SMART WALKING STICK FOR THE BLIND WITH TEXT IMAGE RECOGNITION

Abstract

Reduced privacy, mobility and productivity are some of the major concerns for the visually impaired. Currently, most of the widely available assistive technologies for the blind have limited functionality and can be ineffective for everyday use. In this paper, the design of an electronic-based walking stick for the blind with image recognition capability is discussed. The device uses echolocation to guide the user in order to avoid collision with obstacles. It uses a Raspberry webcam to acquire the image of the surroundings. These acquired images are digitally processed in real-time using a combination of basic morphological operations. Text image detection and recognition is performed using the EAST convolutional neural network (CNN) model and the Tesseract OCR model respectively [9, 10]. Finally, a text-to-speech operation is performed on the detected characters and read out to the user through an interface.