CDF-Matching for Automatic Tone Error Detection in Mandarin Call System

Author

Si Wei ; Hai-Kun Wang ; Qing-sheng Liu ; Ren-Hua Wang

Author_Institution

Lab of iFlytek Speech, Univ. of Sci. & Technol. of China, Hefei, China

Volume

4

fYear

2007

fDate

15-20 April 2007

Abstract

This paper introduces a tone pronunciation error detection algorithm for Mandarin CALL system. HMM is introduced to build tone model. FO after CDF-matching normalization is used as the feature of tone model. Tone error detection algorithm is based on the posterior probability calculated from HMM tone models. Comparing to the other normalization methods, CDF-matching can get better tone error detection performance with CC´s increasing from 0.76 to 0.79, which is close to that between human evaluators (0.83).

Keywords

computer aided instruction; hidden Markov models; linguistics; natural languages; speech recognition; CDF-matching normalization; HMM; Mandarin CALL system; automatic tone error detection; computer assisted language learning; posterior probability; tone pronunciation error detection algorithm; Computer errors; Databases; Detection algorithms; Distribution functions; Hidden Markov models; Humans; Natural languages; Probability; Speech; Testing; CALL; CDF-matching; Pronunciation error detection; Tone recognition;

fLanguage

English

Publisher

ieee

Conference_Titel

Acoustics, Speech and Signal Processing, 2007. ICASSP 2007. IEEE International Conference on

Conference_Location

Honolulu, HI

ISSN

1520-6149

Print_ISBN

1-4244-0727-3

Type

conf

DOI

10.1109/ICASSP.2007.367199

Filename

4218073