CNN for Handwritten Arabic Digits Recognition Based on LeNet-5

摘要

This paper presents a new unsupervised learning approach with stackedautoencoder (SAE) for Arabic handwritten digits categorization. Recently,Arabic handwritten digits recognition has been an important area due to itsapplications in several fields. This work is focusing on the recognition partof handwritten Arabic digits recognition that face several challenges,including the unlimited variation in human handwriting and the large publicdatabases. Arabic digits contains ten numbers that were descended from theIndian digits system. Stacked autoencoder (SAE) tested and trained the MADBasedatabase (Arabic handwritten digits images) that contain 10000 testing imagesand 60000 training images. We show that the use of SAE leads to significantimprovements across different machine-learning classification algorithms. SAEis giving an average accuracy of 98.5%.