An Initial Study on Recognition of Confusing Sets in Mandarin Base-syllables by using Support Vector Machine

• Main Authors: Tung-Hung Hsieh, 謝東宏
• Other Authors: Ren-Yuan Lyu
• Format: Others
• Language: zh-TW
• Published: 2002
• Online Access: http://ndltd.ncl.edu.tw/handle/09742435067695951959

碩士 === 長庚大學 === 電機工程研究所 === 90 === Owing to Mandarin have some confusing sets, the confusing sets in syllables lead to difficulties for high recognition rate. This thesis is about Mandarin isolated syllable recognition for two-stage. First, we define five INITIAL confusing sets .The five confusing sets are (1)Unaspirated stop set：b,d,g。(2) Aspirated stop set：p,t,k。(3)Nasal-Lateral set：l,m,n。(4) Unaspirated fricative set：f,h。(5)Retroflex-Nonretroflex set：J,C,S,j,c,s。 First stage, we proceeded to train and test for reduce phone set. Confusing set is regarded as the same phone set , call “Generalized INITIAL”. For example, we can reduce phone set from 22 INITIAL sets to 7 INITIAL sets via we define five Generalized INITIAL confusing sets. Second stage, we use Support Vector Machine (SVM) as experiment manner.