| Fingerprint has important state in identity verification for its exclusivity and invariability. Along with the emergence of the low cost fingerprint sensor and the realization of the highly reliable algorithm, the automated fingerprint identification technique is applied more and more in our life and work. On the other hand, because of the fast development of the digital signal processor (DSP), the DSP already can satisfy the request of abounding computing in image processing.This paper first discusses the relevant concept and the traditional algorithm of fingerprint verification, then put forward a new fingerprint verification algorithm based on wavelet transformation. This algorithm is on the base of the character that the wavelet subband coefficient approximately obedient to the generalized Gaussian density. So two parameters can represent the wavelet subband and can be used to match fingerprints as fingerprint's character. The algorithm's advangtage is simplity of pre-processing algorithm, fast processing speed;small data memory needed by the algorithm (need to save 6 data, 24 byte. Be close to the capacity of 'ID+password'), which suit real application. The paper carries on a great deal of simulation in MATLAB, and compares the performance of this algorithm with the traditional algorithm. The result of simulation show that this algorithm has simple steps, fast speed and high identification rate.Fingerprint verification algorithm in this paper is on the hardware platform of TMS320VC5510 DSK and FPC1010 fingerprint sensor (for capturing fingerprint), and is debugged in DSP Code Composer Studio (CCS). This paper introduce the steps of the realization on DSP and in the process of realization, the problems considered in programming, method to debug and optimization technique.The result show that the algorithm needs small memory capacity, and has fast speed, high identification rate and credibility. It provides a new way to develop fingerprint verification technology. |
PDF Full Text Request