CN101206714A - Fingerprint identification method and fingerprint payment system based on DSP - Google Patents

Abstract

The invention discloses a DSP-based fingerprint identification method and a DSP-based fingerprint payment system, wherein, extreme filtering, smoothing filtering and Laplace sharpening are performed on an acquired fingerprint by adoption of a DSP TMS320VC5000; fingerprint image characteristics after pretreatment are extracted through binaryzation of iterated thresholds; fingerprint image characteristic data after binary extraction of the iterated thresholds is bond with user log-on account number data and then stored into a tree structure database on a server; a client DSP processes a field fingerprint, then modulates the field fingerprint through data encryption, is connected with the server through a RS-232/485 interface, demodulates received fingerprint data and then performs search and verification or identification according to the fingerprint characteristic data which is stored in the server database, thereby application aims of determination of fingerprint identity and personal account number payment are reached. The invention has the advantages of quick, flexible and accurate processing speed, strong anti-interference ability, small volume and so on.

Description

Fingerprint identification method and fingerprint payment system based on DSP

Affiliated technical field

The present invention relates to a kind of biometric discrimination method, particularly relate to a kind of fingerprint identification method, also relate to a kind of fingerprint payment system that adopts this method based on DSP.

Technical background

What the transaction in present most bank and shopping market was all adopted is " authority magnetic card+password " mode, has a lot of hidden danger in concrete application and system execution, such as: easily be forged.The easier acquisition of password under the unwitting situation of holder, is illegally used its authority; Easily substituted.Without permission, authorize mutually privately, the holder meets at own authority card other people without authorization and uses or the holder is coerced by the people and oneself authority card is met at other people use; Easily leave behind or lose, make troubles and influence for own and household and colleague's life and work.

At present, the fingerprint identification method that also is useful on personal status's identification is different, but the existing fingerprint identification method problem that all the ubiquity discrimination is low, recognition speed is slow.

Summary of the invention

One object of the present invention is to overcome above-mentioned defective in the prior art: a kind of discrimination height, the fast fingerprint identification method based on DSP of recognition speed are provided.

One object of the present invention is to overcome above-mentioned defective in the prior art: the fingerprint payment system that a kind of processing speed is fast, flexible, antijamming capability is strong is provided.

For achieving the above object, technical scheme provided by the invention is as follows: a kind of fingerprint identification method based on DSP is provided, may further comprise the steps:

A, employing DSP TMS320VC5000 carry out extreme value filtering, smothing filtering, Laplce's sharpening processing to the fingerprint of gathering;

B, extract through pretreated fingerprint image characteristics point by the iteration threshold binaryzation;

C, the fingerprint image characteristics point and the Account Data of user's registration that will extract through the iteration threshold binaryzation are bound and are stored in the database on the server;

D, DSP are connected with server by the RS-232/485 interface, and server is connected with service terminal.

The process that described fingerprint to collection carries out the extreme value Filtering Processing is as follows:

A, to be provided with a pending pixel region be s0, and 8 neighborhood territory pixels are arranged as follows around it:

$\left[\begin{array}{ccc}s1& s2& s3\\ s4& s0& s5\\ s6& s7& s8\end{array}\right]$

Get the average Ai of neighborhood related pixel earlier, i ∈ 1,2 ... 8} is one group of processing unit with four pixels, can improve correlation technique document extreme value filtering algorithm, is expressed as follows:

If b is A0＞max (Ai), i ∈ 1,2 ... 8} then

s1＝s2＝s4＝s0＝max(A1，A2，A4)

s2＝s3＝s5＝s0＝max(A2，A3，A5)

s4＝s6＝s7＝s0＝max(A4，A6，A7)

s5＝s?7＝s8＝s0＝max(A5，A7，A8)

If c is A0＜min (Ai), i ∈ 1,2 ... 8} then

s?1＝s2＝s4＝s0＝min(A1，A2，A4)

s2＝s3＝s5＝s0＝min(A2，A3，A5)

s4＝s6＝s7＝s0＝min(A4，A6，A7)

s5＝s7＝s8＝s0＝min(A5，A7，A8)

If d is min (Ai)＜=Ai＜=max (Ai), i ∈ 1,2 ... 8} then with the output of pixel initial value, does not process.

