A novel offline handwritten text recognition technique to convert ruled-line text into digital text through deep neural networks

Offline Handwritten Text Recognition (HTR) has been an active area of research due to its wide range of applications and challenges. Recently, many offline HTR techniques have been developed. However, most of the existing techniques were trained on the datasets that contain the handwritten text images on plain pages. Nevertheless, in real life, the handwritten text can be written on either plain pages or ruled-line pages. Therefore, the approaches proposed in recent literature are unable to convert the digital text accurately written on ruled-line pages. Hence, this study proposes a tailor-made end-to-end offline HTR technique that can accurately convert the offline handwritten text written on ruled-line pages into digital text with the help of computer vision and deep neural network-based techniques. To Evaluate the performance of our proposed technique, we developed a relatively complex dataset that contains the hand-written text images on the ruled-line pages. Our experimental results show that our proposed technique is capable of converting the hand-written text on ruled-line pages into digital text with an overall accuracy of 76.7%. Moreover, the experimental results show that our proposed technique obtained 20% more accurate results compared to baseline techniques. We believe that our proposed technique will contribute positively in the body of knowledge in the field of offline HTR. Moreover, the modular design of our proposed technique allows tailored modifications with respect to data while eliminating the need to retrain the neural network-based models.