Retrieval performance analysis of multibiometric database using optimized multidimensional spectral hashing based indexing

Abstract

Demand for efficient retrieval of data from biometric databases is increasing due to its widespread authentication applications that range from e-passport to attendance system. Major research contribution in this area is by using tree based indexing methods and data independent random hashing methods. In this work, the data dependent hashing technique using optimized multidimensional spectral hashing that uses hash table lookup is employed. Palmprint, Iris and Face biometric features are generated using GIST, optimized features are fused based on bio-inspired cuckoo search algorithm and then converted into binary hash code. The compact binary codes representing the fused features form the multibiometric database on which the retrieval performance is analyzed. Simulation results obtained indicate that the hit rate and penetration rate have improved considerably for the desired recognition accuracy.