Handwritten Character Recognition of Kannada Scripts using Novel Feature Extraction Techniques and BMCNN Classifier
Supreetha Patel Tiptur Parashivamurthy, Sannangi Viswaradhya Rajashekararadhya · International Journal of Electrical and Electronics Engineering · 2023
Handwritten Character Recognition (HCR) is one of the most popular research in recent years. Many HCR systems were developed based on various languages. However, only a few works are based on handwritten Kannada characters. Recognising handwritten Kannada characters is challenging because of the curvy and symmetric nature of the Kannada characters. Although various works were conducted for Kannada HCR, some issues must be solved. Hence, this work proposed BMCNN-based Kannada HCR. In the preprocessing phase, 3M filtering and the CLAHE techniques perform noise reduction and contrast enhancement. Then, the image is resized, angle rotated and mirror-inverted to obtain better accuracy of the input image. Then, the zonal, pattern and gradient features are extracted from the preprocessed image. Next, the significant features are selected by ISSA and then given to the BMCNN classifier to recognise the input Kannada character. To prove the efficiency of the proposed framework, the experimental analysis is conducted in terms of various measures and compared with state-ofart techniques. The results showed that the proposed recognition technique performs better than the existing techniques.