Merge remote-tracking branch 'origin/master'

README.md

@@ -4,7 +4,7 @@ A speech to text to control an IoT object
 >This project is launched with a python 3.6
 
 > On Ubuntu, you need to run:  
-> `sudo apt update && sudo apt-get install portaudio19-dev`  
+> `sudo apt update && sudo apt-get install portaudio19-dev swig libpulse-dev libasound2-dev`   
 > `pip install --upgrade pyaudio`
 
 To run the underlying rasa server __look at [this documentation](plant_intent_recognizer/README.md)__

basic_speech_to_text.py

@@ -7,6 +7,27 @@
 import speech_recognition as sr
 
 
+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
+    :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
+    """
+    r = sr.Recognizer()
+    with sr.Microphone() as source:
+        if noise_level:
+            r.energy_threshold = noise_level
+        else:
+            r.adjust_for_ambient_noise(source)
+        audio = r.listen(source)
+    text = None
+    with suppress(sr.UnknownValueError, sr.RequestError):
+        text = r.recognize_sphinx(audio)
+    return keyword in text if text else False
+
+
 def speech_to_text(noise_level: int = None) -> Union[None, str]:
     """
     Function called to listen and convert to text the answer of the user.

plant_intent_recognizer/README.md

@@ -9,4 +9,4 @@ We do not use the full capacities of rasa but [only the NLU part](https://rasa.c
 - Start the server with:  
   `rasa run -m models --enable-api --cors “*” --debug`
 - Interact with the server with an HTTP POST request like:  
-  `curl localhost:5005/model/parse -d '{"text":"hello"}`
+  `curl localhost:5005/model/parse -d '{"text":"hello"}'`

voice_controller.py

@@ -1,6 +1,8 @@
+import threading
+from time import sleep, time
 from typing import Callable, Dict, List
 
-from basic_speech_to_text import speech_to_text
+from basic_speech_to_text import speech_to_text, is_keyword_said
 from plant_intent_recognizer.detect_intent import RasaIntent, Intent
 
 CALLBACK_INTENTS: Dict[Intent, List[Callable[[], None]]] = {}
@@ -33,17 +35,47 @@ def _trigger_function_on_intent(intent: Intent):
 
 class VoiceController:
 
-    def __init__(self):
-        self.rasa_intent = RasaIntent()
+    def __init__(self, active_time_delay=10):
+        """
+        :param active_time_delay time in seconds after the keyword was said before being not "active"
+        """
+        self._rasa_intent = RasaIntent()
+        self.active = False
+        self._stop = False
+        self.active_time_delay = active_time_delay
+        self.last_active_time = None
+
+        self._thread = threading.Thread(target=self.run, args=())
+        self._thread.daemon = True  # Daemonize thread
+
+    def stop(self):
+        """Stopping gracefully, might take a few seconds"""
+        self._stop = True
+        self._thread.join()
+
+    def start(self):
+        self._thread.start()  # Call run()
 
     def run(self):
-        while True:
-            text = speech_to_text()
-            print(f"text: {text}")
-            if text:
-                intent = self.rasa_intent.detect_intent(text)
-                print(f"intent: {intent}\n")
-                _trigger_function_on_intent(intent)
+        self._stop = False
+        while not self._stop:
+            if self.active:  # We actively listen to user command
+                text = speech_to_text()
+                print(f"text: {text}", flush=True)
+                if text:
+                    intent = self._rasa_intent.detect_intent(text)
+                    print(f"intent: {intent}\n", flush=True)
+                    _trigger_function_on_intent(intent)
+                    self.last_active_time = time()
+                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):
+                print("ACTIVE MODE", flush=True)
+                self.active = True
+                self.last_active_time = time()
+            else:
+                print("😴", end='', flush=True)  # Still in sleep mode
 
 
 if __name__ == '__main__':
@@ -64,4 +96,8 @@ def goodbye():
 
 
     vc = VoiceController()
-    vc.run()
+    vc.start()
+    print("I can continue to do stuff")
+    sleep(60)
+    print("Time to stop")
+    vc.stop()