期刊名称：International Journal of Innovative Research in Computer and Communication Engineering
印刷版ISSN：2320-9798
电子版ISSN：2320-9801
出版年度：2015
卷号：3
期号：8
DOI：10.15680/IJIRCCE.2015. 0308120
出版社：S&S Publications
摘要：We recommend a digital camera-founded assistive text studying framework to help blind humans studytextual content material labels and product packaging from hand held objects of their everyday lives. To isolate theitem from cluttered backgrounds or special surrounding objects in the digicam view, we first endorse an effective andpowerful movement centered procedure to outline a neighborhood of curiosity (ROI) within the video via utilisingasking the user to shake the thing. This approach extracts moving object area with the help of a mixture-of-Gaussianssituated old previous subtraction process. Within the extracted ROI, text localization and realization are carried out togather text understanding. To robotically localize the textual content regions from the thing ROI, we recommend anovel textual content localization algorithm by means of studying gradient aspects of stroke orientations anddistributions of area pixels in an Adaboost model. Textual content characters within the localized text areas are thenbinarized and famous by way of off-the-shelf optical persona cognizance (OCR) application. The famous text codes areoutput to blind purchasers in speech. Efficiency of the proposed textual content localization algorithm is quantitativelyevaluated on ICDAR-2003 and ICDAR-2011 potent learning Datasets. Experimental end result expose that ouralgorithm achieves the state-of-the-arts. The proof-of-proposal prototype can be evaluated on a dataset gatheredutilising 10 blind men and women, to investigate the effectiveness of the approach’s hardware. We discover patroninterface issues, and examine robustness of the algorithm in extracting and studying textual content from specialobjects with complicated backgrounds..
关键词：blindness; assistive instruments; text learning; hand held objects; text neighborhood localization; stroke;orientation; distribution of discipline pixels; OCR
