Tulu Text Recognition with Speech Based Output

The Tulu language, spoken predominantly in the coastal region of Karnataka, India, possesses a rich cultural heritage and a unique script.Despite its historical significance, the Tulu script is underrepresented in the digital realm.This paper presents a Tulu language character recognition system designed to bridge this digital gap.The system processes images containing handwritten Tulu characters and outputs the corresponding text in English.The workflow begins with pre-processing the images to enhance quality and normalize variations.This is followed by image segmentation using region-based techniques to isolate individual characters.A Convolutional Neural Network (CNN) model, trained on a comprehensive dataset, is then employed for the classification and recognition of these characters.The system meticulously recognizes each character, ensuring best recognition efficiency achieved for Tulu characters from collected dataset.The results verified that the proposed methodology outperforms from the present state of art models.To further enhance the system's utility, the recognized text is mapped to pre-recorded audio clips corresponding to each character.These audio clips are then synthesized sequentially to generate a complete auditory output, providing a spoken version of the recognized character.This feature not only enriches the accessibility of the Tulu language in digital formats but also promotes its preservation and dissemination by catering to auditory learners and visually impaired users.