Singer identification model using data augmentation and enhanced feature conversion with hybrid feature vector and machine learning

EURASIP Journal on Audio, Speech, and Music Processing

Table 1 Best results for each feature extractor with selected classifiers

Feature extraction method	Classifier	Accuracy	Precision	Recall	F-score
MFCC	Logistic regression	0.500	0.429	0.500	0.424
Spectogram	Extra tree	0.333	0.304	0.333	0.290
Beats	KNN	0.310	0.330	0.310	0.305
Spectral contrast	Logistic regression	0.500	0.526	0.500	0.505
Constant-q	SVM	0.357	0.334	0.357	0.321
Chroma-CENS	Random forest	0.405	0.406	0.405	0.379
Short-time Fourier transform (STFT)	Logistic regression	0.333	0.405	0.333	0.321
Spectral centroid	Decision tree	0.262	0.255	0.262	0.234
Bandwith	SVM	0.262	0.239	0.262	0.241
Root-mean-square error	SVM	0.381	0.278	0.381	0.313