A Design of Chinese Speech Synthesis System

• Main Authors: HUANG, YII-CHERNG, 黃以承
• Other Authors: CHEN, CHIH-CHIEN
• Format: Others
• Language: zh-TW
• Published: 1994
• Online Access: http://ndltd.ncl.edu.tw/handle/34968875311477460879

碩士 === 國立中山大學 === 電機工程研究所 === 82 === This thesis describes an analysis and synthesis system of Mandarin speech signal from frequency domain. From the KLT, many speech frames can be segmented into several sub- words. The KLT also can be used for signal compression and signal reconstruction in the analysis part of this system. We can also use the critical band concept in acoustics to reduce the feature dimension in the frequence domain. The voiced/unvoiced band system is replaced for voiced/unvoiced frame system in the synthesis part of this system. It can effectively detect the period of the fre- quency bands between the relation of different signals from the periodic pulse and the source signal. At last, the voiced bands and the unvoiced bands signal portions are synthesized in the time domain and then added gether to prodduce Mandarin speech. The synthesized speech is quite distinguishable and it alsodeserves the tone of the original speakers.