Optimized Motion Capture for Cricket Shot Classification Using Minimal Hardware and Machine Learning

Optimized Motion Capture for Cricket Shot Classification Using Minimal Hardware and Machine Learning

Motion capture technology has been used in many sports analytics, particularly for analyzing player movements in games like cricket. However, the high cost of commercial systems and the disturbing nature of wearable devices limit their accessibility and practicality. This study presents a two-fold s...

Full description

Saved in:
Bibliographic Details
Main Authors: J. Ishan Randika, Kanishka Rajamanthri, Avishka Kothalawala, Niroshan Gunawardana, Ashan Induranga, Pathum Weerakkody, Kaveendra Maduwantha, B. T. G. S. Kumara, Kaveenga Koswattage
Format: Article
Language:English
Published: IEEE 2025-01-01
Series:IEEE Access
Subjects:
Cricket batting
deep learning
machine learning
motion capturing
neural network
Online Access:https://ieeexplore.ieee.org/document/11062611/
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Motion capture technology has been used in many sports analytics, particularly for analyzing player movements in games like cricket. However, the high cost of commercial systems and the disturbing nature of wearable devices limit their accessibility and practicality. This study presents a two-fold solution: first, the development of a low-cost, non-intrusive optical motion capture system; and second, the implementation of a machine learning model to classify cricket batting shots based on body angle variations over time. After evaluating multiple optical techniques, MediaPipe was identified as the most effective and affordable solution. A system comprising four Logitech 720p cameras was deployed using Python multiprocessing to capture player motion in real time, ensuring efficient handling of both computational and input/output operations. Motion data collected from the system was analyzed to extract distinct angle variation patterns associated with different batting shots. These patterns were used to train a hybrid machine learning model combining Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) networks. The proposed model achieved an accuracy of 95% in classifying various batting shots, demonstrating the feasibility and effectiveness of the system as an economical alternative for sports motion analysis.
ISSN:2169-3536

Similar Items