How to use SpInprocRecoContext and receive recognition events with C#? RRS feed

  • Question

  • I'm very close to completing a personal project for modifying the Windows speech dictionary via C# (SAPI 5.4). The last bit I'm working on is how to get the SAPI phone set for a given word. I've found a way to do this via a C# form and spoken recognition obtained through SpSharedRecoContext. However I'm trying to get the recognition to work with a voice file (*.wav) as the input. I understand that this needs to be done via an SpInprocRecoContext.

    Nearly every example from Microsoft I've found regarding SAPI 5.4 recognition (like this one for VB) is for SpSharedRecoContext and not SpInprocRecoContext (and I believe I've seen comments that some of these examples are missing details). It also looks like the SAPI 5.1 and 5.4 SDK examples with C# only cover SpSharedRecoContext. Additionally, I've found multiple topics on Stack Overflow (mostly answered by Eric Brown, see topic 1topic 2topic 3) that mention using a SpInprocRecoContext requires more setup than SpSharedRecoContext, but I have yet to find a definitive answer for how to capture voice recognition events when using SpInprocRecoContext in C#.

    How can I proceed on this? Any help would be extremely appreciated!!

    Here is my code so far:

    using SpeechLib;
    using System;
    namespace SpeechTest
        class Program
            static void Main(string[] args)
                string MyText = "dolphins"; // Text string of interest
                string WaveFile = @"C:\Reco\MYWAVE.wav"; // Path to wav file used for voice recognition
                // Declare important recognition objects
                SpInprocRecognizer Recognizer;
                SpInProcRecoContext RecoContext;
                ISpeechRecoGrammar grammar;
                ISpeechFileStream MyFileStream;
                ISpeechVoice MyVoice;
                // Create recognizer and recognition context
                RecoContext = new SpInProcRecoContext();
                Recognizer = (SpInprocRecognizer)RecoContext.Recognizer;            
                // Set up recognition event handling            
                RecoContext.Recognition += new _ISpeechRecoContextEvents_RecognitionEventHandler(RecoContext_Recognition);
                // Set up the grammar
                grammar = RecoContext.CreateGrammar(); // Initialize the grammar
                grammar.DictationLoad("", SpeechLoadOption.SLOStatic); // Set up dictation grammar
                grammar.DictationSetState(SpeechRuleState.SGDSActive); // Activate the grammar
                // Set up audio input for SpInProcRecoContext
                SpObjectTokenCategory Category = new SpObjectTokenCategory();
                SpObjectToken AudioToken = new SpObjectToken();
                Recognizer.AudioInput = AudioToken;
                //Category.SetId(SpeechStringConstants.SpeechCategoryRecognizers); // <== generates a runtime error!!!
                //SpObjectToken EngineToken = new SpObjectToken();
                //Recognizer.Recognizer = EngineToken;
                //Category.SetId(SpeechStringConstants.SpeechCategoryRecoProfiles); // <== generates a runtime error!!!
                //SpObjectToken ProfileToken = new SpObjectToken();
                //Recognizer.Profile = ProfileToken;
                // Create an audio file stream from MyText
                MyFileStream = new SpFileStream(); // Create new SpFileStream instance
                TextToWave(MyText, WaveFile); // Call function to create a wav file that voices MyText
                MyFileStream.Open(WaveFile, SpeechStreamFileMode.SSFMOpenForRead, true);
                // Activate the recognizer and input the audio file stream into the recognizer
                RecoContext.State = SpeechRecoContextState.SRCS_Enabled;
                Recognizer.State = SpeechRecognizerState.SRSActive;
                Recognizer.AudioInputStream = MyFileStream; // <== generates a runtime error!!!
                // Output info and cleanup
                Console.WriteLine(MyText + " = " + SAPIPhonemes);
            static void TextToWave(string text, string file)
                SpFileStream fileStream = new SpFileStream();
                SpVoice voice = new SpVoice();
                fileStream.Open(file, SpeechStreamFileMode.SSFMCreateForWrite, true);
                voice.AudioOutputStream = fileStream;
            public static string SAPIPhonemes = null;
            public static void RecoContext_Recognition(int StreamNumber, object StreamPosition, SpeechRecognitionType RecognitionType, ISpeechRecoResult Result)
                // This event is recognized and all the below code works fine when using SpSharedRecoContext
                string SAPIPhonemes = null;
                SpPhoneConverter MyPhoneConverter = new SpPhoneConverter();
                MyPhoneConverter.LanguageId = 1033;
                foreach (ISpeechPhraseElement MyPhrase in Result.PhraseInfo.Elements)
                    SAPIPhonemes += " " + MyPhoneConverter.IdToPhone(MyPhrase.Pronunciation);

