On-line Handwritten Isolated Symbol Recognition using Bidirectional Long Short-term Memory (BLSTM) Networks
Abstract
In this article, we studied BLSTM networks to
recognize on-line handwritten isolated symbols (including both
digits and mathematical symbols). BLSTM networks are suitable
for sequence classification tasks as they can access the contextual
information from two directions. We tested the performance of
the BLSTM model for two different problems using well known
datasets and compared our results to the state of the art on the
same datasets.