DocumentCode :
2770481
Title :
Hybrid Features for Neural Network-Based Bangla ASR Incorporrating Velocity Coefficients (?)
Author :
Kotwal, Mohammed Rokibul Alam ; Hassan, Foyzul ; Daud, Shakib Ibn ; Alam, Md Shafiul ; Ahmed, Faisal ; Huda, Mohammad Nurul
Author_Institution :
Dept. of Comput. Sci. & Eng., United Int. Univ., Dhaka, Bangladesh
fYear :
2011
fDate :
7-9 Oct. 2011
Firstpage :
416
Lastpage :
420
Abstract :
This paper presents a Neural Network-based Bangla phoneme recognition method for Automatic Speech Recognition (ASR). The method consists of three stages: at first stage, a multilayer neural network (MLN) converts acoustic features, mel frequency cepstral coefficients (MFCCs), into phoneme probabilities, where the second stage computes velocity (Δ) coefficients from the phoneme probabilities by using three point linear regression (LR). Finally, the phoneme probabilities, velocity (Δ) coefficients and the input MFCCs, combined as hybrid features, are fed into a hidden Markov model (HMM) based classifier to obtain more accurate phoneme strings. From the experiments on Bangla speech corpus prepared by us, it is observed that the proposed method provides higher phoneme recognition performance than the existing method. Moreover, it requires a fewer mixture components in the HMMs.
Keywords :
cepstral analysis; hidden Markov models; neural nets; probability; signal classification; speech recognition; Bangla speech corpus; Mel frequency cepstral coefficients; acoustic features conversion; automatic speech recognition; hidden Markov model based classifier; multilayer neural network; neural network-based Bangla ASR; neural network-based Bangla phoneme recognition method; phoneme probabilities; phoneme strings; three point linear regression; velocity coefficients; Accuracy; Feature extraction; Hidden Markov models; Mel frequency cepstral coefficient; Speech; Speech recognition; automatic speech recognition; hidden Markov model; mel frequency cepstral coefficient; multilayer neural network; velocity coefficient;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Computational Intelligence and Communication Networks (CICN), 2011 International Conference on
Conference_Location :
Gwalior
Print_ISBN :
978-1-4577-2033-8
Type :
conf
DOI :
10.1109/CICN.2011.87
Filename :
6112900
Link To Document :
بازگشت