Handwriting Recognition for Predicting Gender and Handedness Using Deep Learning

Abstract

This research focuses on predicting the gender and handedness and also combined prediction through Handwriting Recognition. In this paper we have analyzed various Deep learning approaches using different CNN models. Our proposed approach uses two data sets IAM English dataset and Real-Time dataset that was collected by us for recognizing the handwriting pattern to perform the defined task. Using different Deep learning approaches, experiments were conducted to analyze the performance on the given two datasets for predicting. Gender and handedness. Experiments showed Google Net using CNN is the most efficient for the prediction and recognition with the handwriting samples. For combined prediction the accuracy rate is around 94% with Real-Time dataset and 95% with the IAM dataset.