The process that described fingerprint to collection carries out The disposal of gentle filter is as follows:

Each pixel in the fingerprint and M are carried out convolution, and level and smooth convolution kernel is:

$M=\frac{1}{15}\left[\begin{array}{ccc}1& 2& 1\\ 2& 5& 2\\ 1& 2& 1\end{array}\right]$

The process that described fingerprint to collection carries out Laplce's sharpening processing is as follows:

The sharpening convolution kernel adopts the La Pulashi operator, and is as follows,

$h=\left[\begin{array}{ccc}-1& -1& -1\\ -1& 9& -1\\ -1& -1& -1\end{array}\right]$

By this convolution kernel image pixel is carried out the convolution budget, realize high-pass filtering, thereby make

Laplace operator is used on the fingerprint image, and obtains the fingerprint ridge line after the sharpening.

Described as follows through the process of pretreated fingerprint image characteristics point by the extraction of iteration threshold binaryzation:

Take the method for iteration threshold, in the iteration threshold computing:

A, setting Initial Hurdle T, as make T=127 (gray level), the average gray value of fingerprint image is divided into two groups

R1、R2；

B, calculating two groups average gray value u1, u2;

C and then reset new gray scale threshold values T, new T is defined as: T=(u1+u2)/2;

D, fingerprint image is carried out threshold segmentation according to this T.

Database on the described server is based on the tree construction fingerprint database, is divided into a certain classification according to the fingerprint characteristic fingerprint of naming a person for a particular job, and searches for the affirmation user identity fast according to the classification of fingerprint, in fingerprint base design based on tree construction, and unique fingerprint vector At tree node [i ₁..., i _l] go up according to Gaussian distribution N (0, σ _W ²) produce, wherein l is the degree of depth of node, and i is certain user node, and { a} is separate, for user u for the fingerprint vector collection simultaneously ⁽ⁱ⁾Its index is i=[i ₁, i ₂...., i _l], distribute to u ⁽ⁱ⁾Fingerprint copy X ⁽ⁱ⁾In, the L of a total embedding fingerprint is

Also can carry out the modulating transformation of orthogonal vector to the finger print data sequence in the database: conversion coefficient is divided into set of vectors according to ID number of fingerprint registration, make the vector quantization (COVQ of channel optimization by combination, channel-optimizedVQ) the Euclid vector quantization index i* of acquisition standard, to given input information source vector v, from coded vector y _iAnd size is N _CbThe quantization index i* that obtains of information source code book can be expressed as

$i*=\arg \min (\underset{j=1}{\overset{N_{\mathrm{cb}}}{\mathrm{\Σ}}}{P}_{j/i}\·{\left|\right|v-y_j\left|\right|}^2),$ i∈{0，1，Λ，N _cb-1}

Here, P _J/iReceive the conditional probability of j vector signal when being known index i, hiding finger print data at first carries out information source coding (depending on data character) and produces the symbol sebolic addressing { s of Q direction ₁, s ₂, Λ, s _Q, at the k dimension space, to each vector coefficients v of Q direction _JSymbol of middle embedding obtains to upset vector

, for the constraint condition of satisfied transmission,

The scope of value should be at radius

Circle in, like this, size just can be formed channel codebook for the k of Q dimension upset value; This channel codebook can be adjusted by the embedment strength α of decision transmission performance, promptly Can be expressed as $\widehat{v_j}=v_j=\mathrm{\α}\·C\left(s_i\right),$ I=1,2, Λ, Q, wherein C (s _i) be exactly that size is the channel codebook of Q, before the transmission, the coefficient of upset is changed to host signal by contravariant; The data S that fingerprint is hidden carries out transform domain and decomposes after the pre-service of ranks direction, obtains transmission signals at last

, again will

(x y) is divided into different fingerprint embedded blocks, i user u according to decomposition layer i and direction θ ⁽ⁱ⁾Fingerprint W ⁽ⁱ⁾Produce by formula (7):

