Effect of Speaker Variation on the performance of Bangla ASR

Author

Ahamed, Bulbul ; Israt, Farhana ; Chowdhury, Syed Mustafizur Rahman ; Huda, Mohammad Nurul

Author_Institution

Northern Univ. Bangladesh, Dhaka, Bangladesh

fYear

2013

fDate

17-18 May 2013

Firstpage

1

Lastpage

5

Abstract

Accent based on Speaker Variation plays an important role on the performance of Bangla (widely used as Bengali) automatic speech recognition (ASR). In this paper, we have analyzed how much performance variation because of these accents is found. The analysis has showed a significant variation of word correct rates, word accuracies and sentence correct rates because of the different accents due speaker variation. It is expected that this experimental analysis will lead the researcher to improve the performance of ASR by minimizing the effects due to speakers´ variation.

Keywords

speech recognition; Bangla ASR performance; Bangla automatic speech recognition; Bengali; experimental analysis; performance variation; sentence correct rates; speaker variation effect; word accuracy; word correct rates variation; Automatic speech recognition; Educational institutions; Feature extraction; Hidden Markov models; Mel frequency cepstral coefficient; Speech; Bangla; Speaker variation; automatic speech recognition; hidden Markov model; triphone model;

fLanguage

English

Publisher

ieee

Conference_Titel

Informatics, Electronics & Vision (ICIEV), 2013 International Conference on

Conference_Location

Dhaka

Print_ISBN

978-1-4799-0397-9

Type

conf

DOI

10.1109/ICIEV.2013.6572578

Filename

6572578