Monday, November 12, 2012

Gender Effect Canonicalization for Bangla ASR

During completing my Masters in Computer Science and Engineering at United International University. I have started and completed working in following research work. Here is a brief information of my research work,

Paper Title: Gender Effect Canonicalization for Bangla ASR
Main Author: Engr. Md.Asfak-Ur-Rahman
Other Authors:
Supervisor: Mohammad Nurul Huda

Abstract: "This paper presents a Bangla (widely used as Bengali) automatic speech recognition system (ASR) by suppressing gender effects. Gender characteristic plays an important role on the performance of ASR. If there is a suppression process that represses the decrease of differences in acoustic-likelihood among categories resulted from gender factors, a robust ASR system can be realized. In the proposed method, we have designed a new ASR incorporating the Local Features (LFs) instead of standard mel frequency cepstral coefficients (MFCCs) as an acoustic feature for Bangla by suppressing the gender effects, which embeds three HMM-based classifiers for corresponding male, female and geneder-independent (GI) characteristics. In the experiments on Bangla speech database prepared by us, the proposed system has achieved a significant improvement of word correct rates (WCRs), word accuracies (WAs) and sentence correct rates (SCRs) in comparison with the method that incorporates Standard MFCCs."

Keywords: acoustic model; automatic speech recognition; gender effects suppression; hidden Markov model

Research Objectives:
  • To design a Bangla ASR system by normalizing gender effects through a canonicalization process based on maximization of output probability.
  • To consider male, female and gender-independent (GI) voices.
  • To use local features (LFs) instead of mel frequency cepstral coefficients (MFCCs)
  • To increase the word correct rate, word accuracy and sentence  correct  rate.
Research Outcomes:
  1. We proposed an automatic speech recognition technique based on LFs for Bangla language by suppressing the gender effect incorporating HMM-based classifiers for male, female and Gender-Independent characteristics. 
  2. Our research paper includes following information,
  • The MFCC-based method incorporating GI classifier provides the higher performance than the method that does not incorporate GI classifier. The MFCC-based method incorporating GI exhibits its superiority at all the mixture components investigated.
  • The incorporation of GI HMM classifier improves the word correct rates, word accuracies and sentence correct rates significantly.
  • The proposed LF-based method shows a significant improvement of word correct rates, word accuracies and sentence correct rates for mixture component one.
Related Important Link:
  • Please CLICK HERE to view my complete research paper.
  • My research work is also accepted as JOURNAL in International Journal of Advanced Computer Science and Applications (IJACSA). But B.K.M Mizanur Rahman, Bulbul Ahamed and Khaled Mahmud has misused my research and submitted my paper to IJACSA without my approval. I am the main author of this paper. But they have changed naming order and added new names as author who was not involve during our research also. 
Copyright Notice

Contact authors for permission
The concept and writing of this article are protected by copyright. Excepting translations, you may not reproduce this articles online or in print without first obtaining written permission. Please contact me to obtain my written consent.

Include required attribution text
You must include the following text on the same page as your reprint and if the article is used in a print publication, the attribution must read:
"Reprinted with the permission of Engr. Md.Asfak-Ur-Rahman,"
Artwork and design cannot be reproduced
You may translate my articles into languages other than English, free of charge, without prior notice, and without soliciting our written permission, under the following conditions:
  1. Your publication must be freely available. You may not sell my content and concept that you have translated. 
  2. You must faithfully translate the article and concept as written.Except for minor issues of vernacular, you may not alter the my meaning. 
  3. You must link to my web page in which the article appeared, and must credit "Engr. Md.Asfak-Ur-Rahman,". You may not claim to have written the material yourself. (But you know that.)
You must include the following text on the same page as your translation:
"Translated with the permission of Engr. Md.Asfak-Ur-Rahman,