Abstract: Linear Discriminant Analysis (LDA)is one of methods in pattern recognition, and is widely used in many fields such as pattern recognition and data analysis. LDA is to find an effective classification direction. While the sample dimention is much larger than its quantity, it is hard for LDA to deal with this problem. In order to effectively solve small sample size problem in LDA，this paper presented a modified LDA algorithm MLDA. This new algorithm turns within-class scatter matrix into scalarization in order to avoid computing the inverse of within-class scatter matrix. A series of experiments verify MLDA solves the small sample size problem to some extend.

Key words: Feature extraction,Linear Discriminant Analysis(LDA),Small sample size problem,Between-class scatter matrix,Within-class scatter matrix,Scalarization