Volume 9 Issue 2 - November 2017

  • 1. A hybrid approach to identify a singer in a video song

    Authors : S.metilda Florence, Dr. S. Mohan

    Pages : 128-131

    DOI : http://dx.doi.org/10.21172/1.92.22

    Keywords : Video annotation, Classification, Combined Classifiers, Artist Identification, Hybrid system

    Abstract :

    Extraction of data concerning the video contents automatically refers to Automatic Video Annotation System. The extracted information will function as the initial step for numerous information access strategies like searching, comparison, and classification. The vast number of video songs accessible to the public requires tools to efficiently retrieve and manage the music of interest to the users. Thus, the projected system would enable them to look for his or her favourite singer’s video song. Underlying Singer within the video songs is distinguished by mining their Linear Prediction Coefficients (LPC) and Spectral features. A unique methodology is applied to boost the performance of the classifiers. A new hybrid system by combining the outputs of two classifiers’ posterior possibilities using product and Mean rule are proposed. The combined classification result shows significant improvement in the outcome. Naïve Bayes and K–Nearest Neighbor (K-NN) algorithms are used for Statistical Analysis. The projected system gives 93% accuracy in identifying a Singer in a video song. Experimental outcomes show that users will retrieve the songs of their selection.

    Citing this Journal Article :

    S.metilda Florence, Dr. S. Mohan, "A hybrid approach to identify a singer in a video song", Volume 9 Issue 2 - November 2017, 128-131