwatches to enable human activity recognition has brought forth immersive applications. This paper presents an end-to-end approach using deep learning to recognise physical exercises from a commercially available smartwatch. The exercises are recognised based on two different settings namely; constrained and unconstrained workouts in the form of High-Intensity Interval Training. The model reported a 97.35% accuracy for constrained exercise recognition, and a 82.29% accuracy for unconstrained exercise recognition. This method is capable of recognising 18 High-Intensity Interval Training exercises. The model was deployed to Google Cloud Platform to recognise exercises in real-time settings. The method will be further expanded to operate as a real-time 'Fitness Coach', which could automatically suggest optimal workout plans for users and monitor their health conditions during workout sessions.