Evidence-Based Complementary and Alternative Medicine

Research Article

Tongue Images Classification Based on Constrained High Dispersal Network

Table 5

Comparison of the proposed method with other feature extracting approaches.


		LDA	KNN	CART	GBDT	RF	LIBSVM	LIBLEAR SVM

HOG [20]	Sensitivity	%	%	%	%		%	%
HOG [20]	Specificity	%	%	%	%	%	%	%
LBP	Sensitivity	%	%	%	%	%	%	%
LBP	Specificity	%	%	%	%	%	%	%
SIFT	Sensitivity	%	%	%	%	%	%	%
SIFT	Specificity	%	%	%	%	%	%	%
HOG + LBP	Sensitivity	%	%	%	%	%	%	%
HOG + LBP	Specificity	%	%		%	%	%	%
HOG + SIFT	Sensitivity	%	%	%	%	%	%	%
HOG + SIFT	Specificity	%	%	%	%	%	%	%
LBP + SIFT	Sensitivity	%	%	%	%	%	%	%
LBP + SIFT	Specificity	%		%	%	%	%	%
HOG + LBP + SIFT	Sensitivity	%	%	%	%	%	%
HOG + LBP + SIFT	Specificity	%	%	%	%	%	%	%
Doublets [21]	Sensitivity	%
Doublets [21]	Specificity	%	%
Doublets + HOG [22]	Sensitivity		%	%	%	%	%	%
Doublets + HOG [22]	Specificity		%	%	%	%	%	%
PCANet [23]	Sensitivity	%	%	%	%	%	%	%
PCANet [23]	Specificity					%	%	%
Our method	Sensitivity	%	%	%	%	%	%	%
Our method	Specificity	%	%	%	%	%	%	%

Note. The sum rule of feature combination is a cascade operation. Given two types of features and obtained by feature extraction methods and , respectively, then + is equal to . For example, HOG + LBP means, for each sample, we append LBP features just after HOG features. Besides, “our method” is short for the proposed CHDNet feature extraction method.