Investigation of Different Time-Frequency Representations for Intelligibility Assessment of Dysarthric Speech

Speech disorders linked to neurological problems affect person's ability to communicate through speech. Dysarthria is one of the speech disorders caused due to muscle weakness producing slow, slurred and less intelligible speech. Automatic intelligibility assessment of dysarthria from speech ca...

Full description

Saved in:
Bibliographic Details
Published in:IEEE transactions on neural systems and rehabilitation engineering 2020-12, Vol.28 (12), p.2880-2889
Main Authors: H M, Chandrashekar, Karjigi, Veena, Sreedevi, N
Format: Article
Language:English
Subjects:
Acoustics
constant-Q transform
convolutional neural networks
Disorders
Dysarthria
Feature extraction
Fourier transforms
Intelligibility
intelligibility assessment
Muscles
Neurological diseases
perceptual enhancement
short time Fourier transform
Signal resolution
single frequency filtering
Spectrogram
Spectrograms
Speech
Speech disorders
Time-frequency analysis
Transforms
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:Speech disorders linked to neurological problems affect person's ability to communicate through speech. Dysarthria is one of the speech disorders caused due to muscle weakness producing slow, slurred and less intelligible speech. Automatic intelligibility assessment of dysarthria from speech can be used as a promising clinical tool in treatment. This paper explores the use of perceptually enhanced Fourier transform spectrograms and Constant-Q transform spectrograms with CNN to assess word level and sentence level intelligibility of dysarthric speech from UA and TORGO databases. Constant-Q transform and perceptually enhanced mel warped STFT spectrograms performed better in the classification task.
ISSN:1534-4320
1558-0210
DOI:10.1109/TNSRE.2020.3035392