International Journal on Advanced Science, Engineering and Information Technology, Vol. 12 (2022) No. 4, pages: 1449-1456, DOI:10.18517/ijaseit.12.4.14445

Implementation of Gabor Wavelet and Support Vector Machine for Braille Recognition

Rifki Indra Perwira, Mangaras Yanu Florestiyanto, Indah Reforsiana Nurjanah, - Heriyanto, Dessyanto Boedi Prasetyo


Visually impaired people use braille letters to write and read, which not all people with normal vision can read. It causes parents of children with impaired vision to have difficulties in assisting them in learning at home. At the same time, the involvement of parents in children's learning assistance is needed to monitor their learning progress. In this research, braille letters are identified through images taken using a scanner as a tool to input the images. Then, the Canny edge detection method is used obtain all the edges of each braille dot. Feature extraction is applied to obtain all characteristics of each letter, and the method used is the Gabor Wavelet. The features which are utilized include standard deviation, mean, variance, and median with a theta angle of 00, 300, 450, 600, 1200, 1350, 1800 and wavelengths of 3, 6, 13, 28, and 58. These features are combined and used as test data and training data for the Support Vector Machine (SVM) classification stage and produce letters and words in alphabetic letter forms. Braille letters that can be identified in this research are small letters, capital letters, punctuation marks, and numbers. Tests are carried out using a multi-class confusion matrix scenario to determine the level of accuracy, precision, and recall. Based on the results of the tests conducted using 758 braille letters, the accuracy value is 98.15%; the precision value is 97.66%; and the recall value is 98.28%.


Braille; image processing; canny edge detection; gabor wavelet; support vector machine.

Viewed: 1471 times (since abstract online)

cite this paper     download