A program that extracts features with MFCC and finds the accuracy with linear SVM.
The accuracy was 83.6% using cross-validation when the dataset of audio sample was as follows and output to one linear SVM of the machine learning algorithm using the MFCC output by this program.
- Name: Jakobovski / Free Spoken Digit Dataset (FSDD)
- LICENCE: Creative Commons Attribution-ShareAlike 4.0 International
- Link: https://github.com/Jakobovski/free-spoken-digit-dataset
The above result can be executed in linearSVM.py. Download the audio sample from the link above.
This program is a test program to prove my next program. Link: https://github.com/OkamotoDaiki/MFCC
- Python 3.8.10
- numpy 1.21.4
- Scikit-learn 1.0.1
Place the audio file written in DEMO in the folloing file path and execute linearSVM.py
"recordings/*.wav"
- Oka.D.
- okamotoschool2018@gmail.com