$W^{\left(\text{i}\right)}=\sqrt{{\mathrm{\ρ}}_1}{a}^{i_1}+\sqrt{{\mathrm{\ρ}}_2}{a}^{i_1,i_2}+\mathrm{\Λ}+\sqrt{{\mathrm{\ρ}}_L}{a}^{i_1,i_2,..,i_L}---\left(7\right)$

Wherein, { ρ _lBe subjected to the probability decision of collusion attack by the different branches of corresponding fingerprint tree, and 0≤ρ is arranged ₁, Λ, ρ _L≤ 1, ${\mathrm{\Σ}}_{j=1}^L{\mathrm{\ρ}}_j=1$ Relation, adjust between the different user fingerprint that assignment is distributed according to the capacity of each grade fingerprint of embedment strength α control tree and self-adaptation;

For the value of fingerprint vector on the different nodes of tree, therefore, distribute to u ⁽ⁱ⁾The signal form of fingerprint video content be:

$X^{\left(i\right)}=\hat{S}+\mathrm{\α}{W}^{\left(i\right)}---\left(8\right)$

And in the modulation of quadrature fingerprint, host signal S can be divided into the part S of L non-overlapping copies ₁, Λ, S _L, like this at S _lIn the coefficient that can embed satisfy relation:

And satisfy ${\mathrm{\Σ}}_{l=1}^L{N}_l=N---\left(9\right)$

In the fingerprint modulation of this proposition, definition matrix P is one and goes up the triangle square formation, at P ^JIn, in order to realize effective bandwidth,, make p for k＞l _{K, l}=0, for k≤l, make 0＜p _{K, l}≤ 1 to realize robustness, definition E _{K, l}For being embedded into X _l ⁽ⁱ⁾Middle k level fingerprint vector Capacity, then $E_l\stackrel{\mathrm{\Δ}}{=}{\mathrm{\Σ}}_{k=1}^L{E}_{k,l}$ Be W _l ⁽ⁱ⁾All told,

At the 1st grade, p _l, 1=1 is at the 2nd≤l≤L level, known p _{L, l}, seek { p _{K, l}} _K＜lRelation value, satisfying a kind of condition is E _{1, l}: E _{2, l}: Λ: E _{L-1, l}=ρ ₁: ρ ₂: Λ, ρ _L-1So, but for k＜l relation is arranged:

$p_{k,l}=\frac{{\mathrm{\ρ}}_k(1-p_{l,l})}{{\mathrm{\ρ}}_1+\mathrm{\Λ}+{\mathrm{\ρ}}_{l-1}}---\left(10\right)$

With known { p _{L, l}} _{L=1, Λ, L}The time matrix arranged

$P^j={\left[\begin{array}{cccc}1& 1-p_{\mathrm{2,2}}& \mathrm{\&Lambda;}& (1-p_{L,L})\frac{{\mathrm{\&rho;}}_1}{1-{\mathrm{\&rho;}}_L}\\ 0& p_{\mathrm{2,2}}& \mathrm{\&Lambda;}& (1-p_{L,L})\frac{{\mathrm{\&rho;}}_2}{1-{\mathrm{\&rho;}}_L}\\ M& \mathrm{<figure-callout\; id="0"\; label="M\; O\; M"\; filenames="A20071002753100241.png"\; state="\{\{state\}\}">M</figure-callout>}& O& M\\ 0& 0& 0& p_{L,L}\end{array}\right]}_{L\&times;L}---\left(11\right)$

In (11), work as $\underset{l=1}{\overset{L}{\mathrm{\&Sigma;}}}{N}_l=N,0\&le;N_l\&le;N$ The time, seek N ₁, N ₂, Λ, N _LSatisfy condition

P ^J[N ₁?N ₂ΛN _L] ^T＝N[ρ ₁?ρ ₂Λρ _L] ^T (12)

In (12) formula, work as p _{L, L}=ρ _LThe time, just be based on a kind of method that the quadrature fingerprint is modulated, so,, realize that the method for the multistage imbedding problem of fingerprint is according to above definition:

