Optical Braille recognition with Haar wavelet features and Support-Vector Machine

Abstract

This paper proposes a Braille character recognition system, based on Haar feature extraction and Support-Vector Machine (SVM) classification. Braille documents are first scanned into full-color image. The images are then passed through a preprocessor which converts the images into grayscale images, and performs geometric correction. Then a sliding window is applied to the image to crop out sub images. For each sub image, Haar feature vector is calculated and then sent to SVM to decide whether the sub image contains a Braille dot; this translates the original grayscale image into a binary image. Then a simple searching algorithm is applied to the binary image to translate Braille characters into English letters. This method is simple, convenient, and easy to operate, also able to extract dots online in real time. The experiments show that the method is effective and accurate for Braille extraction.