在非线性空间中采用新的最大散度差鉴别准则,提出了一种新的核最大散度差鉴别分析方法。该方法不仅有效地抽取了人脸图像的非线性鉴别特征,而且从根本上避免了以往核Fisher鉴别分析中训练样本总数较多时,通常存在的核散布矩阵奇异的问题,计算复杂度大大降低,识别速度有了明显的提高。在ORL人脸数据库上的实验结果验证了该算法的有效性。
A novel kernel maximum scatter difference discriminant analysis(KMSDA) based on scatter difference criterion is developed for extraction of nolinear feature.The proposed method not only extract nolinear feature for faces but also essentially avoid the difficulties caused by the singularity of kernel within-class scatter matrix in traditional kernel Fisher discriminant analysis(KFDA).In addition,much computational time is saved due to its lower computational complexity.The experimental results on ORL face database verify the effectiveness of the proposed method.