Experimental Research Of Biometrics Recognition Based On Finger Sensing Capacitive Reactance Spectrum Measurements

The currently existed mainstream biometrics recognition(BR)technologies are based on different kinds of image information,but have defects of poor collectability,invasive cellection or poor antifalsification.The author's research group previously proposed a new BR method based on palm bioimpedance spectroscopy measurement,but the results showed that such method was vulnerable to the impacts of contact impedance and had low recognition rate.On this basis,this thesis proposes a new BR method based on finger sensing capacitance reactance spectroscopy(CRS),aiming to avoid the disadvantages of contact impedance,and ultimately to improve the recognition rate.First,this thesis developed a finger-sensitive CRS portable measuring instrument.System design of the instrument used today's popular dual-core ARM + FPGA architecture.FPGA control DAC produce multisine excitation signal to the front-end circuit,and dual-channel ADC device synchronous sampling excitation and response signals and data cached in the FPGA FIFO module,then reading data from the ARM processor(STM32)and stored in a U disk,measuring system has achieved the measurement of 256 points CRS in the range of 4~1024kHz frequency.Secondly,this thesis analyzed the relevant factors affecting the performance of the electrode,designed the ideal measuring electrode.Also,the measurement front-end circuit signal line types and sizes were studied experimentally contrast,choosed the ideal signal line types and sizes.Then on external environmental factors(such as sports,water)impact test repeatability doing experimental research,experimental results showed that the smaller environmental impact.Finally,in the Xi'an University of Technology campus on 500 volunteers were finger-sensitive CRS measurement experiment,we established 500 sample libraries and test libraries,respectively,using SVM and BP neural network algorithm for BR experiments,which the SVM algorithm recognition rate can reach 81%,while BP neural network algorithm recognition rate 85%.The results of this thesis have some preliminary evidence of the feasibility of biometric recognition based on finger-sensitive CRS,and for the development of biometric recognition method based on the non-image acquisition provided a new idea.