关键词: 特征提取, 线性判别分析, 类间离散度标量

Abstract: Linear Discriminant Analysis (LDA) is an effective feature extraction method, but there exist at least two critical drawbacks in it: small sample size problem and rank limitation. In order to solve the above problems, this paper presents an improved LDA algorithm (ILDA), which introduces between-class scatter scalar and within-class scatter scalar and extract features through computing the weight of each dimension in sample space. Numerical experiments on ORL facial database and man-made datasets show ILDA achieves good performance in feature extraction.

Key words: feature extraction, Linear Discriminant Analysis (LDA), between-class scatter scalar