BoxerSense: Punch Detection and Classification Using IMUs

Abstract

Physical exercise is essential for living a healthy life since it has substantial physical and mental health benefits. For this purpose, wearable equipment and sensing devices have exploded in popularity in recent years for monitoring physical activity, whether for well-being, sports monitoring, or medical rehabilitation. In this regard, this paper focuses on introducing sensor-based punch detection and classification methods toward boxing supporting system which is popular not only as a competitive sport but also as a fitness standard for people who wish to keep fit and healthy. The proposed method is evaluated on 10 participants where we achieved 98.8% detection accuracy, 98.9% classification accuracy with SVM in-person-dependent (PD) cases, and 91.1% classification accuracy with SVM in person-independent (PI) cases. In addition, we conducted a preliminary experiment for classifying six different types of punches performed from both hands for two different sensor positions (right wrist and upper back). The result suggested that using an IMU on the upper back is more suited for classifying both hand punches than an IMU on the right wrist. To provide feedback in real time, we estimated the real-time performance of each classification method and found out all our methods could classify a single punch in less than 0.1 s. The paper also discussed some points of improvement toward a practical boxing supporting system.