Low-Cost Smart Glasses for Blind Individuals using Raspberry Pi 2

Abstract

Blindness is the inability of an individual to perceive light. Due to the lack of visual sense, blind individuals use guiding tools and human assistance to replace their visual impairment. The study developed a smart glass that can detect objects and text signages and give an audio output as a guiding tool for blind individuals. In creating the prototype, Raspberry Pi 2 Model B will use as the microprocessor. It will be using a camera module that will be a tool for detection. The algorithms used for object detection and text detection are YOLOv3 and OCR, respectively. In-text detection, OCR helps recognize both handwritten and digitalized texts. MATLAB is the software used for the application of OCR. It is composed of three parts (3): image capturing, extraction of text, and conversion of text-to-speech. In object detection, YOLOv3 is the algorithm used in the process. It comprises four (4) parts: data collection, data preparations, model training, and inference. Then the conversion of text-to-speech will take into place. The objects that the prototype can detect are limited to 15 objects only. The prototype can function at both the 150 lux luminance and 107527 luminance in object detection. However, there are discrepancies in the detection of some objects due to distance; the detection cannot detect the specific thing at certain trials. In-text detection, the detection of the text signage has 100% reliability. In addition, text detection used five font styles. In the testing, the font style Calibri has a 30% percentage error (using the word ENTRANCE) and a 20% percentage error (using the phrase EXIT) due to the structure. The processing time of the prototype has an average time of 1.916s at maximal walking and 1.673s at a slow pace walking.