(Old question, but no accepted answer, and appears quite high in google)
If you really want to do this in C++, you have to download the SAPI SDK, which does not come standard with Windows : http://www.microsoft.com/downloads/en/details.aspx?FamilyID=5e86ec97-40a7-453f-b0ee-6583171b4530&displaylang=en , select SpeechSDK51.exe
The best documentation you can find on SAPI is not on the web, it's in the SDK itself, in the Docs/ folder. The .chm explains everything really well. Here is an additional link to get you started.
However, it C++ is not a requirement for you, I strongly recommend you do it in C#. It's really much simpler (no COM components, no separate SDK, more doc on MSDN, more tutorials, ...) . See this CodeProject article; you'll have to remove all the GUI stuff, and all the speech synthesis stuff, and you'll see, speech recognition boild down to 10 lines of code. Quite impressive.
EDIT sample code, not compiled, not tested :
using System.Speech;
using System.Speech.Recognition;
// in constructor or initialisation
SpeechRecognitionEngine recognizer = null;
recognizer = new SpeechRecognitionEngine();
recognizer.SetInputToDefaultAudioDevice();
recognizer.SpeechRecognized += new EventHandler<SpeechRecognizedEventArgs>(recognizer_SpeechRecognized);
recognizer.RecognizeAsync(RecognizeMode.Multiple);
// The callback called when a sentence is recognized
private void recognizer_SpeechRecognized(object sender, SpeechRecognizedEventArgs e){
string text = e.Result.Text;
// Do whatever you want with 'text' now
}
ta dah, done
