Multispectral Palm-vein Fusion for User Identification

Jaekwon Lee, Jooyoung Kim, Donghyun Kim, Seung Ah Lee, Jong Hwan Sung, Kar-Ann Toh

Abstract: In this paper, we propose a system for user identification based on palm-veins extracted from multi-spectral images of the palm. Essentially, a feature level fusion is firstly conducted by stacking the preprocessed palm images from multiple image spectrums to increase the richness of information. Subsequently, a convolution neural network (CNN), which utilizes the residual learning with a linear bottleneck scheme, is adopted to learn the stacked features. The proposed system has been evaluated on a public multispectral palm database where a promising performance in terms of the identification accuracy has been observed.

External IDs:dblp:conf/elinfocom/LeeKKLST23