Acoustic training system for speaker independent continuous Arabic speech recognition system

Author

Nofal, Maged ; Abdel-Raheem, Esam ; El Henawy, Hadia ; Kader, N.A.

Author_Institution

Ain Shams Univ., Cairo, Egypt

fYear

2004

fDate

18-21 Dec. 2004

Firstpage

200

Lastpage

203

Abstract

The paper presents an acoustic training system for building acoustic models for a medium vocabulary speaker independent continuous speech recognition system. A speech database is constructed to train the acoustic models. The acoustic models are constructed, and trained. A test set database is constructed to test the accuracy of the acoustic models. Also 4 language models of two main types: bigram and context free grammar, were built and used in tests. Our results show 5.26% and 2.72% word error rates for 1340 and 306 words bigram based language models, respectively. Our results show also 0.19% and 0.99% word error rates for 1340 and 306 words context free grammar based language models, respectively.

Keywords

Gaussian processes; acoustic signal processing; context-free grammars; database languages; hidden Markov models; natural languages; speaker recognition; vocabulary; acoustic model; acoustic training system; bigram; context free grammar; continuous Arabic speech recognition system; language model; signal construction; speech database; vocabulary speaker independent system; word error rate; Acoustic testing; Automatic speech recognition; Context modeling; Databases; Error analysis; Hidden Markov models; Loudspeakers; Natural languages; Speech recognition; Stochastic processes;

fLanguage

English

Publisher

ieee

Conference_Titel

Signal Processing and Information Technology, 2004. Proceedings of the Fourth IEEE International Symposium on

Print_ISBN

0-7803-8689-2

Type

conf

DOI

10.1109/ISSPIT.2004.1433721

Filename

1433721