Quadri_Syed_G.pdf (2.99 MB)
Download fileIndexing of American Football Video Using MPEG-7 Descriptors And MFCC Features
thesis
posted on 2021-05-22, 08:49 authored by Syed G QuadriIn this work, an application system is proposed to classify American Football Video shots. The application uses MPEG-7 motion and audio descriptors along with MEL Frequency Cepstrum coefficient features to classify the video shots into 4 categories, namely: Pass plays, Run plays, Field Goal/Extra Point plays and Kickoff/Punt plays.
Fisher's Linear Discriminant Analysis is used to classify the 4 events, using a leave-one-out classification technique in order to minimize the sample set bias. For a database of 200 video shots taken from four different games, an overall system performance of 92.5% was recorded. In comparison to other American Football indexing systems, the proposed system performs 8% to 12% better.
We have also proposed an algorithm that uses MPEG-7 motion activity descriptors and mean of the motion vector magnitudes, in a collaborative manner to detect the starting point of play events within video shots. The algorithm can detect starting points of the play with 83% accuracy.