Generation of Handwriting Applying RNN with Mixture Density Network

Abstract

Handwriting recognition is one of the fields where researchers try to apply different machine learning techniques to identify the words and recreate a meaningful sentences. Recognition has two ways to get an input: online and offline. Online handwriting involves the conversion of text along with continuous input as user writes the text. Offline handwriting involves processing handwriting images without any other data as input. This paper has been worked on using a set of online handwriting data to recognize a variety of features using long short-term memory recurrent neural networks (LSTM RNNs). The trained neural network along with mixture density networks (MDNs) is used to generate offline handwriting samples from text.