Handwritten digit recognition using combined feature extraction technique and neural network
Ankita Mishra, Dayashankar Singh
COMPUTER MODELLING & NEW TECHNOLOGIES 2017 21(2) 80-88
Madan Mohan Malaviya University of Technology, Gorakhpur, India
Handwritten digit recognition is established and emerging problem in pattern recognition and computer vision. A very few volume of work related to research has been done in this field till now. Handwritten digit recognition is very useful in cheque processing in bank, form processing systems and many more. In this paper, a robust and novel technique has been introduced for handwritten digit recognition which is tested on well-established MNIST dataset. Histogram of oriented gradient technique and wavelet transform technique is used for feature extraction. Radial basis function neural network and back-propagation neural network have been used as classifier. Experimental analysis has been carried out and result shows that RBF yields good recognition accuracy as compared to back-propagation neural network.