Handwritten tifinagh character recognition using deep learning architectures

Abstract

Tifinagh-IRCAM is the official alphabet of the Amazigh language which is widely used in North Africa. Being normalized recently in 2001, Tifinagh remains a young alphabet and, accordingly, its handwritten character recognition is a young field of research. In this context, we propose two deep learning approaches for handwritten Tifinagh character recognition: Convolutional Neural Networks (CNNs) and Deep Belief Networks (DBNs). Both networks are trained and tested on the AMHCD handwritten character database. DBNs achieve an accuracy of 95.47% while CNNs outperform existent methods with an accuracy of 98.25%.