A Multi-Layer Holistic Approach for Cursive Text Recognition

Urdu is a widely spoken and narrated language in several South-Asian countries communities worldwide. It relatively hard to recognize text compared other languages due its cursive writing style. The script belongs non-Latin family like Arabic, Hindi Chinese. written styles, among which ‘Nastaleeq’ the most popular used font A gap still poses challenge for localization/detection recognition of Nastaleeq as it follows modified version Arabic script. This research study presents methodology classify font, regardless position image. proposed solution comprised two-step methodology. In first step, detection performed using Connected Component Analysis (CCA) Long Short-Term Memory Neural Network (LSTM). second hybrid Convolution Recurrent (CNN-RNN) architecture deployed detected text. image containing binarized segmented produce single-line fed CNN-RNN model, recognizes saves file. technique outperforms existing ones by achieving an overall accuracy 97.47%.