Classification of speaker's gender based on MFCC features
Gaétan Ramet
This project is about gender classification of speakers. We use different Machine learning algorithms to predict the gender of speakers based on the MFCCs in small audio files.
The data used for training and testing comes from the 'dev-clean' dataset of Librispeech. Download the dataset, then extract the archive and copy the folder beside the notebook.
This project make use of a few python libraries :
- numpy
- pysoundfiles for sound extraction
- python_speech_features for MFCCs extraction
- Sickit-learn for Machine learning algorithms
- Tensorflow for Neural networks
Make sure to download and install the necessary libraries before running the notebook.
Results are presented in the notebook, a few functions are written in the lib module for readability.