Condition and relation according to (11) formula $N_L=N-\underset{l=1}{\overset{L-1}{\mathrm{\&Sigma;}}}{N}_l,$ Each fingerprint vector for tree 1≤l≤L level

Have:

$a^{i_1,\mathrm{\&Lambda;},i_l}={a_l}^{i_1,\mathrm{\&Lambda;},i_l}\mathrm{\&Psi;}{a_{l+1}}^{i_1,\mathrm{\&Lambda;},i_l}\mathrm{\&Psi;}{a_L}^{i_1,\mathrm{\&Lambda;},i_l}---\left(13\right)$

Here,

According to Gaussian distribution N (0, σ _W ²) and separate.For k 〉=l, length is N _kFingerprint vector

Be embedded into S _kIn.Symbol " Ψ " has been represented the cascade factor.To u ⁽ⁱ⁾The i part fingerprint video content that the user receives is ${X_l}^{({\mathrm{<figure-callout\; id="1"\; label="i"\; filenames="A20071002753100241.png"\; state="\{\{state\}\}">i</figure-callout>}}_1,\mathrm{\&Lambda;},i_l)}=S_l+\mathrm{\&alpha;}{W_l}^{({\mathrm{<figure-callout\; id="1"\; label="i"\; filenames="A20071002753100241.png"\; state="\{\{state\}\}">i</figure-callout>}}_1,\mathrm{\&Lambda;},i_l)},$ Have

${W_l}^{(i_1,\mathrm{\&Lambda;},i_l)}=\sqrt{{\mathrm{\&rho;}}_{1,l}}{a}_l^{i_1}+\sqrt{{\mathrm{\&rho;}}_{2,l}}{a}_l^{i_1,i_2}+\mathrm{\&Lambda;}+\sqrt{{\mathrm{\&rho;}}_{l,l}}{a}_l^{i_1,\mathrm{\&Lambda;},i_l}---\left(14\right)$

Safe transmission for ease of fingerprint characteristic information, by DSP the transform domain that fingerprint characteristic data carries out Information hiding is handled, at transmitting terminal, fingerprint characteristic information and user ID are carried out mould two computings by the pseudo-random sequence that conversion generates, carry out the transmission error control Error Correction of Coding again, transmit in conjunction with chnnel coding again, receiving end carries out information extraction according to extraction algorithm in without the data of channel decoding, the data that propose are passed through error-correcting decoding again, carry out mould two computings with pseudo-random sequence and can get finger print data.

For the further security of enhanced system, also further comprise detection and identification step to collusion attack person, the detection of system and identification are divided into three processes: the range of nodes of at first adding up collusion attack person; Point out the tree node that collusion attack person is taken place by the correlativity vector again; At last, find the leaf at collusion attack person place, wherein, the correlativity vector that detects at i level subrange is:

${\mathrm{<figure-callout\; id="1"\; label="T\; i"\; filenames="A20071002753100241.png"\; state="\{\{state\}\}">T</figure-callout>}}_i\left(i_m\right)=\frac{{(Y-X)}^T\&CenterDot;a_{\mathrm{<figure-callout\; id="1"\; label="i"\; filenames="A20071002753100241.png"\; state="\{\{state\}\}">i</figure-callout>}1,.,i_m}}{\sqrt{{\left|\right|S\left|\right|}^2}}---\left(15\right)$

Its probability distribution when K user's appearance is:

$P\left({\mathrm{<figure-callout\; id="1"\; label="T\; i"\; filenames="A20071002753100241.png"\; state="\{\{state\}\}">T</figure-callout>}}_i\right|K,S_c,{\mathrm{\&sigma;}}_d^2)=N\left({\mathrm{\&mu;}}_{1,...,i_m}{\mathrm{\&sigma;}}_d^2\right)---\left(16\right)$

In the formula, σ _d ²Be the variance of received signal Y, and

Form be:

