Word Recognition Techniques for Kannada Handwritten Documents
Tulika Sureka, K S N Swetha, Isha Arora, H. R. Mamatha · 2019
Word Spotting has been an evolving field of research in the past few decades. Starting off with the word spotting in printed text the evolution has now dawned upon the spotting of words in handwritten documents. Recognition of handwritten words is a challenging problem because of varied styles of characters and similarity of their shape. Word Spotting makes document indexing easier and furthers helps in interpreting the documents. In this work we compare and analyze various method by which word recognition can be done in Kannada handwritten documents. When a user gives a query in the form of an image [1], the word is recognized displaying its corresponding unique label. Various feature extraction techniques along with Deep Learning techniques have been tried out. We find that CNN with a spatial layer transformation outperforms the other techniques for Kannada Handwritten words.