Detection of Touchscreen-Based Urdu Braille Characters Using Machine Learning Techniques

Table 4

Comparative analysis with previous literatures.

Supported languages	Techniques used	Feature extraction/algorithms/others	Accuracy (%)	References
Hindi	SVM	HOG features	94.5	[24]
Chinese	Image segmentation techniques	Image segmentation techniques	79.63	[52]
Chinese	DNN	DNN	90.47	[53]
Simple Braille recognition	MLP	SDAE	86	[34]
	RBF		80
	SoftMax		92
	MLP	Traditional feature extraction	64
	RBF		55
	SoftMax		65
	KNN		63
	Naïve Bayes		53
	Random forest		65
	SVM		69.6
Tamil	Nil	Image segmentation technique	99.2	[28]
Hindi			98.8
Sinhala	SVM	HOG feature extraction method	80	[35]
Arabic	Newly designed Braille letter recognition and transcription scheme for Braille to Arabic text conversion	Image segmentation	99	[21]
Odia	SVM	HOG features	99	[36]
Korean	CNN model	------	99.6	[54]
Bangla	Deep neural network	VGG-16	94.62	[55]
		ResNet-50	93.58
		DenseNet-121	94.08
Grade 1 Urdu Braille	Decision tree	RICA feature extraction method	99.57	Newly proposed coordinate-based model
	KNN		99.50
	SVM		99.73