基于ICA的人机交互手势的识别  

Hand Gesture Recognition in Human-Machine Interaction Based on Independent Component Analysis

作　　者：戴志强[1] 张延炘[1] 苏晓星[1] 

机构地区：[1]南开大学现代光学研究所教育部光电信息技术科学重点实验室

出　　处：《光电子．激光》2003年第8期866-868,共3页Journal of Optoelectronics·Laser

基　　金：天津市光电子联合研究中心资助项目(003101211)

摘　　要：通过对独立分量分析(ICA)理论的研究以及对人机交互手势特征的分析,提出了一种基于ICA的静态手势特征提取与识别的方法。用ICA方法分别提取各类静态手势图像的独立分量特征(ICF),构成手势图像的独立基函数空间,对手势图像采用独立分量的最小二乘意义下的表示,结合系统的判别阈值实现对手势的分类识别。系统采用4类手势,共计80幅图像,对方法的有效性进行了检测。实验结果表明,这一方法不仅可行,而且能够获得满意的识别结果。A novel method was proposed to extract and recognize features of static hand gestures based on the independent component analysis(ICA) theory and the analysis of hand gesture features in humanmachine interaction.Independent component features(ICF) corresponding to each kind of e individual static hand gestures were extracted to construct independent basic function space for hand gesture images.Hand gestures represented by the ICF in the leastsquare meaning were classified and recognized under the consideration of discriminative threshold value of the system.Four kinds of hand gestures(80 images altogether) were used to evaluate the effectiveness of the method.The experimental results have shown that the method is feasible.

关 键 词：ICA 独立分量分析 人机交互 手势识别 特征提取 最小二乘法 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]