结合CSS与傅里叶描述子的手势特征提取  

Gesture Feature Extraction Combining CSS with Fourier Descriptor

作　　者：李丹娇[1] 彭进业[1] 冯晓毅[1] 王珺[1] 

机构地区：[1]西北工业大学电子信息学院,西安710129

出　　处：《计算机工程》2012年第6期178-180,共3页Computer Engineering

基　　金：国家自然科学基金资助项目(60875016,61075014);教育部博士点基金资助项目(20096102110025)

摘　　要：目前常用的基于视觉的静态手势特征提取方法只从单一方面进行描述,缺乏全局信息和局部信息的有效结合。为此,提出一种结合CSS形状描述子与傅里叶描述子的手势特征提取方法。将CSS形状描述子与傅里叶描述子相结合,以此作为一种融合手势局部特征和全局特征的新的静态手势特征。实验结果表明,与传统方法相比,该方法的正确率更高,达到98.3%。Aiming at the problem that the common static gesture features which are used in vision based recognition often only focus on gestures in a single aspect, and lack of effective gesture description for combination of wholeness and local part, this paper proposes an approach which extracts Fourier descriptor as a kind of global feature, and CSS shape descriptor as a kind of local feature, and then combines them into the final feature. Experimental results verify the effectiveness of this new feature on static gesture recognition by its high correct rate 98.3%, compared with traditional approaches.

关 键 词：基于视觉 静态手势特征 CSS形状描述子 傅里叶描述子 局部特征 全局特征 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]