A pipelined approach for FPGA implementation of multi modal biometric pattern recognition using prototype based supervised neural network

A Biometric system is essentially a pattern recognition system that makes use of biometric traits to recognize individuals. Authentication systems built on only one biometric modality may not fulfill the requirements of demanding applications in terms of properties such as performance, acceptability and distinctiveness. Most of the unimodal biometrics systems have problems such as noise in collected data, intra-class variations, inter-class variations, non universality etc. Some of these limitations can be overcome by multiple source of information for establishing identity; such systems are known as multimodal biometric systems. In this paper a multi modal biometric system of iris and palm print based on Wavelet Packet Analysis and a neural classifier implemented in FPGA is described. The visible texture of a person's iris and palm print is encoded into a compact sequence of 2-D wavelet packet coefficients, which generate a “feature vector code”. The combined pattern vector of palm print features and iris features are formed using fusion at feature level and applied to the pattern classifier. The LVQ neural network which is used as pattern classifier produces better recognition rate of 96.22%. The neural network is implemented for input data of 18 dimensions and 12 input classes and 8 output classes, using virtex-4 xc4vlx15 device.