An automated snick detection and classification scheme as a cricket decision review system

Abstract

Umpire decisions can greatly affect the outcome of a cricket game. When there is doubt about the umpire’scall for a decision, a decision review system (DRS) may be brought into play by a batsman or bowler to validate thedecision. Recently, the latest technologies, including Hotspot, Hawk-eye, and Snickometer, have been employed whenthere is doubt among the on-field umpire, batsman, or bowlers. This research is a step forward in gaging the trueclass of a snick generated from the contact of the cricket ball with either (i) the bat, (ii) gloves, (iii) pad, or (iv) acombination of bat and pad. Preprocessing included noise removal from the snick audios using the Audacity program.Machine learning-based classification is achieved by training neural networks with audio features of the snick. A supportvector machine was employed to enhance the classification system. Twenty-one features comprising time and frequencydomain characteristics were compiled. After one-way analysis of variance was employed and multicomparison analysiswas performed, a set of seventeen features was selected and utilized to increase the accuracy of the proposed DRS.The system was trained on indigenous snick data gathered by ourselves and then tested for real cricket snick scenarios.The system achieves a classification rate of 98.3 percent for the self-collected data while presenting an accuracy of 85.7percentage for the real cricket snick scenarios. The research also developed a dataset of snick audios comprising 132signals for the four categories to aid researchers working in the same field.