Merge pull request spokestack#17 from spokestack/wr-vad-trigger

brentspell · web-flow · commit f552275e247a · 2020-10-08T14:07:52.000-04:00
Added VAD Trigger
diff --git a/spokestack/vad/webrtc.py b/spokestack/vad/webrtc.py
@@ -46,7 +46,7 @@ def __init__(
         self._run_length: int = 0
 
     def __call__(self, context: SpeechContext, frame: np.ndarray) -> None:
-        """Processes a single frame of audio to detemine if voice is present
+        """ Processes a single frame of audio to determine if voice is present
 
         Args:
             context (SpeechContext): State based information that needs to be shared
@@ -78,3 +78,30 @@ def reset(self) -> None:
     def close(self) -> None:
         """ Close interface for use in pipeline """
         self.reset()
+
+
+class VoiceActivityTrigger:
+    """ Voice Activity Detector trigger pipeline component """
+
+    def __init__(self) -> None:
+        self._is_speech = False
+
+    def __call__(self, context: SpeechContext, frame: np.ndarray) -> None:
+        """ Activates speech context whenever speech is detected
+
+        Args:
+            context (SpeechContext): State based information that needs to be shared
+            between pieces of the pipeline
+            frame (np.ndarray): Single frame of PCM-16 audio from an input source
+
+        """
+        if context.is_speech != self._is_speech:
+            if context.is_speech:
+                context.is_active = True
+            self._is_speech = context.is_speech
+
+    def close(self) -> None:
+        self.reset()
+
+    def reset(self) -> None:
+        self._is_speech = False
diff --git a/tests/vad/test_webrtc.py b/tests/vad/test_webrtc.py
@@ -6,7 +6,7 @@
 import numpy as np
 
 from spokestack.context import SpeechContext
-from spokestack.vad.webrtc import VoiceActivityDetector
+from spokestack.vad.webrtc import VoiceActivityDetector, VoiceActivityTrigger
 
 
 @patch("webrtcvad.Vad.is_speech", return_value=True)
@@ -79,3 +79,19 @@ def test_vad_fall_untriggered():
         detector(context, frame)
         assert not context.is_speech
     detector.close()
+
+
+def test_voice_activity_trigger():
+    context = SpeechContext()
+    trigger = VoiceActivityTrigger()
+
+    frame = np.zeros(160, np.int16)
+
+    trigger(context, frame)
+    assert not context.is_active
+
+    context.is_speech = True
+    trigger(context, frame)
+    assert context.is_active
+
+    trigger.close()
