adsapien.pas

Async Professional 4.04

{*********************************************************}
{*                   ADSAPIEN.PAS 4.04                   *}
{*         Copyright (C) TurboPower Software 2002        *}
{*                 All rights reserved.                  *}
{*********************************************************}

{Global defines potentially affecting this unit}
{$I AWDEFINE.INC}

unit AdSapiEn;

interface

uses                                   
  Windows,
  ActiveX,
  Classes,
  Graphics,
  {$IFDEF Delphi5}
  OleServer,
  {$ENDIF}
  OleCtrls,
  StdVCL,
  SysUtils,
  Controls,
  OoMisc,
  AdISapi,
  AdSapiGr,
  {$IFDEF Delphi5}
  Contnrs,
  {$ENDIF}
  Messages,
  Forms,
  Dialogs,  
  AdExcept,
  ComObj;

const

  { Constants from the Speech API }

  { TTS Character Sets }

  ApdCHARSET_TEXT                  = $00;
  ApdCHARSET_IPAPHONETIC           = $01;
  ApdCHARSET_ENGINEPHONETIC        = $02;

  { TTS Options }

  ApdTTSDATAFLAG_TAGGED            = $01;

  { TTS Ages }

  ApdTTSAGE_BABY                   = 1;
  ApdTTSAGE_TODDLER                = 3;
  ApdTTSAGE_CHILD                  = 6;
  ApdTTSAGE_ADOLESCENT             = 14;
  ApdTTSAGE_ADULT                  = 30;
  ApdTTSAGE_ELDERLY                = 70;

  { TTS Features }

  ApdTTSFEATURE_ANYWORD            = $00001;
  ApdTTSFEATURE_VOLUME             = $00002;
  ApdTTSFEATURE_SPEED              = $00004;
  ApdTTSFEATURE_PITCH              = $00008;
  ApdTTSFEATURE_TAGGED             = $00010;
  ApdTTSFEATURE_IPAUNICODE         = $00020;
  ApdTTSFEATURE_VISUAL             = $00040;
  ApdTTSFEATURE_WORDPOSITION       = $00080;
  ApdTTSFEATURE_PCOPTIMIZED        = $00100;
  ApdTTSFEATURE_PHONEOPTIMIZED     = $00200;
  ApdTTSFEATURE_FIXEDAUDIO         = $00400;
  ApdTTSFEATURE_SINGLEINSTANCE     = $00800;
  ApdTTSFEATURE_THREADSAFE         = $01000;
  ApdTTSFEATURE_IPATEXTDATA        = $02000;
  ApdTTSFEATURE_PREFERRED          = $04000;
  ApdTTSFEATURE_TRANSPLANTED       = $08000;
  ApdTTSFEATURE_SAPI4              = $10000;

  { TTS Genders }

  ApdGENDER_NEUTRAL                = 0;
  ApdGENDER_FEMALE                 = 1;
  ApdGENDER_MALE                   = 2;

  { TTS Interfaces }

  ApdTTSI_ILEXPRONOUNCE            = $01;
  ApdTTSI_ITTSATTRIBUTES           = $02;
  ApdTTSI_ITTSCENTRAL              = $04;
  ApdTTSI_ITTSDIALOGS              = $08;
  ApdTTSI_ATTRIBUTES               = $10;
  ApdTTSI_IATTRIBUTES              = $10;
  ApdTTSI_ILEXPRONOUNCE2           = $20;

  { SR Features }

  ApdSRFEATURE_INDEPSPEAKER        = $00001;
  ApdSRFEATURE_INDEPMICROPHONE     = $00002;
  ApdSRFEATURE_TRAINWORD           = $00004;
  ApdSRFEATURE_TRAINPHONETIC       = $00008;
  ApdSRFEATURE_WILDCARD            = $00010;
  ApdSRFEATURE_ANYWORD             = $00020;
  ApdSRFEATURE_PCOPTIMIZED         = $00040;
  ApdSRFEATURE_PHONEOPTIMIZED      = $00080;
  ApdSRFEATURE_GRAMLIST            = $00100;
  ApdSRFEATURE_GRAMLINK            = $00200;
  ApdSRFEATURE_MULTILINGUAL        = $00400;
  ApdSRFEATURE_GRAMRECURSIVE       = $00800;
  ApdSRFEATURE_IPAUNICODE          = $01000;
  ApdSRFEATURE_SINGLEINSTANCE      = $02000;
  ApdSRFEATURE_THREADSAFE          = $04000;
  ApdSRFEATURE_FIXEDAUDIO          = $08000; 
  ApdSRFEATURE_IPAWORD             = $10000;
  ApdSRFEATURE_SAPI4               = $20000;

