论文信息 - A user-adaptive deep machine learning method for handwritten digit recognition

A user-adaptive deep machine learning method for handwritten digit recognition

The HWR (handwritten recognition) problem gains more attention with the development of machine learning. In this work, a user-adaptive HWR method is purposed for the application when only handwritten digits and few limited characters need to be recognized. Five types of CNN (Convolutional Neural Network) classifier are used in three steps: digits recognition, string-type classifier and string recognition. Experiment results show that the purposed method is capable of HWR for digits and few limited characters.

[1] Ching Y. Suen,et al. A genetic framework using contextual knowledge for segmentation and recognition of handwritten numeral strings , 2007, Pattern Recognit..

[2] Geoffrey E. Hinton,et al. Reducing the Dimensionality of Data with Neural Networks , 2006, Science.

[3] Luiz Eduardo Soares de Oliveira,et al. Handwritten digit segmentation: Is it still necessary? , 2018, Pattern Recognit..

[4] Sebastiano Impedovo,et al. A novel prototype generation technique for handwriting digit recognition , 2014, Pattern Recognit..

[5] Youcef Chibani,et al. Segmentation and recognition system for unknown-length handwritten digit strings , 2017, Pattern Analysis and Applications.

[6] Il-Seok Oh,et al. A segmentation-free recognition of two touching numerals using neural network , 1999, Proceedings of the Fifth International Conference on Document Analysis and Recognition. ICDAR '99 (Cat. No.PR00318).

[7] Misha Tsodyks,et al. Neural networks and perceptual learning , 2004, Nature.