On use of task independent training data in tandem feature extraction

The problem we address in this paper is, whether the feature extraction module trained on large amounts of task independent data, can improve the performance of stochastic models? We show that when there is only a small amount of task specific training data available, tandem features trained on task...

Full description

Saved in:
Bibliographic Details
Published in2004 IEEE International Conference on Acoustics, Speech, and Signal Processing Vol. 1; pp. I - 541
Main Authors Sivadas, S., Hermansk, H.
Format Conference Proceeding
LanguageEnglish
Published Piscataway, N.J IEEE 2004
Subjects
Applied sciences
Cepstral analysis
Data mining
Exact sciences and technology
Feature extraction
Hidden Markov models
Information, signal and communications theory
Principal component analysis
Signal processing
Spatial databases
Speech processing
Speech recognition
Stochastic processes
Telecommunications and information theory
Training data
Vocabulary
Learning
Performance evaluation
Stochastic model
Linear prediction
Cepstral analysis
Probabilistic approach
Hidden Markov models
Speech recognition
Signal processing
Feature extraction
Speech processing
Online AccessGet full text

Cover

More Information
Summary:The problem we address in this paper is, whether the feature extraction module trained on large amounts of task independent data, can improve the performance of stochastic models? We show that when there is only a small amount of task specific training data available, tandem features trained on task independent data give considerable improvement over perceptual linear prediction (PLP) cepstral features in hidden Markov model (HMM) based speech recognition systems.
ISBN:9780780384842
0780384849
ISSN:1520-6149
2379-190X
DOI:10.1109/ICASSP.2004.1326042