  { SR Supported grammar types }

  ApdSRGRAM_CFG                    = $1;
  ApdSRGRAM_DICTATION              = $2;
  ApdSRGRAM_LIMITEDDOMAIN          = $4;

  { SR Interfaces }

  ApdSRI_ILEXPRONOUNCE             = $0000001;
  ApdSRI_ISRATTRIBUTES             = $0000002;
  ApdSRI_ISRCENTRAL                = $0000004;
  ApdSRI_ISRDIALOGS                = $0000008;
  ApdSRI_ISRGRAMCOMMON             = $0000010;
  ApdSRI_ISRGRAMCFG                = $0000020;
  ApdSRI_ISRGRAMDICTATION          = $0000040;
  ApdSRI_ISRGRAMINSERTIONGUI       = $0000080;
  ApdSRI_ISRESBASIC                = $0000100;
  ApdSRI_ISRESMERGE                = $0000200;
  ApdSRI_ISRESAUDIO                = $0000400;
  ApdSRI_ISRESCORRECTION           = $0000800;
  ApdSRI_ISRESEVAL                 = $0001000;
  ApdSRI_ISRESGRAPH                = $0002000;
  ApdSRI_ISRESMEMORY               = $0004000;
  ApdSRI_ISRESMODIFYGUI            = $0008000;
  ApdSRI_ISRESSPEAKER              = $0010000;
  ApdSRI_ISRSPEAKER                = $0020000;
  ApdSRI_ISRESSCORES               = $0040000;
  ApdSRI_ISRESAUDIOEX              = $0080000;
  ApdSRI_ISRGRAMLEXPRON            = $0100000;
  ApdSRI_ISRRESGRAPHEX             = $0200000;
  ApdSRI_ILEXPRONOUNCE2            = $0400000;
  ApdSRI_IATTRIBUTES               = $0800000;
  ApdSRI_ISRSPEAKER2               = $1000000;
  ApdSRI_ISRDIALOGS2               = $2000000;

  { SR Sequences }

  ApdSRSEQUENCE_DISCRETE           = 0;
  ApdSRSEQUENCE_CONTINUOUS         = 1;
  ApdSRSEQUENCE_WORDSPOT           = 2;
  ApdSRSEQUENCE_CONTCFGDISCDICT    = 3;

  { SR Interference Types }

  ApdSRMSGINT_NOISE                = $0001;
  ApdSRMSGINT_NOSIGNAL             = $0002;
  ApdSRMSGINT_TOOLOUD              = $0003;
  ApdSRMSGINT_TOOQUIET             = $0004;
  ApdSRMSGINT_AUDIODATA_STOPPED    = $0005;
  ApdSRMSGINT_AUDIODATA_STARTED    = $0006;
  ApdSRMSGINT_IAUDIO_STARTED       = $0007;
  ApdSRMSGINT_IAUDIO_STOPPED       = $0008;

  { SR Training Requests }

  ApdSRGNSTRAIN_GENERAL            =  $01;
  ApdSRGNSTRAIN_GRAMMAR            =  $02;
  ApdSRGNSTRAIN_MICROPHONE         =  $04;

  { SS Error codes }

  ApdTTSERR_NONE                   = $00000000;
  ApdTTSERR_INVALIDINTERFACE       = $80004002;
  ApdTTSERR_OUTOFDISK              = $80040205;
  ApdTTSERR_NOTSUPPORTED           = $80004001;
  ApdTTSERR_VALUEOUTOFRANGE        = $8000FFFF;
  ApdTTSERR_INVALIDWINDOW          = $8004000F;
  ApdTTSERR_INVALIDPARAM           = $80070057;
  ApdTTSERR_INVALIDMODE            = $80040206;
  ApdTTSERR_INVALIDKEY             = $80040209;
  ApdTTSERR_WAVEFORMATNOTSUPPORTED = $80040202;
  ApdTTSERR_INVALIDCHAR            = $80040208;
  ApdTTSERR_QUEUEFULL              = $8004020A;
  ApdTTSERR_WAVEDEVICEBUSY         = $80040203;
  ApdTTSERR_NOTPAUSED              = $80040501;
  ApdTTSERR_ALREADYPAUSED          = $80040502;

