Timeout

[DKanarsky]

I wonder why handling timeout code in SpeechRecognizer.cs is commented?

        //// If we met timeout signal that speech ended
        //if (timeoutSamples != NO_TIMEOUT && remainingSamples <= 0)
        //{
        //    Log.Debug(TAG, "Timeout");
        //    OnTimeout();
        //}

[BrunoMoureau]

Hello, I think I didn't need to handle timeout in the project in which I used PocketSphinx. I guess I kept it commented only for information purpose.

If I remember well, I just had to start and stop the speech recognition at a given time.

[DKanarsky]

Hi! Ok, thanks for the reply. Your example helped me a lot!

[DKanarsky]

One more question... What do your think about CMU Sphinx vs. Google API? What is the benefit of using CMU Sphinx in adnroid os?

[BrunoMoureau]

For my project, I had to make a continuous speech recognition system. I tried to use Google API first but I noticed some limitations :

• (I don't know if it is a bug but) I wasn't able to last the listening more than ten seconds. PocketSphinx allows you to start and stop the recognition when you want to.
• The speech recognition needs an access to internet to be performed. PocketSphinx is installed on your device so it doesn't required any internet access to work. This is a lightweight engine so it doesn't take much space.
• I wasn't able to recognize a given word or a given sentence as "Ok Google" to wake up the system. You can do that with PocketSphinx. You can also adjust the speech recognition threshold to make the system much accurate.

These are three advantages that was really relevant in my case but there are certainly many others. If you have more questions, I'll try to answer you the best as I can.

Sorry if I made some mistakes but english isn't my native language :)

[DKanarsky]

Thanks! I tested my first xamarin speech recognize application based on your code. I needed to implement some commands recognition after a keyword. I slightly edited your code and also set up a timeout (3 sec). But I noticed that if a person kept speeking long after a keyword timeout didn't work. It was listening till the end of the speech. I wonder if it could be handled. I looked through your code again and found _decoder.GetInSpeech() method and that timeout depended on it. I think this is what I need but I don't completely undestand how it works. Besides I looked through Sphinx tutorial again and android demo project. It is said that "the SpeechRecognizer is the main class to access decoder functionality" but I didn't find implemetation of that class. Is your SpeechRecognizer.cs based on the original class?

Sorry if I made some mistakes but english isn't my native language :) Neither my)

[BrunoMoureau]

Well, I am sorry but I can't help you for the timeout because I don't really know how to make it works. To answer your last question, the SpeechRecognizer.cs file is based on the original java class you can find in Sphinx package. There is a private field in the class for the decoder : _decoder = new Decoder(config); I think this is what you are looking for when you talk about "access to decoder functionality" (for example) : _recorder.StartRecording();