Recognition of Meitei Mayek Using Statistical Texture and Histogram Features

Abstract

Recognition of handwritten characters is one of the challenging and interesting problems in recent trends considering the empirical aspect. The research work on Meitei Mayek (Manipuri Script) recognition is still in the infant stage due to its intricate patterns and being a regional language. Moreover, the language has been reinstated recently, and there is no standard database available for research work. Therefore, we attempt to develop a recognition system on our developed dataset of Meitei Mayek using statistical texture and histogram features. A total of 4, 900 samples of 35 letters of Meitei Mayek have been collected in 140 pages from 90 different people of varying age group. Feature extraction technique like Uniform Local Binary Pattern (ULBP) and Projection Histogram (PH) have been considered to evaluate and validate the developed dataset. Finally, recognition is performed using the K-nearest neighbor (KNN) classifier. The combination of ULBP and PH has given the recognition rate of 97.85%.