  { SR Error Codes }

  ApdSRERR_NONE                    = $00000000;
  ApdSRERR_OUTOFDISK               = $80040205;
  ApdSRERR_NOTSUPPORTED            = $80004001;
  ApdSRERR_NOTENOUGHDATA           = $80040201;
  ApdSRERR_VALUEOUTOFRANGE         = $8000FFFF;
  ApdSRERR_GRAMMARTOOCOMPLEX       = $80040406;
  ApdSRERR_GRAMMARWRONGTYPE        = $80040407;
  ApdSRERR_INVALIDWINDOW           = $8004000F;
  ApdSRERR_INVALIDPARAM            = $80070057;
  ApdSRERR_INVALIDMODE             = $80040206;
  ApdSRERR_TOOMANYGRAMMARS         = $8004040B;
  ApdSRERR_INVALIDLIST             = $80040207;
  ApdSRERR_WAVEDEVICEBUSY          = $80040203;
  ApdSRERR_WAVEFORMATNOTSUPPORTED  = $80040202;
  ApdSRERR_INVALIDCHAR             = $80040208;
  ApdSRERR_GRAMTOOCOMPLEX          = $80040406;
  ApdSRERR_GRAMTOOLARGE            = $80040411;
  ApdSRERR_INVALIDINTERFACE        = $80004002;
  ApdSRERR_INVALIDKEY              = $80040209;
  ApdSRERR_INVALIDFLAG             = $80040204;
  ApdSRERR_GRAMMARERROR            = $80040416;
  ApdSRERR_INVALIDRULE             = $80040417;
  ApdSRERR_RULEALREADYACTIVE       = $80040418;
  ApdSRERR_RULENOTACTIVE           = $80040419;
  ApdSRERR_NOUSERSELECTED          = $8004041A;
  ApdSRERR_BAD_PRONUNCIATION       = $8004041B;
  ApdSRERR_DATAFILEERROR           = $8004041C;
  ApdSRERR_GRAMMARALREADYACTIVE    = $8004041D;
  ApdSRERR_GRAMMARNOTACTIVE        = $8004041E;
  ApdSRERR_GLOBALGRAMMARALREADYACTIVE = $8004041F;
  ApdSRERR_LANGUAGEMISMATCH        = $80040420;
  ApdSRERR_MULTIPLELANG            = $80040421;
  ApdSRERR_LDGRAMMARNOWORDS        = $80040422;
  ApdSRERR_NOLEXICON               = $80040423;
  ApdSRERR_SPEAKEREXISTS           = $80040424;
  ApdSRERR_GRAMMARENGINEMISMATCH   = $80040425;
  ApdSRERR_BOOKMARKEXISTS          = $80040426;
  ApdSRERR_BOOKMARKDOESNOTEXIST    = $80040427;
  ApdSRERR_MICWIZARDCANCELED       = $80040428;
  ApdSRERR_WORDTOOLONG             = $80040429;
  ApdSRERR_BAD_WORD                = $8004042A;
  ApdE_WRONGTYPE                   = $8004020C;
  ApdE_BUFFERTOOSMALL              = $8004020D;

