Palmprint Recognition Based on Histograms of Sparse Codes

In this paper, we propose a palmprint recognition scheme using histograms of sparse codes (HSC) as a new feature for palmprint image. In the feature extraction stage, the HSC feature is obtained by computing sparse codes for a given dictionary from a palmprint image, which results in a feature image. In the feature encoding stage, a hash table is designed from the feature image using the binary hashing technique. Finally, the hash table is matched with the templates of hash tables for the purpose of identifying an individual. Extensive experiments are performed on three publicly-available palmprint databases. Experimental results show that the performance of the palmprint recognition system using the proposed scheme is superior to that of other schemes in terms of equal error rate (EER), genuine acceptance rate (GAR) at 11% false acceptance rate (FAR), and processing time.