Piwho

Piwho is python wrapper around MARF <http://marf.sourceforge.net/>__ speaker recognition framework for the Raspberry pi and other SBCs. With the Piwho you can implement speaker recognition in your projects.

.. image:: https://travis-ci.org/Adirockzz95/Piwho.svg?branch=v1.3.0 :target: https://travis-ci.org/Adirockzz95/Piwho

|

Blink example

.. code:: python

import RPi.GPIO as GPIO
import time
from piwho import recognition, vad

def blink(pin):
    GPIO.setmode(GPIO.BOARD)
    GPIO.setup(pin, GPIO.OUT) 

    for i in range(0,10):
        GPIO.output(pin,GPIO.HIGH)
        time.sleep(1)
        GPIO.output(pin,GPIO.LOW)
        time.sleep(1)
   
if __name__ == "__main__":
   recog = recognition.SpeakerRecognizer('./')
   vad.record()
   name = recog.identify_speaker()
   if name[0] == 'Abhishek':
       blink(11)

Tested on

• Ubuntu 15.10
• Pi 1 model B (raspbian wheezy, jessie)
• Pi 2 model B (raspbian wheezy, jessie, stretch)
• CHIP

Installation

Update the Pi

.. code:: bash

$ sudo apt-get update
$ sudo apt-get upgrade

You need to have JDK (min version: 1.7) installed on your Pi.

.. code:: bash

# verify jdk is installed
$ java -version

Pyaudio is required to run audio recording script. (Optional)

.. code:: bash

# Install portaudio
$ sudo apt-get install portaudio19-dev
# Install python dev package
$ sudo apt-get install python2.7-dev
# Install pyaudio
$ pip install pyaudio 

Piwho is on PyPI

.. code:: bash

$ pip install piwho

or clone the project from github

.. code:: bash

$ git clone https://www.github.com/Adirockzz95/Piwho.git
$ cd Piwho
$ python setup.py install

Tests

Tests are implemented using unittest framework:

.. code:: bash

$ pip install -r requirements.txt
$ python -m unittest discover -v ./tests

Documentation

• Training the model <docs/trainingmodel.rst>__\
• Recognition <docs/recognition.rst>__\
• Gender detection <docs/gender_piwho.rst>__\
• integrating with Jasper <docs/jasper.rst>__

Tips / Caveats

• Recognition/Training time depends on the length of an audio file.
• If possible overclock your Pi- use Turbo mode.
• Give maxmium RAM to CPU.
• Read MARF manual <http://marf.sourceforge.net/docs/marf/0.3.0.5/report.pdf>__ to know how the it works.
• Recognition speed is directly proportional to the CPU power.

Misc

Here are resources/similar projects I came across while working on this project.

• recognito <https://github.com/amaurycrickx/recognito>__ : Works very well on Pi, but data storage is not implemented.
• voiceid <https://code.google.com/archive/p/voiceid/>__
• ALIZE <http://mistral.univ-avignon.fr/>__
• Shout <http://shout-toolkit.sourceforge.net/use_case_diarization.html>__
• MARF manual <http://marf.sourceforge.net/docs/marf/0.3.0.6/report.pdf>__
• LIUM site <http://www-lium.univ-lemans.fr/diarization/doku.php/welcome>__

LICENSE

MIT <./LICENSE>__