${\mathrm{\&mu;}}_{i1,...,i_{m-1}(i_m})=\frac{{\mathrm{<figure-callout\; id="1"\; label="k\; i"\; filenames="A20071002753100241.png"\; state="\{\{state\}\}">k</figure-callout>}}_i\&CenterDot;\sqrt{{\mathrm{\&rho;}}_m}}{K}\left|\right|S\left|\right|---\left(17\right)$

Wherein,

Be among K the user some survivor of i level subrange, by formula (17), we search for the place in i level subrange be that leaf is an index:

$\widehat{j_m}={\arg }_{i_m}^L\{{\mathrm{<figure-callout\; id="1"\; label="T\; i"\; filenames="A20071002753100241.png"\; state="\{\{state\}\}">T</figure-callout>}}_i\left(i_m\right)\&GreaterEqual;h_m\}---\left(18\right)$

Wherein, h _mDetermine by the false-alarm probability of detection of packets and the variance of received signal, when needing design system, take all factors into consideration the threshold value adjusted of setting.

A kind of fingerprint payment system is provided, comprises fingerprint capturer, peripheral control circuit, also comprise fingerprint image processor and fingerprint memory, outlet terminal, entry terminal, fingerprint image processor is to utilize said method that fingerprint image is handled and discerned.

Of the present inventionly be: handle by adopting DSP CCS2.2 that the fingerprint of gathering is carried out extreme value filtering, smothing filtering, Laplce's sharpening based on the fingerprint identification method of DSP and the beneficial effect of fingerprint payment system; Extract through pretreated fingerprint image characteristics point by the iteration threshold binaryzation; The cryptographic binding that the fingerprint image characteristics point that will extract through the iteration threshold binaryzation and user register also is stored in the database on the server; DSP is connected with server by the RS-232/485 interface, and server is connected with service terminal, utilizes DSP that fingerprint is carried out digital signal processing, has advantages such as processing speed is fast, flexible, accurate, antijamming capability is strong, volume is little.

Below in conjunction with drawings and Examples fingerprint identification method and the fingerprint payment system based on DSP of the present invention is described further:

Figure of description

Fig. 1 is the schematic diagram that the present invention is based on the fingerprint identification method of DSP;

Fig. 2 is the process flow diagram that the present invention is based on the fingerprint identification method fingerprint processing of DSP;

Fig. 3 be the fingerprint identification method that the present invention is based on DSP take the fingerprint unique point process flow diagram;

To be the fingerprint identification method that the present invention is based on DSP carry out the schematic diagram that the transform domain of Information hiding is handled with fingerprint characteristic data to Fig. 4;

Fig. 5 is the schematic diagram that the fingerprint identification method that the present invention is based on DSP embeds fingerprint characteristic data channel;

Fig. 6 is the schematic diagram of tree construction that the present invention is based on the fingerprint identification method of DSP;

Fig. 7 is the schematic diagram that the fingerprint identification method that the present invention is based on DSP carries out fingerprint characteristic orthogonal modulation;

Fig. 8 is that the server end finger print data that the present invention is based on the fingerprint identification method of DSP contrasts the process flow diagram of handling;

Fig. 9 is the systematic schematic diagram of a kind of system of fingerprints of the present invention;

Figure 10 is the structure principle chart of a kind of system of fingerprints of the present invention.

Embodiment

Below of the present invention based on the fingerprint identification method of DSP and the most preferred embodiment of fingerprint payment system, therefore do not limit protection scope of the present invention.

With reference to Fig. 1, Fig. 2, Fig. 3, Fig. 4, Fig. 5, Fig. 6, Fig. 7, Fig. 8, a kind of fingerprint identification method based on DSP is provided, may further comprise the steps:

A, employing DSP TMS320VC5000 carry out extreme value filtering, smothing filtering, Laplce's sharpening processing to the fingerprint of gathering;

B, extract through pretreated fingerprint image characteristics point by the iteration threshold binaryzation;

C, the fingerprint image characteristics point and the Account Data of user's registration that will extract through the iteration threshold binaryzation are bound and are stored in the database on the server;

D, DSP are connected with server by the RS-232/485 interface, and server is connected with service terminal.

