Speech User Interface Pervasive Information Access Dragon dictation

Speech User Interface 語音介面

無所不在的資訊取得 Pervasive Information Access

應用實例: 文字輸入 Dragon dictation (聲龍聽寫) http: //itunes. apple. com/us/app/dragondictation/id 341446764? mt=8

完整之語音對話系統架構 Dialogue Management Automatic Natural Language Speech Recognition Understanding signal words Natural Language Generation Planning logical form words Text-tospeech

語音辨識的型態 連續vs. 非連續語音(continuous vs. non-continuous) 語者相關或無關 (speaker independent vs. dependent) 即興或朗讀文章(spontaneous vs. read) 關鍵字搜尋或全句辨識(keyword spotting vs. continuous recognition of spoken words) 字彙集大或小(small vs. large vocabulary set)

語音辨識技術 隱藏式馬可夫模型 (Hidden Markov Model) 參考論文: A tutorial on hidden Markov models and selected applications in speech recognition

語音辨識系統評估 透過 word error rate (WER) 來評估語音辨 識系統的表現 Error. Rate = 100*(Subs + Ins + Dels) / Nwords REF: I WANT TO GO HOME *** REC: * WANT TWO GO HOME NOW SC: D C S C C I 100*(1 S+1 I+1 D)/5 = 60%

英文 TTS 線上展示 AT & T Natural Voices Good evening, class. Today we are going to discuss an important type of human-computer interface: speech UI, also known as voice UI. We will demonstrate a TTS engine developed by AT & T, which, in my opinion, is the best TTS so far.

語音合成技術 Raw Text in Text Analysis Text Normalization Part-of-Speech tagging Homonym Disambiguation Phonetic Analysis Dictionary Lookup Grapheme-to-Phoneme (LTS) Prosodic Analysis Boundary placement Pitch accent assignment Duration computation Waveform synthesis Speech out

波形合成方法 Concatenative synthesis: based on the concatenation (or stringing together) of segments of recorded speech (將預錄的語音片 段串連起來) Formant synthesis: created using additive synthesis and an acoustic model with various fundamental frequency, voicing, and noise levels. Articulatory synthesis: synthesizing speech based on models of the human vocal tract

波形合成: 連鎖合成法 目前所有商業語音合成系統均採用 Concatenative Synthesis連鎖合成法，可再細分 為以下三類: Diphone Synthesis – Units are diphones; middle of one phone to middle of next. – Why? Middle of phone is steady state. – Record 1 speaker saying each diphone Unit Selection Synthesis – Larger units (Record 10 hours or more, so have multiple copies of each unit) – Use search to find best sequence of units Domain-specific synthesis: concatenates prerecorded words and phrases to create complete utterances

語音對話系統 Speech conversational system SIRI: 基於美國國防部 Cognitive Assistant that Learns and Organizes (CALO) project 以語音為基礎的個人虛擬助理 http: //en. wikipedia. org/wiki/Siri_(softwa re)

展示影片 A conversation with Siri on the i. Phone 4 S

主要技術 Conversational Interface: 語音辨識核心由 Nuance 所提供。 Personal Context Awareness: CALO 計畫 相關技術。 Service Delegation: 資訊搜尋與服務提供， 有多家公司參與。

資料與服務蒐尋 Open. Table, Gayot, City. Search, Boo. Rah, Yelp, Yahoo Local, Reserve. Travel, Localeze for restaurant and business questions and actions; Eventful, Stub. Hub, and Live. Kick for events and concert information; Movie. Tickets, Rotten. Tomatoes and the New York Times for movie information and reviews; True Knowledge, Bing Answers, and Wolfram Alpha for factual question answering; Bing, Yahoo and Google for web search.

語音介面: 實用面之問題 Major problems: – modes (no feedback) • certain commands only work when in specific states – deep hierarchies (also known as voice mail hell) Verbose feedback wastes time/patience – only confirm consequential things – use meaningful, short cues Interruption – half-duplex communication (i. e. , no barge-in support) Too much speech on the part of customer is tiring Speech takes up space in working memory – can cause problems when problem solving

語音介面開發標準 Voice. XML (VXML) is the W 3 C's standard XML format for specifying interactive voice dialogues between a human and a computer. 目前版本 Voice. XML 2. 1 Voice. XML 3. 0 (working draft)

語音介面開發 具 語音辨識: CMU Sphinx; Open Source Toolkit For Speech Recognition http: //cmusphinx. sourceforge. net/ 語音合成 festvox: http: //festvox. org/index. html 語音介面: Microsoft Speech API (SAPI 5. 3) Java Speech API

參考資料 X. Huang, A. Acero and H. W. Hn, Spoken Language Processing: A Guide to Theory, Algorithm and System Development, 2001. Rabiner and Schafer, Theory and Applications of Digital Speech Processing, 2010. Why is Siri Important?