A flexible and portable currency note recognizer for the blind

• Main Author: Neo, Shi Lei
• Other Authors: Kong Wai-Kin Adams
• Format: Final Year Project
• Language: English
• Published: 2016
• Subjects: DRNTU::Engineering::Computer science and engineering::Computing methodologies::Image processing and computer vision
• Online Access: http://hdl.handle.net/10356/69157
• Institution: Nanyang Technological University

While most people are blessed with good vision, many people live their life with little or limited vision, and some no vision at all. According to World Health Organization (WHO), the number of people suffering from visually impairment is approximately 285 million. Furthermore, 82% of the visually impaired are 50 years old or even older. With the rapid increase of ageing population in Singapore, the country might soon experience an increase in the visually impaired population. Most of the visually impaired people are able to gain their independence by going through the blind skill technique trainings. However, more often than not, the trainings take years of practices before they can fully master it. On the other hand, with the use of image recognizing technology, some of the training can shorten tremendously. The project aims to improve the lifestyle and independence of the visually impaired people by developing an Android application to recognize different value of Singapore dollar notes. The currency recognition application in this project is named MoneyAid. It is designed to recognize Singapore dollar notes by capturing an image of the note in real time. The application produces a digital voice of the predicted note after each prediction. This is achieved through the implementation of image recognizing algorithm. The algorithm uses the approach of bag of word to simplify the representation of the local feature extracted using Scale-invariant feature transform. The bag of word model is fitted into the Radian Basis function kernel to produce a classifier that can predict the Singapore dollar notes with an accuracy of 97.6884%. Finally, the trained classifier model is integrated into the application, MoneyAid to detect and predict the capture image of Singapore dollar notes.