【摘要】 采用独立成份分析(ICA)方法对单眼宽约为40 pixel的低分辨率人眼图像进行预处理,并把得到的一组组合系数输入支持向量机(SVM)判断图像中人眼的注视与否。对于单用户的注视探测,一维组合系数输入实现了100%的正确识别率;对于多用户的注视探测,二维组合系数输入实现了100%的正确率识别率。更多还原

【Abstract】 Different from the costly gazing direction measurement and gazing tracking systems,the gazing detection system always uses low-resolution images to realize human-machine interaction.In this paper,the independent component analysis(ICA) is applied to preprocess the low-resolution(40-pixel wide of a single eye) eye images.After the processing,each eye image,which is represented by a low-dimensional vector,will serve as the input of Support Vector Machine(SVM) to give the judgment of gazing or not.The 1-dimensional and 2-dimensional vector input have achieved 100% classification rate to single-user and multi-user,respectively.更多还原