Article | Dataset | Features | Classifiers | Rate(%) |
---|---|---|---|---|
Liu and Huang (2002) [46] | Ten different singers, 30 different music for each singer (Chinese) | FMCV, PMCV | KNN | 80.0 |
Tsai and Wang (2006) [47] | Twenty-three different singers, 10 different music for each singer | MFCC | GMM | 87.8 |
Dharini and Revathy (2014) [26] | Ten different singers, 20 different soundtracks for each singer (Indian, Bengali) | PLP | K-means | 55.56 |
Eghbal-Zadeh et al. (2015) [18] | Artist20 (20 singers, 1413 songs) | MFCCs | KNN | 84.31 |
Xing (2017) [48] | Ten different singers, 10 different music for each singer | LPC | GMM | 81.8 |
Shen et al. (2019) [1] | MIR-1K dataset | MFCCs | LSTM | 88.4 |
Loni and Subbaraman (2019) [17] | Twenty-six different singers, 550 different songs (Indian) | Formants, vibrato, timbre, and harmonic spectral envelope | SVM | 86 |
Murthy et al. (2021) [8] | Indian popular singers’ database (IPSD), Artist20 | MFCCs, LPCCs, SDCs, chroma, spectogram | YSA-RF-CNN | 61.69–75.50 |
Noyum et al. (2021) [49] | Four different singers, 50 different songs for each singer | DWT | Linear SVM | 83.96 |
Costa et al. (2017) [29] | Latin Music Database, ISMIR 2004, and African music collection dataset | Spectrogram, RLBP, rhythm patterns (RP), statistical spectrum descriptors (SSD), and rhythm histograms (RH) | CNN SVM | 92 |
Li et al. (2021) [30] | Artist20, singer32 vs singer60 | Spectrogram | CRNN | 99.0–85.0 |
Nasrullah et al. (2019) [12] | Artist20 | Spectrogram | CRNN | 93.7 (F1) |
Sharma et al. (2019) [31] | Artist20 | MFCC | UBM T-matrix | 89.97 |
Zhang et al. (2022) [32] | Artist20 | Mel-spectrogram, articulation, rhythmic complexity, rhythmic stability, dissonance, tonal stability, modality, x-vector | CRNN | 81 (F1) |
Proposed model | Nine different singers, 20 different songs for each singer | MFCC, octave-based spectral contrast | Extra Tree | 89.4 |
Proposed model | Artist20 | MFCC, octave-based spectral contrast | KNN | 85.4 |