Skip to main content
. 2020 Jul 6;20(13):3790. doi: 10.3390/s20133790

Table 1.

Comparison with existing literature.

Ref Year Dataset Classes Features Classifier Results
[9] 2016 Physionet Challenge 2016 [39] Normal(2488), Abnormal(665) Time-frequency,
Wavelet and statistical
LogitBoost, Random Forest Acc: 84.48%
[11] 2016 Physionet Challenge 2016 [39] Normal(2575), Abnormal(665) Dynamic time warping SVM Acc: 82.4%
[15] 2016 Physionet Challenge 2016 Normal(2575), Abnormal(665) 124 Time-frequency features Adaboost, CNN Acc: 89%
[12] 2016 Self-collected Normal(132), Abnormal seven classes(131) Arash-Band SVM Acc: 87.45%
[36] 2017 Self-collected Small VSD(60), Large VSD(60) Statistical, DWT features Multilayer Perceptron (MLP) Acc: 96.6%
[35] 2017 Self-collected Normal, VSD (STFT), MFCC KNN Acc: 93.2%
[13] 2018 PhysioNet Computing in Cardiology Challenge Normal(2575), Abnormal(665) GFCC Weighted SVM Sen: 90.3% Spec: 89%
[17] 2018 UoC-murmur database, PhysioNet-2016 Normal(336), CHD(130), Normal/Abnormal(2435) Mel-Spectrogram, MFFC and sub-band envelopes CNN Acc: 81.5% Sen: 84.5%
[10] 2018 PhysioNet Computing in Cardiology Challenge-2016 Normal(50), Abnormal(50) Cepstrum Analysis SVM Acc: 95%
[38] 2018 Self-collected Normal(40), Abnormal(58) CBFE, FEUAP, FSDA, DDE KNN Acc: 84.39%
[32] 2019 Self-collected Normal(175), Abnormal(108) MFCC, normalized
average Shannon energy
SVM Acc: 92.6%
This work 2020 Self-collected Normal(140), Abnormal(140) MFCC + 1D-LTPs SVM Acc: 95.63%
This work 2020 Self-collected Normal(140), ASD(85), VSD(55) MFCC + 1D-LTPs SVM Acc: 95.24%