抄録
We propose a system which helps the visually impaired people to understand the character information in environment using a note PC and a CCD camera. The system captures a still image by the camera connected to the note PC, and extracts character regions from the image by using the general features of a character and a character sequence and the color information of a signboard region on which the character is drawn. Finally, OCR(Optical Character Recognition) analyzes the extracted character regions and outputs recognized characters by synthesized voice. We conducted an experiment by using 171 images captured in the real environment and obtained satisfactory results: among 417 character lines, 56.1% of lines are completely extracted, and 85.4% are done including lines in which more than half characters are extracted.
