Speech Recognition with Python

I stumbled across this great tutorial, so why not try it out?

http://realpython.com/python-speech-recognition/

As recommended, we will use SpeechRecognition.

After setting up my own repo, I found the author's:

http://github.com/realpython/python-speech-recognition

This also works with Raspberry Pi (using Python 3).

Contents

The contents are as follows:

• Prerequisites
  • For microphone use
  • Optional: monotonic (for Python 2)
  • For speech recognition
• Speech Engine
  • Smoke Test
  • Ambient Noise
• Speech testing
• And finally, the guessing game
• Raspberry Pi
• To Do

Prerequisites

Python 3 and pip installed (Python 2 is scheduled for End-of-life, although the instructions and code have been tested with Python 2 and an approprate requirements file for Python 2 is provided).

For microphone use

1. Check for pyaudio:

   >>> import pyaudio as pa
   Traceback (most recent call last):
     File "<stdin>", line 1, in <module>
   ImportError: No module named pyaudio
   >>>

[The next step is for linux; check the pyaudio requirements first.]

2. Install portaudio19-dev:

   $ sudo apt-get install portaudio19-dev
   

3. Install pyaudio:

   $ pip install --user pyaudio
   

4. Verify installation:

   >>> import pyaudio as pa
   >>> pa.__version__
   '0.2.11'
   >>>

Optional: monotonic (for Python 2)

SpeechRecognition recommends installing monotonic for Python 2 users.

1. Check for monotonic:

   $ pip list --format=freeze | grep monotonic
   

2. Install monotonic:

   $ pip install --user monotonic
   

3. Verify installation:

   $ pip list --format=freeze | grep monotonic
   monotonic==1.4
   $
   

For speech recognition

SpeechRecognition can be used as a sound recorder:

http://github.com/Uberi/speech_recognition/blob/master/examples/write_audio.py

This is probably fine for occasional use - but there are better options available.

1. Check for SpeechRecognition:

   $ pip list --format=freeze | grep SpeechRecognition
   

2. Install SpeechRecognition:

   $ pip install --user SpeechRecognition
   

3. Verify:

   >>> import speech_recognition as sr
   >>> sr.__version__
   '3.8.1'
   >>>

Speech Engine

The tutorial uses the Google Web Speech API, however installing PocketSphinx (which can work offline) is fairly easy.

Snowboy (which can also work offline) is an option for Hotword Detection, but perhaps unsuitable for speech recognition (SpeechRecognition tellingly refers to Snowboy as "Snowboy Hotword Detection").

For another online option, there is Wit.ai (which also has a Node.js SDK).

Smoke Test

The final step may take a few seconds to execute:

>>> import speech_recognition as sr
>>> r = sr.Recognizer()
>>> harvard = sr.AudioFile('audio_files/harvard.wav')
>>> with harvard as source:
...     audio = r.record(source)
... 
>>> type(audio)
<class 'speech_recognition.AudioData'>
>>> r.recognize_google(audio)
u'the stale smell of old beer lingers it takes heat to bring out the odor a cold dip restores health and zest a salt pickle taste fine with ham tacos al Pastore are my favorite a zestful food is the hot cross bun'
>>> 

Ambient Noise

>>> jackhammer = sr.AudioFile('audio_files/jackhammer.wav')
>>> with jackhammer as source:
...     audio = r.record(source)
... 
>>> r.recognize_google(audio)
u'the snail smell of old beer drinkers'
>>> with jackhammer as source:
...     r.adjust_for_ambient_noise(source)
...     audio = r.record(source)
... 
>>> r.recognize_google(audio)
u'still smell old gear vendors'
>>> 

[Slightly different from the tutorial's the snail smell of old gear vendors and still smell of old beer vendors.]

And:

>>> with jackhammer as source:
...     r.adjust_for_ambient_noise(source, duration=0.5)
...     audio = r.record(source)
... 
>>> r.recognize_google(audio)
u'the snail smell like old beermongers'
>>>

[Pretty much the same as the snail smell like old Beer Mongers.]

Speech testing

Using the speech recognition module:

$ python -m speech_recognition
A moment of silence, please...
Set minimum energy threshold to 259.109953712
Say something!
Got it! Now to recognize it...
You said hello hello
Got it! Now to recognize it...
You said the rain in Spain
Say something!
^C$

And:

>>> with mic as source:
...     audio = r.listen(source)
... 
>>> r.recognize_google(audio)
u'Shazam'
>>>

And, as stated in the article, a loud hand-clap generates an exception:

>>> with mic as source:
...     audio = r.listen(source)
... 
>>> r.recognize_google(audio)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/home/owner/.local/lib/python2.7/site-packages/speech_recognition/__init__.py", line 858, in recognize_google
    if not isinstance(actual_result, dict) or len(actual_result.get("alternative", [])) == 0: raise UnknownValueError()
speech_recognition.UnknownValueError
>>>

And finally, the guessing game

Run the guessing game as follows:

$ python guessing_game.py
I'm thinking of one of these words:
apple, banana, grape, orange, mango, lemon
You have 3 tries to guess which one.

Guess 1. Speak!
You said: banana
Incorrect. Try again.

Guess 2. Speak!
You said: Orange
Incorrect. Try again.

Guess 3. Speak!
You said: mango
Sorry, you lose!
I was thinking of 'apple'.
$

Raspberry Pi

After hooking up a Raspberry Pi with a Logitech 4000 webcam (for its microphone) and configuring with AlsaMixer, everything worked pretty much as expected with Python 3.

There were some installation stumbles, but sudo apt-get update fixed them.

It turned out that flac was required so it was also installed.

To Do

• Add original License (this is probably 'fair use' but better safe than sorry)
• Add monotonic as an optional component for Python 2
• Retry with PocketSphinx (works offline)
• Retry with Snowboy (works offline)
• Retry with Wit.ai (which also has a Node.js SDK)
• Try with Raspberry Pi (works nicely)
• Update for recent versions of pip
• Update code to conform to pylint, pycodestyle and pydocstyle
• Update requirements files to fix Snyk.io quibbles
• Update code for Python 3
• Add table of Contents