Text string detection for the first grade visually impaired pupils “reading mandarin textbooks”

Abstract

A text string detection method is proposed to help visually impaired first grade pupils read mandarin textbooks. To detect the actual text in view, and recognizing that a textbook is normally held open with both hands, the proposed method sequentially detects seeing a textbook, and then detects skin colors, index fingers and thumbs, text string region, and text strings. Experimental results show that the proposed method performs better than conventional two frame differencing methods. Furthermore, the proposed method reduces time complexity and has 98.5% text string total detection rate for the experimental videos.