Identify multiple speakers using machine learning models like SVM, Decision Trees, XgBoost etc. Mel Coefficients(MFCC) of the audio are used as features for the models
System packages:
ffmpeg == 4.2.4
python == 3.8.5
Python packages:
librosa == 0.7.2
pandas == 1.1.3
numpy == 1.19.2
sklearn == 0.23.2
six == 1.15.0
pydotplus
xgboost
Audio clips of each class are to be placed in a separate folder and their names start from '0','1','2'... with as many folders as there are speakers/classes. Here I had used audio clips of 5 seconds each. I had 50 audio clips for each class '0','1' and '2'.
preprocess.py - Run this to find audio features for each clip in the dataset. It creates a .csv file.
train.py - Multiple models are used to predict classes. Accuracies of each model is printed out.
test.py - Use this to test on new audio samples.