Table 5.

Comparison results of proposed method with previous works.

Feature	Classifier	Dataset	Result	Reference
Wavelet transform	K-means clustering	Bonn	Acc. 96.67%	[12]
Wavelet scalogram	AlexNet	Bonn	Acc. 100%	[13]
Wavelet entropy	SVM	Bonn	Acc. 100%	[8]
Sample entropy, distributed entropy	GA-SVM	Bonn	AUC.96.67%	[4]
Optimum allocation technique	Logistic model tree	Bonn	Acc. 95.33%	[14]
Wavelet transform	Random forest	Bonn	Acc. 95.00%	[16]
Fuzzy distribution entropy and wavelet packet decomposition	KNN	Bonn	Acc. 98.33%	[18]
Discrete wavelet transform and wavelet energy distribution	ANN	Bonn	Acc. 95%	[19]
Hurst exponent	KNN	Bonn	Acc. 100%	[20]
Symlet wavelet and grid search optimizer feature	Gradient boosting machine	Bonn	Acc. 96.1%	[2]
Self-organized map	RBF neural network	Bonn	Acc. 97.47%	[21]
Spectral, spatial and temporal feature	3D CNN	CHB-MIT	Acc. 99.4%	[22]
Short time Fourier transform	CNN	Bern Barcelona	Acc. 91.8%	[23]
Fourier-based SST	CNN	CHB-MIT	Acc. 99.63%	[24]
Multi view feature learning	Convolutional deep learning	CHB-MIT	Acc. 94.37%	[25]
Waveform level classification by CNN	SVM	Clinical dataset	Acc. 83.86%	[26]
Time domain and frequency domain feature	Bayesian Net	ARMOR project dataset	Acc. 95%	[15]
Teager energy	Supervised backpropagation neural network	(Institute of Neuroscience, India)	Sensitivity 96.66%	[17]
Spectrogram	CNN	TUH	Acc. 88.3%	[27]
Automated identification without feature extraction	One dimensional deep convolutional neural network	TUH	Acc. 79.34%	[1]
Automatic feature extraction	CNN	TUH	Acc. 100%	This study