Discriminant analysis and similarity measure

作者：

Highlights：

• The connection between the cosine similarity and the Bayes decision rule.

• Discuss some inherent challenges of the cosine similarity.

• A new similarity measure that integrates both angular and distance measure.

• Experiments on the FRGC version 2 database (36,818 images).

摘要

Highlights•The connection between the cosine similarity and the Bayes decision rule.•Discuss some inherent challenges of the cosine similarity.•A new similarity measure that integrates both angular and distance measure.•Experiments on the FRGC version 2 database (36,818 images).

论文关键词：Bayes decision rule,Cosine similarity measure,Discriminant analysis,Face recognition,Face Recognition Grand Challenge (FRGC),New similarity measure,Normalized correlation,Pattern recognition

论文评审过程：Received 4 June 2012, Revised 18 June 2013, Accepted 23 June 2013, Available online 29 June 2013.

论文官网地址：https://doi.org/10.1016/j.patcog.2013.06.023