Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Videos

AANN: an alternative to GMM for pattern recognition.

B Yegnanarayana1, S P Kishore

  • 1Department of Computer Science and Engineering, Indian Institute of Technology Madras, Chennai, Tamil Nadu. yegna@cs.iitm.ernet.in

Neural Networks : the Official Journal of the International Neural Network Society
|July 20, 2002
PubMed
Summary
This summary is machine-generated.

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Processing group delay spectrograms for study of formant and harmonic contours in speech signals.

The Journal of the Acoustical Society of America·2024
Same author

Analysis of phase derivatives of speech signals.

The Journal of the Acoustical Society of America·2022
Same author

Group delay spectrogram of speech signals without phase wrapping.

The Journal of the Acoustical Society of America·2022
Same author

Analysis of aperiodicity in artistic Noh singing voice using an impulse sequence representation of excitation source.

The Journal of the Acoustical Society of America·2020
Same author

Subsegmental level analysis of high arousal speech using the zero-time windowing method.

The Journal of the Acoustical Society of America·2019
Same author

Determination of glottal open regions by exploiting changes in the vocal tract system characteristics.

The Journal of the Acoustical Society of America·2016

Autoassociative neural networks (AANNs) offer a non-linear approach to pattern recognition, outperforming Gaussian models by capturing complex data distributions. AANNs provide a flexible error surface for enhanced data distribution matching, as demonstrated in speaker verification.

Area of Science:

  • Pattern Recognition
  • Machine Learning
  • Artificial Intelligence

Background:

  • Traditional Gaussian mixture models (GMMs) have limitations in capturing complex data distributions due to fixed Gaussian shapes and a priori determined mixture counts.
  • Pattern recognition aims to identify class characteristics from training data feature vectors.

Purpose of the Study:

  • To investigate the potential of non-linear models, specifically autoassociative neural networks (AANNs), for pattern recognition.
  • To demonstrate AANNs' ability to capture data distributions more effectively than GMMs.
  • To propose a method for generating an error surface that matches data distributions.

Main Methods:

  • Utilized autoassociative neural network (AANN) models, which perform identity mapping.

Related Experiment Videos

  • Analyzed the training error surface generated by AANNs in the feature space.
  • Developed a method to create an error surface tailored to the input data distribution.

    • Main Results:

    • The training error surface of AANN models provides valuable insights into input data distribution characteristics.
    • AANN models demonstrate superior ability in capturing complex data distributions compared to GMMs.
    • The proposed method effectively generates an error surface that aligns with the data distribution.

    Conclusions:

    • Autoassociative neural networks (AANNs) offer a powerful non-linear alternative for pattern recognition tasks.
    • AANNs' error surface analysis is a viable method for understanding data distributions.
    • The effectiveness of AANNs in distribution capturing is validated through speaker verification applications.