【摘要】 针对现有核 Fisher鉴别分析方法的弱点 ,提出了一种基于图像矩阵的非线性不相关鉴别特征抽取技术。该方法的基本思路是 :首先 ,通过经验核映射将原始输入空间 Rn 映射到某特征空间 RN,然后将特征空间 RN 上的训练样本向量变换为一个 p×k( N=p×k)的图像矩阵 ,最后基于该图像矩阵直接构造该空间上的散布矩阵。在Concordia大学的 CENPARMI手写体数字数据库上的试验结果验证了本文方法的有效性。更多还原

【Abstract】 Considering the weakness of existing kernel Fisher discriminant analysis (KFDA) method,a nonlinear and uncorrelated discriminant feature extraction technique based on image matrix (I-UKFDA) is proposed. Firstly,the original input space R n is mapped into a feature space R N via an empirical kernel map. Then,each of the mapped training sample vectors in the feature space is transformed into a p×k image matrix with N=p×k . Based on the image matrices,the scatter matrices of the feature space are constructed. Finally,experimental results in the CENPARMI handwritten digital database of Concordia University indicate that the method is effective than KFDA.更多还原