type

  TApdSapiDuplex = (sdFull, sdHalf, sdHalfDelayed);
  TApdSapiWaitMode = (wmNone, wmWaitSpeaking, wmWaitListening,
                      wmRestoreListen);
  TApdCharacterSet = (csText, csIPAPhonetic, csEnginePhonetic);
  TApdTTSOptions = set of (toTagged);
  TApdTTSAge = (tsBaby, tsToddler, tsChild, tsAdolescent, tsAdult, tsElderly,
                tsUnknown);
  TApdTTSFeatures = set of (tfAnyWord, tfVolume, tfSpeed, tfPitch, tfTagged,
                            tfIPAUnicode, tfVisual, tfWordPosition,
                            tfPCOptimized, tfPhoneOptimized, tfFixedAudio,
                            tfSingleInstance, tfThreadSafe, tfIPATextData,
                            tfPreferred, tfTransplanted, tfSAPI4);
  TApdTTSGender = (tgNeutral, tgFemale, tgMale, tgUnknown);
  TApdTTSInterfaces = set of (tiLexPronounce, tiTTSAttributes, tiTTSCentral,
                              tiTTSDialogs, tiAttributes, tiIAttributes,
                              tiLexPronounce2);

  TApdSRFeatures = set of (sfIndepSpeaker, sfIndepMicrophone, sfTrainWord,
                           sfTrainPhonetic, sfWildcard, sfAnyWord,
                           sfPCOptimized, sfPhoneOptimized, sfGramList,
                           sfGramLink, sfMultiLingual, sfGramRecursive,
                           sfIPAUnicode, sfSingleInstance, sfThreadSafe,
                           sfFixedAudio, sfIPAWord, sfSAPI4);
  TApdSRGrammars = set of (sgCFG, sgDictation, sgLimitedDomain);
  TApdSRInterfaces = set of (siLexPronounce, siSRAttributes, siSRCentral,
                             siSRGramCommon, siSRDialogs, siSRGramCFG,
                             siSRGramDictation, siSRGramInsertionGui,
                             siSREsBasic, siSREsMerge, siSREsAudio,
                             siSREsCorrection, siSREsEval, siSREsGraph,
                             siSREsMemory, siSREsModifyGui, siSREsSpeaker,
                             siSRSpeaker, siSREsScores, siSREsAudioEx,
                             siSRGramLexPron, siSREsGraphEx, siLexPronounce2,
                             siAttributes, siSRSpeaker2, siSRDialogs2);
  TApdSRSequences = (ssDiscrete, ssContinuous, ssWordSpot, ssContCFGDiscDict,
                     ssUnknown);
  TApdSRInterferenceType = (itAudioStarted, itAudioStopped, itDeviceOpened,
                            itDeviceClosed, itNoise, itTooLoud, itTooQuiet,
                            itUnknown);
  TApdSRTrainingType = set of (ttCurrentMic, ttCurrentGrammar, ttGeneral);

  { Event types }

  TApdOnSapiError =
      procedure (Sender : TObject; Error : LongWord;
                 const Details : string; const Message : string) of object;
  TApdSapiNotifyEvent =
      procedure (Sender : TObject) of object;
  TApdSRInterferenceEvent =
      procedure (Sender : TObject;
                 InterferenceType : TApdSRInterferenceType) of object;
  TApdSRPhraseFinishEvent =
      procedure (Sender : TObject; const Phrase : string) of object;
  TApdSRPhraseHypothesisEvent =
      procedure (Sender : TObject; const Phrase : string) of object;
  TApdSRTrainingRequestedEvent =
      procedure (Sender : TObject; Training : TApdSRTrainingType) of object;
  TApdSRVUMeterEvent =
      procedure (Sender : TObject; Level : Integer) of object;
  TApdSSAttributeChanged =
      procedure (Sender : TObject; Attribute: Integer) of object;

  { Hook methods }

  TApdPhraseFinishMethod =
    procedure (Sender : TObject; Phrase : string; Results : Integer) of object;
  PApdPhraseFinishMethod = ^TApdPhraseFinishMethod;

  TApdRegisteredEventTypes = (etPhraseFinish);
  PApdRegisteredEvent = ^TApdRegisteredEvent;
  TApdRegisteredEvent = record
    CallBack : TApdPhraseFinishMethod;
    EventType : TApdRegisteredEventTypes;
    Active : Boolean;
  end;

  { Exceptions }

  EApdSapiEngineException = class (Exception)
    private
      FErrorCode : Integer;

    public
      constructor Create (const ErrCode : Integer; const Msg : string);
      { Note: ErrCode is before the string to prevent problems in compiling
              with C++ Builder }

    published
      property ErrorCode : Integer read FErrorCode;
  end;

  { TApdSSVoices }