Study of noise robustness of First Formant Bandwidth (F1BW) method

The performance of speech recognition application under adverse noisy condition often becomes the topic of researchers regardless of the language used. Applications that use vowel phonemes require high degree of Standard Malay vowel recognition capability.In Malaysia, researches in vowel recognitio...

Full description

Saved in:
Main Authors: Mohd Yusof, Shahrul Azmi, Siraj, Fadzilah, Yaacob, S., M.P, Paulraj, Nazri, Ahmad
Format: Conference or Workshop Item
Language:English
Published: 2011
Subjects:
Online Access:http://repo.uum.edu.my/13624/1/99%20%281%29.pdf
http://repo.uum.edu.my/13624/
http://www.icoci.cms.net.my
Tags: Add Tag
No Tags, Be the first to tag this record!
Institution: Universiti Utara Malaysia
Record Id: my.uum.repo.13624
epo.13624
Summary:The performance of speech recognition application under adverse noisy condition often becomes the topic of researchers regardless of the language used. Applications that use vowel phonemes require high degree of Standard Malay vowel recognition capability.In Malaysia, researches in vowel recognition is still lacking especially in the usage of Malay vowels, independent speaker systems, recognition robustness and algorithm speed and accuracy. This paper presents a noise robustness study on an improved vowel feature extraction method called First Formant Bandwidth (F1BW) on three classifiers of Multinomial Logistic Regression (MLR), K-Nearest Neighbors (k-NN) and Linear Discriminant Analysis (LDA).Results show that LDA performs best in overall vowel classification compared to MLR and KNN in terms of robustness capability.