Permanent link to this recordhttp://hdl.handle.net/10754/662741
MetadataShow full item record
AbstractIn this paper we propose a system that is capable of recognizing raw online handwritten data. The system consists of an advanced type of neural network known as LSTM (Long Short Term Memory) Encoder-decoder combined with a customized attention mechanism layer. The attention mechanism has greatly enhanced the system performance from a low character level accuracy of 53% to an excellent accuracy of 96%. Moreover, the system involves a segmentation algorithm designed to divide the sentences into segments of lines. For the training and testing we employ the IAM On-Line Handwriting database, the source can be found here . The accuracy can be improved even further by integrating our system with a language model to spell check the outputs.
CitationEisa, A., Abdalla, L., & Ahmed, M. (2019). Online Handwriting Recognition Using Encoder-Decoder Model. 2019 International Conference on Computer, Control, Electrical, and Electronics Engineering (ICCCEEE). doi:10.1109/iccceee46830.2019.9071037
Conference/Event name2019 International Conference on Computer, Control, Electrical, and Electronics Engineering (ICCCEEE)