对于国际交流来说,还是用 st (半音,semitone)更好.
如何求得基音同步标记
在PSOLA中基音同步标记的求取是很重要的步骤,我想请问各位高手:应该如何求取基音同步标记.能否提供一些可行的算法.哪里能够下载可参考的程序.谢谢!
The following is a summary I wrote in another place. Just copy it here, hope it's helpful.
QUOTE:
You are right about the importance of pitch marks. When I build a TTS corpus, I did the following:
1) Prepare texts for recording. You have to transcribe the text into phonetic symbols (such as pinyin), considering the phoneme coverage and balance, the prosodic coverage,
2) Record sound pressure and laryngograph signals simultaneously for the texts. The laryngograph signal is optional, but it helps to determine the GCI (glottal closure instant) of the speech.
3) Force the speech signal to be aligned with phonetic transcriptions. Alignment errors have to be checked and corrected. Maybe some sentences have to be re-recorded.
4) Generate pitch marks. Pitch marks are usually assigned at each GCI detected from the signal by certain algrithm. The marks might need to be mannually checked and corrected in case the GCI detection algrithm doesn't work well. The unvoice segements in the speech also need false "pitch" marks in order to make PSOLA work.
5) Generate the TTS inventory. The speech signals, pitch marks, and phonetic transcriptions are indexed and stored in some database. The synthesizer will retrieve the database and concatenate segments into sentences.
**Note: PSOLA is a patent of France Télécom. You could make a PSOLA-like synthesizer, but couldn't use the PSOLA algrithm for free.
用PRAAT SCRIPT 语言实现等时间比例提取基频值 (../index.htm../index.htmdispbbs.asp boardid=1&rootid=41975&id=41975)
-- 作者:清凉山人
-- 发布时间:2005-12-20 20:32:53
-- [讨论]用PRAAT SCRIPT 语言实现等时间比例提取基频值
用PRAAT SCRIPT 语言实现等时间比例提取基频值
看到有人问起如何用Praat软件实现等时间比例来提取不定时长音节的基频值,如将不同时长的每个音节都提取出10个等距离的基频值来.我特写下如下Script语言以供参考:
--------------------------------------------------------------
# name$表示声音文件名,请首先选择您要分析的声音

上一页下一页