Lower Order Krawtchouk Moment-Based Feature-Set for Hand Gesture Recognition

Table 2

Recognition accuracies at Jochen-Triesch images taken at different values of and .

Values of and	Total FV	-NN using ED	-NN using MD	SVM			ELM			MLP
				Poly	RBF	PUK	Poly	RBF	Linear
Raw feature-set for Jochen-Triesch
	9	60%	61.1%	58.4%	59.2%	62.2%	65.2%	62.3%	66.1%	63.3%
	18	65.6%	66.6%	67%	69.3%	70.7%	73.3%	72.1%	79.6%	68.3%
	27	67.9%	67.2%	76%	77.1%	79.1%	80.2%	79.9%	80.4%	78.4%
	36	75.7%	78.9%	78.4%	79%	80.9%	82.7%	81.7%	82.4%	81.1%
	45	80%	81.1%	82.2%	82.7%	82.9%	83.5%	82.9%	83%	82.5%
Normalized feature-set for Jochen-Triesch
	9	60%	61.5%	58.5%	61.4%	63.7%	69.2%	66.6%	69.2%	65%
	18	65.8%	66.9%	67.2%	72%	72%	75%	74%	74.2%	69%
	27	67.9%	67.9%	76.7%	79.1%	80.6%	82.6%	81.2%	82.1%	79.6%
	36	75.9%	78.9%	79%	80.3%	83.2%	83.9%	82.8%	83.3%	82.5%
	45	80.1%	81.8%	83.9%	83.9%	84.5%	84.9%	84%	84.5%	83.5%