A Novel Recurrent and Convolutional Neural Network Technique for Generating Handwriting from Voice

Abstract

This paper presents a way for generating online handwriting using voice. To build this tool, two broad steps are required: Voice Recognition using Google Speech-to-text API and Handwritten Recognition using a combination of Recurrent and Convolutional neural networks (RCNN). The model is evaluated on IAM and Electronic Fonts datasets that contains handwritten images. This research work has reported the result of training data based on Connectionist Temporal Classification (CTC) loss. CTC also has a function named decoder to predict vector data generated by RCNN into understandable text.