基于多模态视觉特征的自然人机交互  

Multimodal Visual Features Based Natural Human-Computer Interaction

作　　者：管业鹏[1,2] 

机构地区：[1]上海大学通信与信息工程学院,上海200072 [2]新型显示技术及应用集成教育部重点实验室,上海200072

出　　处：《电子学报》2013年第11期2223-2229,共7页Acta Electronica Sinica

基　　金：国家自然科学基金(No.11176016;No.60872117);高等学校博士学科点专项科研基金(No.20123108110014)

摘　　要：针对目前人机交互中存在的问题与不足,提出了一种基于多模态视觉特征的人机交互新方法.采用二维Gabor小波提取全局人脸朝向视觉特征,克服人脸显著特征提取困难,以及不同人脸朝向特征区分的难度性;基于人脸几何分布特征,快速定位人眼中心,无需考虑人脸图像分辨率和人眼闭合及其配戴情况;通过对视觉特征的重要性评价,选取分类特性显著的多模态视觉特征进行机器学习与训练,确定用户所指目标,实现非穿戴自然人机交互,用户无需佩戴任何标记,且其活动不受约束,便于充分发挥其日常技能.通过实验对比,验证了文中所提方法有效、可行,可应用于实时非穿戴自然人机交互中.A novel human-computer interaction（HCI）is developed based on multimodal visual features aiming at some limits at present .Two-dimensional Gabor wavelet is adopted to extract some visual features of global face orientation ,which overcomes some difficulties including extraction of some facial distinct features ,discrimination among some different facial orientations .An ef-ficient and fast approach to locating center of eyes is proposed based on facial geometric distributions without considering facial res-olution ,eyes closing or opening and user′s wearing .Some prominent multimodal visual features for classification are selected to ma-chine learning and training to determine the pointing target after evaluating performance of some extracted visual features .Non-wearable and natural HCI modal can be realized in which user can move freely without wearing any markers when he points at some targets .Their daily skills can be exerted fully during HCI .Experiment results indicate that the developed approach is efficient and can be used to natural non-wearable HCI .

关 键 词：人机交互 多模态视觉特征 非穿戴 机器学习 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]