Sports Video Classification – A Review

Abstract: Video classification is a boundless topic. The Sports video classification is fundamental one and has considerable importance for archiving digital content in broadcasting companies. It is the process of classifying unknown sports video into the type of sport being played. This review paper analyses the different methods such as Neural Net, Texture code cue, Principle Component Analysis (PCA) and automatic thresholding, Mel Frequency Cepstral Coefficient (MFCC), Support Vector Machine (SVM), Hidden Markov model (HMM), CNN and RNN based Spatial and Temporal analysis in sports video classification based on sequential frame dataset. Convolutional Neural Network (CNN) is the basic one to extract features of images. Machine learning, as well as deep learning, approaches are popular for image recognition and classification. The simple and sophisticated method for video classification is “Transfer learning”. It uses a simple and publicly available dataset. Transfer learning is the process of using a system or model, trained for one particular purpose to the other similar and specific purpose. VGG16 is a well- known pre-trained model for image classification. This model is considered for video classification.

Keywords: Neural Net, PCA, MFCC, SVM, HMM, CNN and RNN, Transfer learning.