Fractional-order embedding canonical correlation analysis and its applications to multi-view dimensionality reduction and recognition

作者：

Highlights：

• A new method, namely FECCA, is presented for multi-view dimensionality reduction.

• Fractional-order within-set and between-set scatter matrices are constructed by sample spectrum modeling.

• The extracted features by FECCA have strong discriminant power for recognition.

• Experimental results show FECCA has better recognition rates than the existing joint feature extraction methods.

摘要

•A new method, namely FECCA, is presented for multi-view dimensionality reduction.•Fractional-order within-set and between-set scatter matrices are constructed by sample spectrum modeling.•The extracted features by FECCA have strong discriminant power for recognition.•Experimental results show FECCA has better recognition rates than the existing joint feature extraction methods.

论文关键词：Pattern recognition,Canonical correlation analysis,Feature extraction,Dimensionality reduction,Multi-view learning

论文评审过程：Received 9 December 2012, Revised 21 July 2013, Accepted 16 September 2013, Available online 2 October 2013.

论文官网地址：https://doi.org/10.1016/j.patcog.2013.09.009