Added hotword detection

tpusmb · Feb 13, 2020 · a9dfbd6 · a9dfbd6
1 parent 33ee753
commit a9dfbd6
Show file tree

Hide file tree

Showing 23 changed files with 52 additions and 209 deletions.
diff --git a/README.md b/README.md
@@ -9,4 +9,14 @@ A speech to text to control an IoT object
 
 To run the underlying rasa server __look at [this documentation](plant_intent_recognizer/README.md)__
 
-The start script is [voice_controller.py](voice_controller.py)
+The start script is [voice_controller.py](voice_controller.py)
+
+To use is_wake_up_word_said function, you need to figure out the index of your mic, to do that, you can run:
+```python
+import pyaudio
+p = pyaudio.PyAudio()
+for i in range(p.get_device_count()):
+    print(p.get_device_info_by_index(i))
+```
+
+And choose the one with the `'name': 'default'`
diff --git a/basic_speech_to_text.py b/basic_speech_to_text.py
@@ -1,16 +1,55 @@
 """
 File to manage the speech to text for our plant.
 """
+import struct
 from contextlib import suppress
+from time import time
 from typing import Union
 
+import pyaudio
 import speech_recognition as sr
+import pvporcupine
+
+
+def is_wake_up_word_said(input_device_index=13, sensitivity=0.5, keyword='hey pico', timeout=10):
+    keyword_file_path = [pvporcupine.KEYWORD_FILE_PATHS[keyword]]
+    num_keywords = len(keyword_file_path)
+
+    porcupine = pvporcupine.create(
+        library_path=pvporcupine.LIBRARY_PATH,
+        model_file_path=pvporcupine.MODEL_FILE_PATH,
+        keyword_file_paths=keyword_file_path,
+        sensitivities=[sensitivity] * num_keywords)
+
+    pa = pyaudio.PyAudio()
+    audio_stream = pa.open(
+        rate=porcupine.sample_rate,
+        channels=1,
+        format=pyaudio.paInt16,
+        input=True,
+        frames_per_buffer=porcupine.frame_length,
+        input_device_index=input_device_index)
+
+    start = time()
+    keyword_said = False
+    while not keyword_said and time() - start < timeout:
+        pcm = audio_stream.read(porcupine.frame_length)
+        pcm = struct.unpack_from("h" * porcupine.frame_length, pcm)
+
+        if porcupine.process(pcm):
+            keyword_said = True
+    audio_stream.close()
+    porcupine.delete()
+    return keyword_said
 
 
 def is_keyword_said(keyword="ok", noise_level: int = None) -> bool:
     """
     Function called to check if the keyword is said
     Work offline to preserve privacy at home
+
+    This function is DEPRECATED
+
     :param: keyword the word to be said only to start (SHOULD BE CAREFULLY CHOSEN/TESTED)
     :param: noise_level the level of ambient noise used to detect the end of a phrase
     :return: Is the keyword said or not

diff --git a/porcupine/keyword_files/linux/hey pico_linux.ppn b/porcupine/keyword_files/linux/hey pico_linux.ppn
diff --git a/porcupine/keyword_files/linux/hey pico_linux_compressed.ppn b/porcupine/keyword_files/linux/hey pico_linux_compressed.ppn
diff --git a/porcupine/keyword_files/raspberrypi/hey pico_raspberrypi.ppn b/porcupine/keyword_files/raspberrypi/hey pico_raspberrypi.ppn
diff --git a/porcupine/keyword_files/windows/hey pico_windows.ppn b/porcupine/keyword_files/windows/hey pico_windows.ppn
diff --git a/porcupine/keyword_files/windows/hey pico_windows_compressed.ppn b/porcupine/keyword_files/windows/hey pico_windows_compressed.ppn
diff --git a/porcupine/lib/README.md b/porcupine/lib/README.md
diff --git a/porcupine/lib/common/porcupine_compressed_params.pv b/porcupine/lib/common/porcupine_compressed_params.pv
diff --git a/porcupine/lib/common/porcupine_params.pv b/porcupine/lib/common/porcupine_params.pv
diff --git a/porcupine/lib/linux/x86_64/libpv_porcupine.a b/porcupine/lib/linux/x86_64/libpv_porcupine.a
diff --git a/porcupine/lib/linux/x86_64/libpv_porcupine.so b/porcupine/lib/linux/x86_64/libpv_porcupine.so
diff --git a/porcupine/lib/raspberry-pi/arm11/libpv_porcupine.a b/porcupine/lib/raspberry-pi/arm11/libpv_porcupine.a
diff --git a/porcupine/lib/raspberry-pi/arm11/libpv_porcupine.so b/porcupine/lib/raspberry-pi/arm11/libpv_porcupine.so
diff --git a/porcupine/lib/raspberry-pi/cortex-a53/libpv_porcupine.a b/porcupine/lib/raspberry-pi/cortex-a53/libpv_porcupine.a
diff --git a/porcupine/lib/raspberry-pi/cortex-a53/libpv_porcupine.so b/porcupine/lib/raspberry-pi/cortex-a53/libpv_porcupine.so
diff --git a/porcupine/lib/raspberry-pi/cortex-a7/libpv_porcupine.a b/porcupine/lib/raspberry-pi/cortex-a7/libpv_porcupine.a
diff --git a/porcupine/lib/raspberry-pi/cortex-a7/libpv_porcupine.so b/porcupine/lib/raspberry-pi/cortex-a7/libpv_porcupine.so
diff --git a/porcupine/lib/windows/amd64/libpv_porcupine.a b/porcupine/lib/windows/amd64/libpv_porcupine.a
diff --git a/porcupine/lib/windows/amd64/libpv_porcupine.dll b/porcupine/lib/windows/amd64/libpv_porcupine.dll
diff --git a/porcupine/porcupine.py b/porcupine/porcupine.py
diff --git a/porcupine/util_porcupine.py b/porcupine/util_porcupine.py
diff --git a/voice_controller.py b/voice_controller.py
@@ -2,7 +2,7 @@
 from time import sleep, time
 from typing import Callable, Dict, List
 
-from basic_speech_to_text import speech_to_text, is_keyword_said
+from basic_speech_to_text import speech_to_text, is_keyword_said, is_wake_up_word_said
 from plant_intent_recognizer.detect_intent import RasaIntent, Intent
 
 CALLBACK_INTENTS: Dict[Intent, List[Callable[[], None]]] = {}
@@ -70,7 +70,7 @@ def run(self):
                 elif time() - self.last_active_time > self.active_time_delay:
                     print("SLEEP MODE", flush=True)
                     self.active = False
-            elif is_keyword_said(noise_level=1500):
+            elif is_wake_up_word_said():
                 print("ACTIVE MODE", flush=True)
                 self.active = True
                 self.last_active_time = time()