A Comparative Study Of LRC NNLS Knn And Sparse Representation Based Classification Methods Based On Audio Features

Video comprises of two signal modes the acoustic and visual. The visual information is potentially more difficult to extract. In this work only acoustic signal mode are considered and assessed for the task of genre classification. In the video genre classification field there are many genres in the real world such as music sport information education news and so on. In this paper the three genres are only considered to classify such as cartoon sport and music. The experiments with them have also provided a reference of the performance of such systems when dealing with the own video data set. Furthermore it has been experimented with five classification methods SRC NNLS LRC MSRC and kNN in order to improve accuracy and to see relevant aspects and processes of them. It consists of discrete wavelet subband features then computes the mean and variances for all. Furthermore MFCC features are also implemented in feature extraction. Finally a proper evaluation of the solution has been done. The overall accuracy and classification are also shown in the experimental results.