Supervised machine learning in intelligent character recognition of handwritten and printed nameplate
Renuka Kajale, Soubhik Das, Paritosh Medhekar · 2017
We are in the midst of a revolution, digital revolution which is primarily responsible for the advancements in various computing fields. There is a greater need for smart devices, devices that do not just compute faster but also `smarter'. As soon as we encounter the word `smart', we think of intelligence. Intelligence has certainly shaped our modern life in the last decade. One such form of intelligence is seen in character recognition devices. Normal optical character recognition methods seem to be working fine upto a good extent, but challenges arise when the input set starts becoming dynamic. In that case, these techniques fall short when the text is handwritten or sometimes when the there is a variation in styles, fonts, etc. In addition, there is a crucial need to produce accurate results irrespective of what the given dataset is. Supervised machine learning when applied here works by learning attributes and classifying labels (target names). This method has consistently shown accurate results in the past as it helps in training the machine well. This is the driving force behind our selection of this specific algorithm. Even when the dataset becomes extensively large due to multiple language scripts, this method works fine. In this paper, we propose a method for intelligent character recognition using classifiers and transferring the data to excel sheet. The results show such methods can produce accuracies close to 95% for alpha numerals and special characters.