Speech recognition using SetInputToWaveFile ends prematurely - c#-4.0

Question

I want to do speech recognition of an audio file.
My code is pretty basic and derived from here. The problem is that it stops with every wave file prematurely after a few seconds even though some wave files are hours long.
How to make it scan the whole file?
namespace Stimmenerkennung
{
public partial class Form1 : Form
{
//...
Thread erkennung;
bool completed;
private void Form1_Load(object sender, EventArgs e)
{
erkennung = new Thread(erkennen);
erkennung.Start();
}
void erkennen()
{
using (SpeechRecognitionEngine recognizer =
new SpeechRecognitionEngine())
{
// Create and load a grammar.
Grammar dictation = new DictationGrammar();
dictation.Name = "Dictation Grammar";
recognizer.LoadGrammar(dictation);
// Configure the input to the recognizer.
recognizer.SetInputToWaveFile(#"REC01.wav");
// Attach event handlers for the results of recognition.
recognizer.SpeechRecognized +=
new EventHandler<SpeechRecognizedEventArgs>(recognizer_SpeechRecognized);
recognizer.RecognizeCompleted +=
new EventHandler<RecognizeCompletedEventArgs>(recognizer_RecognizeCompleted);
// Perform recognition on the entire file.
db("Starting asynchronous recognition...");
recognizer.RecognizeAsync();
while (!completed)
{
//fs((int)(100 / recognizer.AudioPosition.TotalSeconds * recognizer.AudioPosition.Seconds));
db(recognizer.AudioState.ToString());
Thread.Sleep(100);
}
}
}
// Handle the SpeechRecognized event.
void recognizer_SpeechRecognized(object sender, SpeechRecognizedEventArgs e)
{
if (e.Result != null && e.Result.Text != null)
{
db(e.Result.Text);
}
else
{
db(" Recognized text not available.");
}
}
// Handle the RecognizeCompleted event.
void recognizer_RecognizeCompleted(object sender, RecognizeCompletedEventArgs e)
{
if (e.Cancelled)
{
db(" Operation cancelled.");
}
if (e.InputStreamEnded)
{
db(" End of stream encountered.");
}
completed = true;
}
void db(string t)
{
this.textBox1.Invoke((MethodInvoker)delegate
{
textBox1.Text = textBox1.Text + Environment.NewLine + t;
//textBox1.Text = t;
});
}
}
}

Answer 1

You can split the file on few seconds chunks by the silences and feed the chunk to the recognizer separately. Then you can combine results into a single string.
You can use any voice activity detection implementation to perform the split, a simple energy-based VAD which calculate frame energy will be sufficient.
You can find some existing implementations of the VAD in CMUSphinx projet