Probabilistic Class Histogram Equalization for Robust Speech Recognition

In this letter, a probabilistic class histogram equalization method is proposed to compensate for an acoustic mismatch in noise robust speech recognition. The proposed method aims not only to compensate for the acoustic mismatch between training and test environments but also to reduce the limitatio...

Full description

Saved in:
Bibliographic Details
Published in:IEEE signal processing letters 2007-04, Vol.14 (4), p.287-290
Main Authors: Suh, Y, Mikyong Ji, Mikyong Ji, Hoirin Kim, Hoirin Kim
Format: Article
Language:English
Subjects:
Acoustic distortion
Acoustic noise
Acoustic testing
Acoustics
Additive noise
Classification
Equalization
Feature compensation
histogram equalization
Histograms
Noise robustness
Nonlinear acoustics
Nonlinear distortion
probabilistic class
Probabilistic methods
Probability theory
robust speech recognition
Speech recognition
Working environment noise
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:In this letter, a probabilistic class histogram equalization method is proposed to compensate for an acoustic mismatch in noise robust speech recognition. The proposed method aims not only to compensate for the acoustic mismatch between training and test environments but also to reduce the limitations of the conventional histogram equalization. It utilizes multiple class-specific reference and test cumulative distribution functions, classifies noisy test features into their corresponding classes by means of soft classification with a Gaussian mixture model, and equalizes the features by using their corresponding class-specific distributions. Experiments on the Aurora 2 task confirm the superiority of the proposed approach in acoustic feature compensation.
ISSN:1070-9908
1558-2361
DOI:10.1109/LSP.2006.884903