A Wearable Device for Enhancing Basketball Shooting Correctness with MPU6050 Sensors and Support Vector Machine Classification

Abstract

One of the impacts of Covid-19 is the delay of basketball sports competitions, which influences the athlete’s fitness and the athlete’s ability to play, especially for shooting techniques. Existing research in wearable devices for basketball shooting correctness classification exists. However, there is still an opportunity to increase the classification performance. This research proposes designing and building a smartwatch prototype to classify the basketball shooting technique as correct or incorrect with enhanced sensors and classification methods. The system is based on an Internet of things architecture and uses an MPU6050 sensor to take gyroscope data in the form of X, Y, and Z movements and accelerometer data to accelerate hand movements. Then the data is sent to the Internet using NodeMCU microcontrollers. Feature extraction generates 18 new features from 3 axes on each sensor data before classification. Then, the correct or incorrect classification of the shooting technique uses the Support-Vector-Machine (SVM) method. The research compares two SVM kernels, linear and 3rd-degree polynomial kernels. The results of using the max, average, and variance features in the SVM classification with the polynomial kernel produce the highest accuracy of 94.4% compared to the linear kernel. The contribution of this paper is an IoT-based basketball shooting correctness classification system with superior accuracy compared to existing research.