Snickometer Edge Detection by Feature Extraction in TF Plane and Wavelet Domain

Technical faults leading to wrong decision making are quite common in the game of cricket. Snicko-meter signals show an edge even if the ball has not caught the edge of the bat. This paper distinguishes between bat-ball edge and ball-pad snickometer signals by feature extraction in Time-Frequency Plane (TF plane) and Discrete Wavelet Transform (DWT) domain. The feature vector of the training samples is clustered with Gaussian Mixture Model (GMM) after feature extraction in both TF plane and DWT domain. DWT is preferred over CWT for three reasons i) faster calculation, ii) less time complexity and iii) calculation of features for only a few number of scales as compared to the number of scales required for CWT. Feature vector classification using GMM in TF plane leads to clustering accuracy of 90% and in DWT domain it leads to 95%.