    For reference here is the form-based SpSharedRecoContext code that works:

    using SpeechLib;
    using System;
    using System.Windows.Forms;
    namespace RecoForm
        public partial class Form1 : Form
            // Speech Recognition Object
            SpSharedRecoContext listener;
            // Grammar object
            ISpeechRecoGrammar grammar;
            public Form1()
            private void Form1_Load(object sender, EventArgs e)
                // nothing
            public string ps;
            private void button1_Click(object sender, EventArgs e)
                if (btnListen.Text == "Start Listening")
                   // textBox1.Clear();
                        listener = new SpSharedRecoContext();
                        listener.Recognition += new _ISpeechRecoContextEvents_RecognitionEventHandler(listener_Reco);
                        grammar = listener.CreateGrammar(0);
                        grammar.DictationLoad("", SpeechLoadOption.SLOStatic);
                        btnListen.Text = "Stop Listening";
                        if (ps == "1")
                            ps = "0";
                    catch (Exception ex)
                else if (btnListen.Text == "Stop Listening")
                    btnListen.Text = "Start Listening";
                    if (ps == "0")
                        ps = "1";
            public void listener_Reco(int StreamNumber, object StreamPosition, SpeechRecognitionType RecognitionType, ISpeechRecoResult Result)
                string heard = Result.PhraseInfo.GetText(0, -1, true);
                textBox1.Text += " " + heard;
                SpPhoneConverter MyPhoneConverter = new SpPhoneConverter();
                MyPhoneConverter.LanguageId = 1033;
                foreach (ISpeechPhraseElement MyPhrase in Result.PhraseInfo.Elements)
                    textBox2.Text += " " + MyPhoneConverter.IdToPhone(MyPhrase.Pronunciation);
    // https://stackoverflow.com/questions/11935533/c-sharp-sapi-5-4-languages

    Here is another example (in VB) that combines the Microsoft examples (here and here) which STILL doesn't work (see the comments in Command1_Click to find the location where I encounter a runtime error):

    Imports SpeechLib
    Public Class Form1
        Const WaveFile = "C:\Reco\MYWAVE.wav"
        Dim WithEvents RC As SpInProcRecoContext
        Dim Recognizer As SpInprocRecognizer
        Dim myGrammar As ISpeechRecoGrammar
        Dim MyFileStream As SpeechLib.SpFileStream
        Dim MyVoice As SpeechLib.SpVoice
        Dim MyText As String
        Private Sub Form1_Load(sender As Object, e As EventArgs) Handles MyBase.Load
            On Error GoTo EH
            RC = New SpInProcRecoContext
            Recognizer = RC.Recognizer
            myGrammar = RC.CreateGrammar
            MyVoice = New SpVoice
            MyVoice.Voice = MyVoice.GetVoices("gender=male").Item(0)
            Dim Category As SpObjectTokenCategory
            Category = New SpObjectTokenCategory
            Dim Token As SpObjectToken
            Token = New SpObjectToken
            Recognizer.AudioInput = Token
            TextBox1.Text = "play the eight of clubs"
            If Err.Number Then ShowErrMsg()
        End Sub
        Private Sub Command1_Click(sender As Object, e As EventArgs) Handles Command1.Click
            MyFileStream = MakeWAVFileFromText(TextBox1.Text, WaveFile)
            Recognizer.AudioInputStream = MyFileStream ' ==> produces a runtime error!!!
        End Sub
        Private Sub RC_Recognition(ByVal StreamNumber As Long, ByVal StreamPosition As Object, ByVal RecognitionType As SpeechLib.SpeechRecognitionType, ByVal Result As SpeechLib.ISpeechRecoResult)
            On Error GoTo EH
            TextBox2.Text = Result.PhraseInfo.GetText
            If Err.Number Then ShowErrMsg()
        End Sub
        Private Sub ShowErrMsg()
            ' Declare identifiers:
            Const NL = vbNewLine
            Dim T As String
            T = "Desc: " & Err.Description & NL
            T = T & "Err #: " & Err.Number
            MsgBox(T, vbExclamation, "Run-Time Error")
        End Sub
        Private Function MakeWAVFileFromText(ByVal strText As String, ByVal strFName As String) As SpFileStream
            On Error GoTo EH
            ' Declare identifiers:
            Dim FileStream As SpFileStream
            Dim Voice As SpVoice
            ' Instantiate Voice and FileStream objects:
            Voice = New SpVoice
            FileStream = New SpFileStream
            ' Open specified .wav file, set voice output
            ' to file, and speak synchronously:
            FileStream.Open(strFName, SpeechStreamFileMode.SSFMCreateForWrite, True)
            Voice.AudioOutputStream = FileStream
            Voice.Speak(strText, SpeechVoiceSpeakFlags.SVSFIsXML)
            ' Close file and return reference to FileStream object:
            MakeWAVFileFromText = FileStream
            If Err.Number Then ShowErrMsg()
        End Function
    End Class
    ' https://msdn.microsoft.com/en-us/library/ee125184%28v=vs.85%29.aspx
    ' https://msdn.microsoft.com/en-us/library/ee125344(v=vs.85).aspx

    Tuesday, April 17, 2018 4:46 PM

All replies

  • Hello Exergist,

    >>but I have yet to find a definitive answer for how to capture voice recognition events when using SpInprocRecoContext in C#.

    Maybe you could refer some open source project regarding on  SpInprocRecoContext when MSDN doesn't provide code directly. Checked the below example.(NOT Test)

    C# (CSharp) SpeechLib.SpInProcRecoContext Code Examples

    Best Regards,

    Neil Hu

    MSDN Community Support
    Please remember to click "Mark as Answer" the responses that resolved your issue, and to click "Unmark as Answer" if not. This can be beneficial to other community members reading this thread. If you have any compliments or complaints to MSDN Support, feel free to contact MSDNFSF@microsoft.com.

    Wednesday, April 18, 2018 6:12 AM