The process that described fingerprint to collection carries out the extreme value Filtering Processing is as follows:

A, to be provided with a pending pixel region be s0, and 8 neighborhood territory pixels are arranged as follows around it:

$\left[\begin{array}{ccc}s1& s2& s3\\ s4& s0& s5\\ s6& s7& s8\end{array}\right]$

Get the average Ai of neighborhood related pixel earlier, i ∈ 1,2 ... 8} is one group of processing unit with four pixels, can improve correlation technique document extreme value filtering algorithm, is expressed as follows:

If b is A0＞max (Ai), i ∈ 1,2 ... 8} then

s1＝s2＝s4＝s0＝max(A1，A2，A4)

s2＝s3＝s5＝s0＝max(A2，A3，A5)

s4＝s6＝s7＝s0＝max(A4，A6，A7)

s5＝s7＝s8＝s0＝max(A5，A7，A8)

If c is A0＜min (Ai), i ∈ 1,2 ... 8} then

s1＝s2＝s4＝s0＝min(A1，A2，A4)

s2＝s3＝s5＝s0＝min(A2，A3，A5)

s4＝s6＝s7＝s0＝min(A4，A6，A7)

s5＝s7＝s8＝s0＝min(A5，A7，A8)

If d is min (Ai)＜=Ai＜=max (Ai), i ∈ 1,2 ... 8} then with the output of pixel initial value, does not process.

The process that described fingerprint to collection carries out The disposal of gentle filter is as follows:

Each pixel in the fingerprint and M are carried out convolution, and level and smooth convolution kernel is:

$M=\frac{1}{15}\left[\begin{array}{ccc}1& 2& 1\\ 2& 5& 2\\ 1& 2& 1\end{array}\right]$

The process that described fingerprint to collection carries out Laplce's sharpening processing is as follows:

The sharpening convolution kernel adopts the La Pulashi operator, and is as follows,

$h=\left[\begin{array}{ccc}-1& -1& -1\\ -1& 9& -1\\ -1& -1& -1\end{array}\right]$

By this convolution kernel image pixel is carried out the convolution budget, realize high-pass filtering, thereby make Laplace operator be used on the fingerprint image, and obtain the fingerprint ridge line after the sharpening.

Described as follows through the process of pretreated fingerprint image characteristics point by the extraction of iteration threshold binaryzation:

Take the method for iteration threshold, in the iteration threshold computing:

A, setting Initial Hurdle T, as make T=127 (gray level), the average gray value of fingerprint image is divided into two groups of R1, R2;

B, calculating two groups average gray value u1, u2;

C and then reset new gray scale threshold values T, new T is defined as: T=(u1+u2)/2;

D, fingerprint image is carried out threshold segmentation according to this T.

Database on the described server is based on the tree construction fingerprint database, is divided into a certain classification according to the fingerprint characteristic fingerprint of naming a person for a particular job, and searches for the affirmation user identity fast according to the classification of fingerprint, in fingerprint base design based on tree construction, and unique fingerprint vector

At tree node [i ₁...., i _l] go up according to Gaussian distribution N (0, σ _W ²) produce, wherein l is the degree of depth of node, and i is certain user node, and { a} is separate, for user u for the fingerprint vector collection simultaneously ⁽ⁱ⁾Its index is i=[i ₁, i ₂...., i _l], distribute to u ⁽ⁱ⁾Fingerprint copy X ⁽ⁱ⁾In, the L of a total embedding fingerprint is

Also can carry out the modulating transformation of orthogonal vector to the finger print data sequence in the database: conversion coefficient is divided into set of vectors according to the fingerprint key, make the vector quantization (COVQ of channel optimization by combination, channel-optimized VQ) the Euclid vector quantization index i* of acquisition standard, to given input information source vector v, from coded vector y _iAnd size is N _CbThe quantization index i* that obtains of information source code book can be expressed as

$i*=\arg \min (\underset{j=1}{\overset{N_{\mathrm{cb}}}{\mathrm{\&Sigma;}}}{P}_{j/i}\&CenterDot;{\left|\right|v-y_j\left|\right|}^2),$ i∈{0，1，Λ，N _cb-1}

