Welcome to EPrints UTCC

Jittiwarangkul, Nutthacha and Anusurain, Ekkapol and Miyanaga, Yoshikazu (2005) F0 Adjustment for Noisy Thai Tone Recognition. UTCC Engineering Research Papers. ISSN 1906-1625

[img]
Preview
Text
700fulltext.pdf

Download (349kB) | Preview

Abstract

This paper presents Thai tone recognition. Fundamental frequency feature (F0) or differential F0 (ΔF0) is appended with Mel-Frequecy Cepstral Coefficient by employed Hidden Markov Model (HMM). The system used RSF/DRA (Running Spectrum Filtering/Dynamic Range Adjustment) [4] to eliminate noise and normalize dynamic range speech feature. Average Magnitude Different Function (AMDF) is applied to extract maximum amplitude then F0 is calculated and adjusted by seam 3E method. The Thai tone recognition system is evaluated at office environment speech, white noise 10 dB SNR and 20 dB SNR. The experiments are conducted by means of isolated words: /pa:/ and /pu:/, which have 5 tones, mid/M/, low/L/, fall/F/, high/H/, and rise/R/. They are spoken by 12 speakers: 6 males and 6 females. The Thai tone recognition accuracy rates of F0 tone feature are 80%, 75% and 65% for office environment speech, 20 dB SNR and 10 dB SNR respectively, meanwhile F0 tone feature are 85%, 75% and 65% respectively.

Item Type: Article
Subjects: Engineering > Computer and Multimedia Engineering
Divisions: School > School of Engineering
Depositing User: Kitikun Pongsak
Date Deposited: 16 May 2014 20:44
Last Modified: 16 May 2014 20:44
URI: http://eprints.utcc.ac.th/id/eprint/700

Actions (login required)

View Item View Item