Offline Handwritten Devanagari Vowels Recognition using KNN Classifier

Abstract

The discussion in the paper is regarding to the recognition of handwritten Devanagari vowels by means of a classifier named as K-NN (K-Nearest Neighbour). Before applying classifier, feature extortion is accomplished for extracting the feature points (FP) i.e. also known as division points (DP). In this paper the feature extortion is perform through recursive sub division technique, which is first time implemented on Devanagari vowels. K-NN classifier is functioned for the learning and the testing phases, through which the recognition go ahead to the high performances in terms of recognition rate, pre-processing and classification speed. Authors tested the described approach using the ISI (Indian Statistical Institute), Kolkata"s handwritten Devanagari vowels database containing 9191 samples, which is divided into 1:3 as testing and training samples respectively. In the recognition process using K-NN classifier 88 vowels are total wrongly identified out of 2281vowels. The recognition rate comes out to be 96.14%.

Recognition of handwritten characters using local gradient feature descriptors

“…The recognition performance with this method is 94.12%. In Rakesh Rathi et al (2012), the authors used feature mining algorithms to compute the feature vector and tested the method on the Devanagari vowels database. Here, the KNN algorithm is used as a classification technique, and obtained the accuracy of 96.14%.…”

Thaiwrittennet: Thai Handwritten Script Recognition Using Deep Neural Networks

“…K-nearest neighbor (k-NN) is instance-based supervised learning, based on statistical estimation. Since each Thai-handwritten recognition image is stored in a row of the vector (Rathi, Pandey, Chaturvedi & Jangid, 2012), all rows are kept as a big table (Mookdarsanit & Moorkdarsanit, 2018a) without any independent learning functions (a.k.a. lazy learning).…”

Handwritten Character Recognition on Focused on the Segmentation of Character Prototypes in Small Strips

“…This problem is difficult and may be resolved by jointly segmenting and recognizing the characters [27], [37], such a problem can be formulated as a multi-class classification problem with a large range of small snippets extracted from the different prototypes of characters [25].…”

“…Several approaches have used this method in the resolution of the problem of the handwriting recognition problem [18], [25].…”