• DocumentCode
    3659655
  • Title

    Multi-band sum of spectrogram based audio fingerprinting of Indian film songs for multi-lingual song retrieval

  • Author

    S. Sri Ranjani;K. Karthik;P. K. Bora;V. Abdulkareem

  • Author_Institution
    Dept. of Electronics and Electrical Engg., Indian Institute of Technology Guwahati, Assam, 781039, India
  • fYear
    2015
  • Firstpage
    1414
  • Lastpage
    1420
  • Abstract
    Film music compositions are highly diversified, exhibiting not just changes in background scores and singer´s voices, but even the lyrical embellishments are morphed into different languages to suit regional audiences. Given this diversified prevalence amongst recorded film music, retrieval becomes extremely challenging. In this paper we propose an approach based on a multi-band sum of spectrogram, executing a delicate tradeoff between sensitivity to pitch jitters incurred by lyrical and singer voice changes while keeping the melodic signature intact. The top-3 retrieval accuracy for the multi-band sum of spectrogram has been found to be around 91% for an STFT window size of 128ms.
  • Keywords
    "Spectrogram","Feature extraction","Robustness","Databases","Jitter","Timing","Time-frequency analysis"
  • Publisher
    ieee
  • Conference_Titel
    Advances in Computing, Communications and Informatics (ICACCI), 2015 International Conference on
  • Print_ISBN
    978-1-4799-8790-0
  • Type

    conf

  • DOI
    10.1109/ICACCI.2015.7275811
  • Filename
    7275811