工研院文字轉 語音技術 簡介

30
工工工工工工工工工工工工 工工工工工工工工工工工工工 工工工 2010/10/15

Upload: ziven

Post on 05-Jan-2016

112 views

Category:

Documents


1 download

DESCRIPTION

工研院文字轉 語音技術 簡介. 資通所前瞻技術中心副研究員 黃健紘. 大綱. 文字轉語音技術簡介 工研院 TTS 發展歷程 工研院 TTS 研發成果展示. 文字轉語音. TTS: Text-to-Speech 輸入 :文字 字串 ( text in characters ) 輸出 : 語音訊號 ( speech in samples) 將 輸入文字,轉換合成為語音輸出. 文字轉語音主要 步驟. 資料庫與 控制參數. 文字 處理. 語言 分析. 音韻 生成. 波形 合成. 文字 資訊. 口語文句. 語言參數. 聲學 參數. - PowerPoint PPT Presentation

TRANSCRIPT

(spoken text)(Markup Command Processing) (Text Normalization)2010/10/15 38.38% 6:21 or or ? (Sentence Segmentation)2010/10/155Copyright 2010 ITRI Arial(Linguistic)/ (Word Segmentation/Bracketing) (Text to Pronunciation) (Part-Of-Speech Tagging)2010/10/156Copyright 2010 ITRI Arial Mapping (acoustic) (energy) (pitch) (duration)Mapping MethodsRecurrent Neural Network (RNN)Classification And Regression Tree (CART)Hidden Markov Model (HMM)2010/10/157Copyright 2010 ITRI Arial(Pitch Synchronous Overlap and Add, PSOLA) (Formant Synthesis) (Linear Predictive Coding, LPC)MLSA (Mel Log Spectrum Approximation Filter)2010/10/158Copyright 2010 ITRI Arial TTS Demo (3/7)2010/10/1522 () Copyright 2010 ITRI Arial TTS Demo (4/7)2010/10/1523 ( ) Copyright 2010 ITRI Arial vs.TTS TTS TTS 2010/10/159Copyright 2010 ITRI ArialTTS 2010/10/1510 / Telematics()//IVR/CTINet-BookMIDPDA//Copyright 2010 ITRI ArialStephen Hawkings VoiceProfessor Stephen Hawking selects NeoSpeech Text-to-Speech as his new voice. Mar. 15, 20042010/10/1511

Copyright 2010 ITRI Arial TTS (1/3)cTTSRNN-based2010/10/1512

Copyright 2010 ITRI Arial TTS (2/3)iTTSCorpus-based102010/10/1513

Copyright 2010 ITRI Arial TTS (3/3)mTTSModel-based ()2010/10/15141

2

Copyright 2010 ITRI Arial mTTS (1/2) (Model Training)2-3/2010/10/1515Copyright 2010 ITRI Arial mTTS (2/2) (Text Analysis)TTS Microsoft Speech API (SAPI) Markup SAPI event SAPI markup 2010/10/1516Copyright 2010 ITRI Arial TTS ///2010/10/1517Copyright 2010 ITRI Arial TTS

2010/10/1518Copyright 2010 ITRI Arial-

2010/10/1519-Copyright 2010 ITRI Arial TTS Demo (1/7)2010/10/1520 ( )Copyright 2010 ITRI Arial TTS Demo (5/7) TTSGon-ling pat-lng, toh-s sin-thi ka-t.()Ti ka-t i sn-sim, ti pat-lng i sn-jm.()i pe-ing hoa-h sim, hoa-h sim, toh-s i ka-t chiok-hok. ()Chit kha-chhi kin-chon, khiok m-khng ch s lng, toh tng- b kha-chhi lng.()2010/10/1524

Copyright 2010 ITRI Arial TTS Demo (6/7) TTS ()Koan-ch-chi-ph-sat, hng-chhim poat-ch pho-l-bit-to s, chiu-kin g-n kai khong, t it-chh kh-eh.()Si-l-ch, sek put- khong, khong put- sek, sek chek-s khong, khong chek-s sek. Si sing hng sek, ek hok j-s.()Kiat t, kiat t, pho-l kiat t, pho-l cheng kiat t, ph-th sat p ho. ()2010/10/1525

Copyright 2010 ITRI Arial TTS Demo (7/7)TTS Model-based TTS TTS 2010/10/1526

Copyright 2010 ITRI Arial TTS Demo2010/10/1527

http://atc.ccl.itri.org.tw/ITRI TTSCopyright 2010 ITRI Arial TTS DemoGoogle Translate TTS Bing Translator TTS iFLYTEK ()SVOXAndroid TTS EngineNeoSpeechStephen Hawking TTS CepstralAT&T Labs ()Youtube 2010/10/1528Copyright 2010 ITRI ArialTTS / (expressive/emotional) (speaker adaptation)TTS///// (multilingual)2010/10/1529Copyright 2010 ITRI ArialTHANKS FOR YOUR ATTENTIONQ & A2010/10/1530E-mail: [email protected] 2010 ITRI Arial