speaker_recognizer_robot

Using MFCCs and HMMs to recognize a speaker

If you are looking for data or wav files to run the recognizer, request access for the drive here: https://drive.google.com/drive/u/0/folders/0BwXkbvw4LufrUzVNSUxiX1VEQUE.

Speaker Recognizer

To run the code, simply

roscore

rosrun speaker_recognizer recognizer.py

Audio Localization

We used a Raspberry Pi with two microphones to localize using a speaker's voice.

To record wav files on your RasPi, use amixer with the steps outlined in this script.

To copy your recorded file from the RasPi to your machine, use the secure copy command. This file can now be used as input to the angle_from_audio function in audio_localizer.py, which calculates the angle of the speaker from the robot.

scp pi@[IP Address]:~/test.wav .

Dependencies

1. python_speech_features

   Install it from pypi: pip install python_speech_features