对人脸识别中Fisher线性判别的改进方法

论文对传统的Fisher线性判别作了深入的分析,由于样本类间离散度矩阵不能很好的分开距离较近的样本类,为此该文又重新定义了样本类间离散度矩阵。通过改进其投影方向来减轻边缘类对特征分解的主导作用,从而提高了识别效率。

To Face Recognition in Fisher Linearity Distinction Corrective Method

The paper to the traditional Fisher linearity distinguished has made the thorough analysis,Because between the sample class relative dispersion matrix cannot very good be away from the near sample class separately,Has redefined between the sample class relative dispersion matrix for this this article .Through improves its projection direction to reduce the edge class the leading role which decomposes to the characteristic,thus enhanced the recognition efficiency.