A Novel Two-Stage Spectrum-Based Approach for Dimensionality Reduction: A Case Study on the Recognition of Handwritten Numerals

Table 4

Recognition rate corresponding to different feature vectors for the English dataset MNIST and Arabic/Farsi dataset Hoda.

	Feature set	Number of features in feature vector	Feature selection method			Accuracy %
			Proposed method	PCA method	RP method
English dataset MNIST	Initial_S	133	—	—	—	91.93
	E-𝑆1	103	*			93.17
	E-S 2	79	*			94.88
	E-𝑆3	103		*		93.82
	E-𝑆4	79		*		90.71
	E-𝑆5	103			*	93.51
	E-𝑆6	79			*	88.39
Arabic/Farsi dataset Hoda	Initial_S	133	—	—	—	90.41
	A/F-𝑆1	94	*			92.60
	A/F-S 2	58	*			95.12
	A/F-𝑆3	94		*		94.04
	A/F-𝑆4	58		*		89.00
	A/F-𝑆5	94			*	91.07
	A/F-𝑆6	58			*	83.66