基于注意模型的视觉替代方法  

A Vision Substitution Method Based on Visual Attention Models

作　　者：田亚男[1] 雷红玮[1] 王旭[1] 

机构地区：[1]东北大学信息科学与工程学院,辽宁沈阳110819

出　　处：《电子学报》2014年第5期890-895,共6页Acta Electronica Sinica

基　　金：国家自然科学基金(No.50477015);教育部基本科研业务费项目青年教师科研启动基金(No.N110304002)

摘　　要：现有的视觉替代方法均是在特定环境中,通过目标识别获得映射特征,缺少普遍的适用性.针对这一问题,提出了一种基于注意模型的视觉替代方法.利用人眼的视觉特性,提取图像的感兴趣区域,并根据听觉显示的基本原理,提出了将感兴趣区域的位置、尺寸和颜色映射为音符响度、音长和音调的PSC映射方法.实验结果表明:将图像中引起视觉注意的区域映射为电子音符,符合人类视觉认知过程,有利于盲人获得外部环境的重要信息,降低了盲人训练和学习的难度,并且听感良好,不会造成听觉疲劳.The existing vision substitution methods are only used in special environments because their mapping features are obtained by object recognition. Aiming at this problem, a vision substitution method based on visual attention models is proposed. Firstly, regions of interest are extracted according to human visual properties. And on the basis of the auditory display theory, PSC mapping method is presented that the position,size and color of regions of interest are respectively mapped into note loudness,duration and pitch. Results show that this method corresponds with the cognitive processes of human vision. It is good at helping the blind to obtain the important information and reducing the difficulty of tr^ing and learning. Moreover, the mapped electronic music is pleasing to ear and enjoyable.

关 键 词：视觉替代 听觉显示 视觉注意 感兴趣区域 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]