Abstract
The role of a good feature extractor is to represent an object using numerical measurements. A good feature extractor should generate the features, which must support the classifier to classify similar objects into one category and the distinct objects into separate category. In this paper, we present a method based on HOG (Histogram of Oriented Gradients) for the recognition of handwritten kannada numerals. HOG descriptors are proved to be invariant to geometric transformation and hence they are one among the best descriptors for character recognition. We have used Multi-class Support Vector Machines (SVM) for the classification. The proposed algorithm is experimented on 4,000 images of isolated handwritten Kannada numerals and an average accuracy of 95% is achieved.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have