Dynamic Principal Component Analysis with Nonoverlapping Moving Window and Its Applications to Epileptic EEG Classification

Table 3

A comparison of classification accuracy obtained by various methods for epileptic seizure detection problem using EEGs dataset from the University of Bonn.

Papers	Method (feature extraction + classification method)	Problems	Accuracy
Nigam and Graupe [6]	Nonlinear preprocessing filter + diagnostic neural network	A–E	97.2%
Srinivasan et al. [7]	Time-frequency domain features + recurrent neural network	A–E	99.6%
Kannathal et al. [8]	Entropy measures + adaptive neurofuzzy inference system	A–E	92.22%
Polat and Güneş [9]	Fast Fourier transform + decision tree	A–E	98.72%
Subasi [10]	Discrete wavelet transform + mixture of expert model	A–E	95%
Tzallas et al. [11]	Time-frequency analysis + artificial neural network	A–E	100%
Guo et al. [12]	Multiwavelet transform and entropy + MLPNN	A–E	99.85%
This work	DPCA with PCPEM + 1-NN,	A–E	100%
Kim and Rosen [13]	AR model + PCA	B-C-E	96.6%
Tzallas et al. [11]	Time-frequency analysis + artificial neural network	A, B, C, D-E	97.73%
Guo et al. [12]	Multiwavelet transform and entropy + MLPNN	A, B, C, D-E	98.27%
This work	DPCA with PCPEM + 1-NN	A, B, C, D-E	100%