Here, P _J/iReceive the conditional probability of j vector signal when being known index i, hiding finger print data at first carries out information source coding (depending on data character) and produces the symbol sebolic addressing { s of Q direction ₁, s ₂, Λ, s _Q, at the k dimension space, to each vector coefficients v of Q direction _JSymbol of middle embedding obtains to upset vector

, for the constraint condition of satisfied transmission,

The scope of value should be at radius

Circle in, like this, size just can be formed channel codebook for the k of Q dimension upset value; This channel codebook can be adjusted by the embedment strength α of decision transmission performance, promptly

Can be expressed as $\widehat{v_j}=v_j=\mathrm{\&alpha;}\&CenterDot;C\left(s_i\right),i=\mathrm{1,2},\mathrm{\&Lambda;},Q,$ C (s wherein _i) be exactly that size is the channel codebook of Q, before the transmission, the coefficient of upset is changed to host signal by contravariant; The data S that fingerprint is hidden carries out transform domain and decomposes after the pre-service of ranks direction, obtains transmission signals at last

, again will (x y) is divided into different fingerprint embedded blocks, i user u according to decomposition layer i and direction θ ⁽ⁱ⁾Fingerprint W ⁽ⁱ⁾Produce by formula (7):

$W^{\left(\text{i}\right)}=\sqrt{{\mathrm{\&rho;}}_1}{a}^{{\mathrm{<figure-callout\; id="1"\; label="i"\; filenames="A20071002753100241.png"\; state="\{\{state\}\}">i</figure-callout>}}_1}+\sqrt{{\mathrm{\&rho;}}_2}{a}^{{\mathrm{<figure-callout\; id="1"\; label="i"\; filenames="A20071002753100241.png"\; state="\{\{state\}\}">i</figure-callout>}}_1,i_2}+\mathrm{\&Lambda;}+\sqrt{{\mathrm{\&rho;}}_L}{a}^{{\mathrm{<figure-callout\; id="1"\; label="i"\; filenames="A20071002753100241.png"\; state="\{\{state\}\}">i</figure-callout>}}_1,i_2,.,i_L}---\left(7\right)$

Wherein, { ρ _lBe subjected to the probability decision of collusion attack by the different branches of corresponding fingerprint tree, and 0≤ρ is arranged ₁, Λ, ρ _L≤ 1, ${\mathrm{\&Sigma;}}_{j=1}^L{\mathrm{\&rho;}}_j=1$ Relation, adjust between the different user fingerprint that assignment is distributed according to the capacity of each grade fingerprint of embedment strength α control tree and self-adaptation;

For the value of fingerprint vector on the different nodes of tree, therefore, distribute to u ⁽ⁱ⁾The signal form of fingerprint video content be:

$X^{\left(i\right)}=\hat{S}+\mathrm{\&alpha;}{W}^{\left(i\right)}---\left(8\right)$

And in the modulation of quadrature fingerprint, host signal S can be divided into the part S of L non-overlapping copies ₁, Λ, S _L, like this at S _lIn the coefficient that can embed satisfy relation:

And satisfy ${\mathrm{\&Sigma;}}_{l=1}^L{N}_l=N---\left(9\right)$

In the fingerprint modulation of this proposition, definition matrix P is one and goes up the triangle square formation, at P ^JIn, in order to realize effective bandwidth,, make p for k＞l _k, l=0 for k≤l, makes 0＜p _{K, l}≤ 1 to realize robustness, definition E _{K, l}For being embedded into X _l ⁽ⁱ⁾Middle k level fingerprint vector

Capacity, then $E_l\stackrel{\mathrm{\&Delta;}}{=}{\mathrm{\&Sigma;}}_{k=1}^L{E}_{k,l}$ Be W _l ⁽ⁱ⁾All told,

