Digit recognition using decimal coding and artificial neural network

• DOI: 10.48129/kjs.v49i1.9556
• Publication Date: 2021-12-06T18:30:28Z
• Authors: Toufik Datsi; Khalid Aznag; ; Ahmed El Oirrak
• Keywords: Decimal; Multilayer perceptron; Backpropagation; Digit recognition; Numerical digit

Current artificial neural network image recognition techniques use all the pixels of an as input. In this paper, we present efficient method for handwritten digit that involves extracting characteristics a by coding each row decimal value, i.e., transforming binary representation into value. This is called rows. The set values calculated from initial arranged vector and normalized; these represent inputs to network. approach proposed in work uses multilayer perceptron classification, recognition, prediction digits 0 9. study, dataset 1797 samples were obtained database imported Scikit-learn library. Backpropagation was used learning algorithm train results show achieves better performance than two other schemes terms accuracy execution time.