A Reinforcement Learning-Based Basketball Player Activity Recognition Method Using Multisensors

Abstract

It is an effective means to use a computer auxiliary system to assist athletes in training. In this paper, we design a technical activity recognition system for basketball players. The system uses the sensing module bound to the basketball player to collect the activity data and uses the proposed Multilayer Parallel Long Short Term Memory (MP-LSTM) algorithm to recognize the activity. Moreover, in order to extend the working time of the system and reduce the energy consumption of the sensing module, we also utilize the classical reinforcement learning algorithm DQN to adaptively control the sampling frequency of the sensing module for making a trade-off between recognition accuracy and energy consumption. Experiment results show that the recognition accuracy of the proposed MP-LSTM algorithm reaches 94%, while the recognition accuracy of the system remains at about 90% after applying the DQN algorithm, and the energy consumption is reduced by 76%.

Hang-Time HAR: A Benchmark Dataset for Basketball Activity Recognition Using Wrist-Worn Inertial Sensors

“…The group used data from five full IMUs attached to the participants' shoes, knees, and lower back in order to classify frequently occurring basketball activities such as walking, running, jogging, pivot, jumpshot, layup, sprinting, and jumping. The most relevant studies to this work are [67,[71][72][73][74]. Ref.…”

“…In comparison to our study, the majority of these studies predominantly utilize machine learning techniques or statistical analysis. The only exceptions are Eggert et al [75] and Bo et al [74], both of which employ deep learning-based approaches. However, these studies lack data with equivalent scope, complexity, and diversity.…”