An HMM based semi-automatic syllable labeling system for Manipuri language

Author

Nandakishor, Salam ; Dutta, S.K. ; Singh, L. Joyprakash

Author_Institution

Dept. of ECE, Speech & Image Process. Lab., Shillong, China

fYear

2015

fDate

15-16 May 2015

Firstpage

1044

Lastpage

1047

Abstract

This paper describes the development of a SemiAutomatic Syllable Labeling System for the Manipuri language which is a scheduled Indian language of Tibeto-Burman origin and very highly agglutinative language. A system is developed using HMM tool kit (HTK), version 3.4 and later analyzed using popular speech analysis software (WaveSurfer), version 8.5.8. The HTK has been used for training, segmentation and labeling of the data. The system can segment the manual phonetic transcription into syllables with timing information. The syllable offset and onset of the automatic and the manual syllable labeling are compared to calculate the “Time Deviation” (W) of each syllable. We have found the “Average Deviation” of the syllables of our system is 25 ms. The Detection Rate of the syllables of the system is calculated by considering various “Time Deviations”.

Keywords

hidden Markov models; natural language processing; speech processing; HMM based semiautomatic syllable labeling system; HMM tool kit; Indian language; Manipuri language; Tibeto-Burman language; WaveSurfer speech analysis software; data labeling; data segmentation; data training; hidden Markov model; phonetic transcription; syllable average deviation; syllable detection rate; syllable time deviation; timing information; Acoustics; Feature extraction; Hidden Markov models; Labeling; Manuals; Speech; Speech recognition; HTK; Manipuri; Phonetic segmentation; Phonetic transcription; Syllable labeling; WaveSurfer;

fLanguage

English

Publisher

ieee

Conference_Titel

Computing, Communication & Automation (ICCCA), 2015 International Conference on

Conference_Location

Noida

Print_ISBN

978-1-4799-8889-1

Type

conf

DOI

10.1109/CCAA.2015.7148548

Filename

7148548