Script identification in handwritten and printed documents using convolutional recurrent connection

Identification of the script in multi-script handwritten or printed documents is one of the essential component to recognize the text. The script identification module helps Optical Character Recognition (OCR) to digitize the text present in the multi-script handwritten or printed documents. The similarity of characters between two or more scripts create this task tedious. The factors such as noise and writing style creates identification of the script more tedious. The present research work has proposed a deep learning method having a set of optimized convolutional layers followed by recurrently connected layers to identify the script of any word sample present in the handwritten or printed document. The proposed method has two components to extract deep hierarchical features and identify the temporal features. The experiments have been carried out on MDIW-13 and PHDIndic_11 datasets having handwritten and printed documents. The experimental results from the proposed method has improved the performance over existing methods in this regard. © The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature 2024.