00項目預(yù)研語音

1、項目預(yù)研Engine 項目組華技大學(xué)軟件學(xué)院2005信 息父項名稱父項標(biāo)識版本子項文檔名稱子項文檔標(biāo)識版本修 改 信 息修 改 者日 期舊版本修改標(biāo)識原 因?qū)?核日 期新版本批 準(zhǔn)日 期配 置 信 息項目名稱移動項目標(biāo)識Engine-BM-2005-01版 本1.0文檔名稱項目預(yù)研文檔標(biāo)識PDS-2005-01版 本1.0編 輯撰寫人時 間2005-5-15版 本1.0審 核 批 準(zhǔn)審 核日 期2005-5-17版 本1.0批 準(zhǔn)日 期2005-5-17項目預(yù)研1 引言1.1 編寫目的項目預(yù)研目的旨在梳理語音開發(fā)流程，確定應(yīng)用開發(fā)框架。1.2 背景項目名稱：移動（b-mobile 讓隨身移動）

2、項目委托項目開發(fā)系統(tǒng)開發(fā)：華：華技大學(xué)軟件學(xué)院技大學(xué)軟件學(xué)院Engine 項目組：VS.NET、Speech Application SDK VerBeta 1.11.3 定義Speech Application SDK： 微軟語音應(yīng)用開發(fā)包1.4 參考資料Speech Application SDK 開系統(tǒng)需求說明檔2 應(yīng)用程序組織模式Speech Application SDK 程序模式之2.1基礎(chǔ)篇（SALT）The SAPI API provides a high-levelerface betspeech engines. SAPI implements all the low-le

3、veln an application ands needed tocontrol and manage the real-time operations of various speech engines.The two basic types of SAPI engines are text-to-speech (TTS) systems andspeechspoken spokenrecognizers. TTS systems synthesize text strings and filesoaudio using synthetic voi. Speech recognizers

4、convert humanaudioo readable text strings and files.Speech Application SDK 程序模式之2.2componentsoftheSpeechApplication PlatformRequired ComponentsDeploying a speech-enabled Web application using SALT markuprequires three components.1.An ASP.NET serverThe Web server generates Wges containing HTML, SALT,

5、and embedded script. The script controls the dialogue flow forvoice-onlyeractions. For example, if there are severalprompts on a page, the script defines the order in which the audio prompts play.2.A Speech ServerSpeech Server recognizes speech, and plays audio prompts and responses.3.A cntThe Speec

6、h Platform supports two types of cnts:ephonyApplication Servicnts, and multimodal cnts with averofernet ernetExplorer running either Speech Add-in for Explorer or Speech Add-in forPocketernet Explorer.The following diagram illustrates these elements and the types ofinformation they pros. It also ill

7、ustrates the relationship of theseelements to the Visual Studio .NET 2003 Speech Development Tools.3Common Usage ScenariosThis section illustrates three deployment configurations for commondeployment scenariost the Speech Platform supports.3.1ephony Scenariohis scenario,ephony Application Servi(TAS)

8、 is the cnt. Aephone acts as the terminal device, and connects to TAS through astandardephony board. Theephony board provides theerfacebetn theephone and TAS. At run time, TAS res on the Webserver for application logic, and on Speech Server for audio signalprosing.When the user dials a phone number

9、for aephony service, the callconnects to TAS. TAS assotes theephone call wivoice-onlySALTreter. Then TAS connects to the Web server and loads thedefault page for the applicationt provides the service for which thecaller is dialing. As the callereracts with the application, TASpasses audio and dual t

10、one multi-frequency (DTMF) input from thecaller to Speech Server, which performs speech recognition (SR),text-to-speech (TTS), and DTMF prosing.The SASDK includes a number of Dialog Speech Controlst supportComputer-Supportedmunications Applications (CSTA)servi. These include the AnswerCall, Transfer

11、Call, MakeCall,and DisconnectCall controls. Developers can use these controls toanswer, transfer, initiate, and disconnectephone calls, as well asgather call information, and send and receive CSTA events. TheSASDK also includes a SmexMessage (Simple Messaging Exten) controlt developers can use to se

12、nd and receive raw CSTA messages.3.2Desktop Multimodal Scenariohis scenario, the cnt isernet Explorer with SpeechAdd-in forernet Explorer installed. ASP.NETspeech-enabled Web application pages reside on the Web server.When the user enters a URL inernet Explorer, the Web serveropens the applications

13、default page. The Web server sends HTML,SALT, and JScript to the Speech Add-in on the desktop. SALT markuphe pagest the Web server sends to the cnt trigger speechrecognition and text-to-speech synthesis. In order to implement SALTfunctionality, at run time the Speech Add-in instantiates a sharedSAPI

14、 SR engine. If nesary, the Speech Add-in also instantiates aTTS and a prompt engine on the cnt. These engines on the desktopcnt perform all prompting, speech recognition, and text-to-speechsynthesis.Note Multimodal applications using a desktop c nt can be deployed using only the SASDK.3.3Windows Mob

15、iMultimodal Scenarioased Pocket PC 2003 (Pocket PC)his scenario, the cnt is Pocketernet Explorer with the SpeechAdd-in forPocketernet Explorer installed. ASP.NETspeech-enabled Web application pages reside on the Web server,along with the application grammars, and a configuration filecontaining the U

16、RL to the Speech Servert performs speechprosing.When the user enters a URL on Pocket PC, the Web server opens theapplications default .aspx page. The Web server also sends the URLpoing to Speech Server. The paget the Web server sendscontains HTML, SALT, and JScript. When the user taps aspeech-enable

17、d HTML element and talks, Pocket PC sends the audio to Speech Server. Along with the compressed audio, Pocket PC sendseither an inline recognition grammar, or a poer to the location of anexternally-d recognition grammart is bound totspeech-enabled element. If the recognition grammar is an inlinegram

18、mar, Speech Server loads the grammar and performs speechrecognition. If the grammar is an externally-d grammar, SpeechServerdownloads a copy of the grammar, loads the grammar,and then performs speech recognition.After the recognizer finishes, Speech Server sends SemMarkupLanguage (SML) output to the Pocket PC along wiudio for promptsif the application dialogue flow requires the appli