Does SAPI 5.1 had good accuracy?

I had a doubt that does SAPI has good accuracy in voice recognition? when i try to read numbers from one to ten , the accuracy is not even 3%. i had a doubt that whether i did wrong or the SAPI itself is unstable in terms of accuracy? I shocked that the engine is not able to recognize 1 to 10 numbers perfectly. I fed up with its accuracy, do we have any other open source for C# for voice recognition

and my code is

i will give my code snippet here

    private Grammar Grammar Creation()
            {
                Choices NumberChoice = new Choices("one", "two", "three", "four", "five", "six", "seven", "eight","nine","ten");                    
                GrammarBuilder NumGrammarBuilder = new GrammarBuilder(NumberChoice);
                var NumGrammar = new Grammar(NumGrammarBuilder);
                return NumGrammar;
            }

and i am recognizing like this
  recognitionEngine.UnloadAllGrammars();
                Grammar cg = CreateSampleGrammar();
                recognitionEngine.LoadGrammar(cg);
                recognitionEngine.RecognizeAsync(RecognizeMode.Multiple);
       recognitionEngine.SetInputToDefaultAudioDevice();               

  recognitionEngine.SpeechRecognized += (s, args) =>
            {

                foreach (RecognizedWordUnit word in args.Result.Words)
                {
  开发者_高级运维                  // You can change the minimun confidence level here                    
                    if (word.Confidence > 0.8f)
                        freeTextBox.Text += word.Text + " ";
                }
                freeTextBox.Text += Environment.NewLine;
            };

Regards Bhuvan

here is a late reply but i hope it will help others, before judging any speech recognition engine you need to make sure somethings.

1. Is your dictionary model of engine is exactly same as of the accent of the voice you are testing. for example by default SAPI works on american accent , so if you're trying to run this engine on British accent or any other accent, its accuracy will surely be less than satisfactory. for this you need to add appropriate model !!

   2.Secondly which type of gram models you are testing?

   3.there are other factor like voice clarity, distortion back ground noise etc but I am sure you already ensures these basic factors.

If your accuracy less than claimed then you should make sure that you have configured your engine according to your inputs, like accent etc If you are trying to convert already saved sounds to speech then you should check the format of the sound, It should be uncompressed like in wav format

Google shows a few, but for a good one you would prob need to pay $$.