Sample code for KeenASR speech recognition engine usage from Unity

TestKeenASR-Unity.cs

using UnityEngine;
using System.Collections;
using KeenResearch;

// A rudimentary showcase of how to use KeenASR SDK from within Unity.
// Upon pressing Start Listening button the app will listen for several words defined in the phrases array below (feel free to augment the array to your liking)
// Results will only be shown on the consol (XCode or Android Studio logcat)

public class TestKeenASR : MonoBehaviour {

	//GUI.T
	// Use this for initialization
	void Start () {
		// setup a few things before we initialize the SDK
		KeenASR.SetLogLevel (KeenASR.LogLevelInfo);
		KeenASR.onInitializedReceived += KeenASRInitialized;

		// init the SDK with the ASR bundle name
		Debug.Log ("Keen: Initializing KeenASR Plugin");
		// Initialization works in a slghtly different manner on Android and iOS. On Android
		// it will be done asynchroniously, whereas on iOS it's done 
		KeenASR.Initialize ("keenB2mQT-nnet3chain-en-us");			
		KeenASR recognizer = KeenASR.Instance;

		// setup events with the instance of the recognizer
		recognizer.onFinalASRResultReceived += FinalASRResult;
		recognizer.onPartialASRResultReceived += PartialASRResult;
		recognizer.onRecognizerReadyToListenAfterInterruptReceived += KeenASRReadyToListenAfterInterrupt;
		recognizer.onUnwindAppAudioBeforeAudioInterruptReceived += UnwindAppAudio;
	}

    // Update is called once per frame
    //void Update () {
    //	if (KeenASR.Instance!=null && KeenASR.Instance.GetRecognizerState()==KeenASR.RecognizerStateListening) {
    //		// here we just show how input levels change; you would poll the input levels somewhere else
    //		// and use it to drive a UI component (not print in the log file)
    //		//Debug.Log ("\tLevel: " + KeenASR.Instance.InputLevel ());
    //	}
    //}

	// After recognizer is initalized we can setup other resources (this could be done else where and in some cases
	// multiple decoding graphs may be setup and used independently
	public void KeenASRInitialized(bool status) {
		if (!status) {
			Debug.Log ("KeenASR SDK was not initialized properly");
			return;
		}

		KeenASR recognizer = KeenASR.Instance;
		recognizer.SetCreateJSONMetadata (true);

		string dgName = "sampleDG";
		//		if (!recognizer.CustomDecodingGraphWithNameExists (dgName)) {
		Debug.Log ("Keen: Creating decoding graph");
		string[] phrases = new string[] { "YES", "NO", "MAYBE", "SURE", "HOW ARE YOU", "I AM GOOD", "I'M GOOD",
			"I DON'T FEEL GOOD", "I FEEL GOOD", "I AM OKAY", "I'M OKAY", "I AM ALRIGHT", "I'M ALRIGHT" };
		// we don't have to recreate the decoding graph every time; we can instead just
		// check for the existance of the graph with specific name. Note however, that with
		// the latter approach you WILL NEED to force recreation of the graph if you change
		// the list of input phrases
		recognizer.CreateCustomDecodingGraphFromSentences (dgName, phrases);
		//		} else {
		//			Debug.Log ("Keen: Decoding graph already exists");
		//		}
		//		if (recognizer.IsEchoCancellationAvaialable ()) {
		//			Debug.Log ("Echo cancellation is available, turining it on");
		//			recognizer.PerformEchoCancellation (true);
		//		} else {
		//			Debug.Log ("Echo cancellation is not available on this device");
		//		}

		// we now use this decoding graph for recognition. Multiple decoding graphs can
		// exist on the device and be switched back and forth
		recognizer.PrepareForListeningWithCustomDecodingGraph(dgName);

		// when set to true, SDK will create audio recordings capturing audio that was 
		// passed to the engine (between start and end listening)
		// you can get the file pat via GetLastRecordingFilename(), once the recognizer 
		// stopped listening (e.g. in onFinalASRResultReceived callback
		recognizer.SetCreateAudioRecordings (true);

		// VAD (Voice Activity Detection) is used to automatically stop listening
		// It can be changed at any time (e.g. slightly reduced in partial callbacks, based 
		// on semantic interpretation of the partial result)
		// final result will be reported after this many seconds end silence
		recognizer.SetVADParameter (KeenASR.VadParamTimeoutEndSilenceForGoodMatch, 1f);
		recognizer.SetVADParameter (KeenASR.VadParamTimeoutEndSilenceForAnyMatch, 1f);
		// also review  KeenASR.VadParamTimeoutForNoSpeech and KeenASR.VadParamTimeoutMaxDuration
		// which also control when stopListening kicks in automatically
	}

    public void FinalASRResult(ASRResult result) {
		Debug.Log ("Keen FINAL RESULT:" + result.cleanText + ", conf: " + result.confidence + ", numWords: " + result.words.Length);
		foreach (ASRWord word in result.words) {
			if (word.isTag)
				Debug.Log ("Word " + word.text + " is a tag word");
			if (word.confidence < 0.8)
				Debug.Log ("Word " + word.text + " has LOW confidence");
		}
		KeenASR recognizer = KeenASR.Instance;
		if (recognizer != null)
			recognizer.ResetSpeakerAdaptation ();
		// For testing/demo purposes only; it's unlikely you would need to call this method from within
		// the FinalASRResult callback
//		Debug.Log("final callback recognizer state returns: " + KeenASR.Instance.GetRecognizerState());
		//Debug.Log("Audio file saved in: " + KeenASR.Instance.GetLastRecordingFilename());

	}

	public void PartialASRResult(string result) {
		Debug.Log ("Keen PARTIAL RESULT:" + result);
		// For testing/demo purposes only; it's unlikely you would need to call this method from within
		// the FinalASRResult callback
//		Debug.Log("partial callback, GetRecognizerState returns: " + KeenASR.Instance.GetRecognizerState());
	}

	public void UnwindAppAudio() {
		Debug.Log ("Unwinding app audio");
	}

	public void KeenASRReadyToListenAfterInterrupt() {
		Debug.Log ("App ready to listen again...");
		// TODO reanable UI elements, etc.
	}


	void OnGUI() {
		GUIStyle buttonStyle = new GUIStyle();		
		buttonStyle.fontSize = 70;
		if (GUI.Button (new Rect (100, 450, 400, 100), "Start Listening", buttonStyle)) {
			#if UNITY_IPHONE || UNITY_ANDROID
				KeenASR.Instance.StartListening();
			#endif
		}	
	}

	private void Awake() {
		
	}
}