The 5th Umpire: Automating Cricket's Edge Detection System

Abstract

The game of cricket and the use of technology in the sport have grown rapidly over the past decade. However, technology-based systems introduced to adjudicate decisions such as run outs, stumpings, boundary infringements and close catches are still prone to human error, and thus their acceptance has not been fully embraced by cricketing administrators. In particular, technology is not employed for bat-pad decisions. Although the snickometer may assist in adjudicating such decisions it depends heavily on human interpretation. The aim of this study is to investigate the use of Wavelets in developing an edgedetection adjudication system for the game of cricket. Artificial Intelligence (AI) tools, namely Neural Networks, will be employed to automate this edge detection process. Live audio samples of ball-on-bat and ball-on-pad events from a cricket match will be recorded. DSP analysis, feature extraction and neural network classification will then be employed on these samples. Results will show the ability of the neural network to differentiate between these key events. This is crucial to developing a fully automated edge detection system.

Keywords

• Neural networks
• Edge-Detection
• Feature Classification
• Cricket
• Wavelets