Vowel-driven connected mandarin digit recognition system

Ruey Ching Shyu, Jhing Fa Wang, Chang Ching Huang, Chung Hsien Wu, Jau Yien Lee

Research output: Chapter in Book/Report/Conference proceedingConference contribution

2 Citations (Scopus)

Abstract

A vowel-driven connected Mandarin digit recognition system is presented in this paper. Since Mandarin is a monosyllabic language, each syllable has a consonant (may be null) and a vowel. Hence, we develop a segmentation algorithm to segment a connected digit string into the consonant and stable-region parts, where a stable region may contain one or more vowels. The stable-region parts are recognized first. The Mandarin digits, except (1,7) and (6,9), can be recognized just by their vowel parts. While the remaining digit pairs (1,7) and (6,9), which have the same vowel parts, are further distinguished by the special acoustic characteristic of the consonant part. The recognition system is tested in a speaker-dependent mode and under two different cases; known string length and unknown string length. The string error rates of 5.3 % and 7.4 % are achieved in the known-length and unknown-length case respectively.

Original languageEnglish
Title of host publicationProc 1993 IEEE Reg 10 Conf Comput Commun Control Power Eng
Editors Anon
PublisherPubl by IEEE
Pages638-641
Number of pages4
ISBN (Print)0780312333
Publication statusPublished - 1993
EventProceedings of the 1993 IEEE Region 10 Conference on Computer, Communication, Control aand Power Engineering. Part 3 (of 5) - Beijing, China
Duration: 1993 Oct 191993 Oct 21

Publication series

NameProc 1993 IEEE Reg 10 Conf Comput Commun Control Power Eng

Other

OtherProceedings of the 1993 IEEE Region 10 Conference on Computer, Communication, Control aand Power Engineering. Part 3 (of 5)
CityBeijing, China
Period93-10-1993-10-21

All Science Journal Classification (ASJC) codes

  • General Engineering

Fingerprint

Dive into the research topics of 'Vowel-driven connected mandarin digit recognition system'. Together they form a unique fingerprint.

Cite this