Real-Time Text Tracking for Text-to-Speech Translation Camera for the Blind

Abstract

Some mobile devices have been developed for helping the visually-impaired people to obtain useful information from text on documents, goods, and signboards. However, it is still difficult for the blind to find or even notice the text in the environment and to capture the images suitable for character recognition and speech synthesis. We developed a prototype of reading assistant device with a scene text locator that shows the text location by sound signals. To improve the device further, this paper proposes a real-time text tracking method that enables character recognition on-the-fly and also helps the users to obtain the text information easily and efficiently with less searching efforts. The presented method is based on our former one, and provides a new feature that it is tolerant of temporary occlusion or out-of-view of text regions.