Latent Topic Model Based on Gaussian-LDA for Audio Retrieval

  • Pengfei Hu
  • Wenju Liu
  • Wei Jiang
  • Zhanlei Yang
Conference paper
Part of the Communications in Computer and Information Science book series (CCIS, volume 321)


In this paper,we introduce a new topic model named Gaussian-LDA, which is more suitable to model continuous data. Topic Model based on latent Dirichlet allocation (LDA) is widely used for the statistical analysis of document collections and other discrete data. The LDA model assumes that the words of each document arise from a mixture of topics, each of which is a multinomial distribution over the vocabulary. To apply the original LDA to process continuous data, discretization based vector quantization must be done beforehand, which usually results in information loss. In the proposed model, we consider continuous emission probability, Gaussian instead of multinomial distribution. This new topic model demonstrates higher performance than standard LDA in the experiments of audio retrieval.


Topic model LDA Gaussian distribution Audio retrieval 


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.


  1. 1.
    Wold, E., Blum, T., Keislar, D., Wheaton, J.: Content-based classification, search, and retrieval of audio. IEEE Multimedia 3(2) (1996)Google Scholar
  2. 2.
    Aucouturier, J.J., Defreville, B., Pachet, F.: The bag-of-frames approach to audio pattern recognition: a sufficient model for urban soundscapes but not for Polyphonic Music. The Journal of Acoustic Society of America 122(2), 881–891 (2007)CrossRefGoogle Scholar
  3. 3.
    Helen, M., Virtanen, T.: Query by example of audio signals using Euclidean distance between Gaussian mixture models. In: IEEE International Conference on Acoustic Speech and Signal Processing (ICASSP), Hawaii, USA (2007)Google Scholar
  4. 4.
    Foote, J.: Content-based retrieval of music and audio. Multimedia Storage Archiving Systems II 3229, 138–147 (1997)Google Scholar
  5. 5.
    Sundaram, S., Narayanan, S.: Audio Retrieval by Latent Perceptual Indexing. In: IEEE International Conference on Acoustic Speech and Signal Processing (CASSP), Las Vegas, USA (2008)Google Scholar
  6. 6.
    Kim, S., Sundaram, S., Narayanan, S.: Acoustic topic models for audio information retrieval. In: IEEE Workshop on Applications of Signal Processing to Audio and Acoustics (2009)Google Scholar
  7. 7.
    Kim, S., Sundaram, S., Georgiou, P., Narayanan, S.: Audio scene understanding using topic models. In: Neural Information Processing System (NIPS) Workshop (2009)Google Scholar
  8. 8.
    Blei, D.M., Ng, A.Y., Jordan, M.I.: Latent dirichlet allocation. Journal of Machine Learning Research 3 (2003)Google Scholar
  9. 9.
    Winn, J.M.: Variational Message Passing and its Applications. PHD thesis, University of Cambridge (2003)Google Scholar
  10. 10.
    Attias, H.: A variational Bayesian framework for graphical models. In: Advances in Neural Information Processing Systems (2000)Google Scholar
  11. 11.

Copyright information

© Springer-Verlag Berlin Heidelberg 2012

Authors and Affiliations

  • Pengfei Hu
    • 1
  • Wenju Liu
    • 1
  • Wei Jiang
    • 1
  • Zhanlei Yang
    • 1
  1. 1.National Laboratory of Pattern Recognition (NLPR), Institute of AutomationChinese Academy of SciencesBeijingChina

Personalised recommendations