Parkinson’s Disease Diagnosis in Cepstral Domain Using MFCC and Dimensionality Reduction with SVM Classifier

Table 3

Performance comparison with recently published work.

Study	Method	Acc (%)	Sen. (%)	Spec. (%)
Sarkar et al. [42]	KNN + SVM	55.00 (LOSO on training database), 68.45 (LOSO on testing database)	60 (Training database)	50 (training database)
Canturk and Karabiber [43]	4 feature selection methods + 6 classifiers	57.5 (LOSO CV), 68.94 (10-fold)	54.28 (LOSO), 70.57 (10-fold)	80 (LOSO), 66.92 (10-fold)
Eskidere et al. [39]	Random subspace classifier ensemble	74.17 (10-fold CV)	Did not report	Did not report
Vadovský and Parali [40]	C4.5 + C5.0 + random forest + CART	66.5 (4-fold CV with pronouncing numbers), 65.86 (5-fold CV with pronouncing numbers)	Did not report	Did not report
Kraipeerapun and Amornsamankul [41]	Stacking + complementary neural Networks (CMTNN)	Average 75% (10-fold CV)	Did not report	Did not report
Ali et al. [44]	Multimodal approach	70	Not reported	Not reported
Benba et al. [29]	HFCC-SVM	87.5	90.00	85.00
Li et al. [32]	SVM + FS	82.50	85.00	80.00
Ali et al. [3]	LDA–NN–GA	95.00	95.00	95.00
Proposed method	MFCC-LDA-SVM	97.5% (LOSO CV)	100.0%	97.5%