Combination of local and global vision modelling for Arabic handwritten words recognition

We propose in this paper a recognition system of Arabic handwritten words issued from literal amounts of Arabic checks. This system is based on the idea of the PERCEPTRO system developed by M. Côté for Latin word recognition. It is a specific NN, named Transparent Neural Network (TNN), combining a global and a local vision modelling (GVM-LVM) of the word. In the forward propagation movement, the former (GVM) proposes a list of structural features characterising the presence of some letters in the word. GVM proposes a list of possible letters and words containing these characteristics. Then, in the back-propagation movement, these letters are confirmed or not according to their proximity with corresponding printed letters. The correspondence between the letter shapes and the corresponding printed letters is performed by LVM using the correspondence of their Fourier descriptors, playing the role of a letter shape normalizer.