At the 1st grade, p _{L, l}=1, at the 2nd≤l≤L level, known p _{L, l}, seek { p _{K, l}} _K＜lRelation value, satisfying a kind of condition is E _{1, l}: E _{2, l}: Λ: E _{L-1, l}=ρ ₁: ρ ₂: Λ, ρ _L-1So, but for k＜1 relation is arranged:

$p_{k,l}=\frac{{\mathrm{\&rho;}}_k(1-p_{l,l})}{{\mathrm{\&rho;}}_1+\mathrm{\&Lambda;}+{\mathrm{\&rho;}}_{l-1}}---\left(10\right)$

With known { p _{L, l}} _{L=1, Λ, L}The time matrix arranged

$P^J={\left[\begin{array}{cccc}1& 1-p_{\mathrm{2,2}}& \mathrm{\&Lambda;}& (1-p_{L,L})\frac{{\mathrm{\&rho;}}_1}{1-{\mathrm{\&rho;}}_L}\\ 0& p_{\mathrm{2,2}}& \mathrm{\&Lambda;}& (1-p_{L,L})\frac{{\mathrm{\&rho;}}_2}{1-{\mathrm{\&rho;}}_L}\\ M& \mathrm{<figure-callout\; id="0"\; label="M\; O\; M"\; filenames="A20071002753100241.png"\; state="\{\{state\}\}">M</figure-callout>}& O& M\\ 0& 0& 0& p_{L,L}\end{array}\right]}_{L\&times;L}---\left(11\right)$

In (11), work as $\underset{l=1}{\overset{L}{\mathrm{\&Sigma;}}}{N}_l=N,0\&le;N_l\&le;N$ The time, seek N ₁, N ₂, Λ, N _LSatisfy condition

P ^J[N ₁ N ₂ΛN _L] ^T＝N[ρ ₁ ρ ₂Λρ _L] ^T (12)

In (12) formula, work as p _{L, L}=ρ _LThe time, just be based on a kind of method that the quadrature fingerprint is modulated, so,, realize that the method for the multistage imbedding problem of fingerprint is according to above definition:

Condition and relation according to (11) formula $N_L=N-\underset{l=1}{\overset{L-1}{\mathrm{\&Sigma;}}}{N}_l,$ Each fingerprint vector for tree 1≤l≤L level

Have:

$a^{{\mathrm{<figure-callout\; id="1"\; label="i"\; filenames="A20071002753100241.png"\; state="\{\{state\}\}">i</figure-callout>}}_1,\mathrm{\&Lambda;},i_l}={a_l}^{i_1,\mathrm{\&Lambda;},i_l}\mathrm{\&Psi;}{a_{l+1}}^{{\mathrm{<figure-callout\; id="1"\; label="i"\; filenames="A20071002753100241.png"\; state="\{\{state\}\}">i</figure-callout>}}_1,\mathrm{\&Lambda;},i_l}\mathrm{\&Psi;}{a_{\mathrm{<figure-callout\; id="1"\; label="L\; i"\; filenames="A20071002753100241.png"\; state="\{\{state\}\}">L</figure-callout>}}}^{i_{}}---\left(13\right)$

Here,

According to Gaussian distribution N (0, σ _W ²) and separate.For k 〉=l, length is N _kFingerprint vector

Be embedded into S _kIn.Symbol " Ψ " has been represented the cascade factor.To u ⁽ⁱ⁾The i part fingerprint video content that the user receives is ${X_l}^{({\mathrm{<figure-callout\; id="1"\; label="i"\; filenames="A20071002753100241.png"\; state="\{\{state\}\}">i</figure-callout>}}_1,\mathrm{\&Lambda;},i_l)}=S_l+\mathrm{\&alpha;}{W_l}^{({\mathrm{<figure-callout\; id="1"\; label="i"\; filenames="A20071002753100241.png"\; state="\{\{state\}\}">i</figure-callout>}}_1,\mathrm{\&Lambda;},i_l)},$ Have

${W_l}^{({\mathrm{<figure-callout\; id="1"\; label="i"\; filenames="A20071002753100241.png"\; state="\{\{state\}\}">i</figure-callout>}}_1,\mathrm{\&Lambda;},i_l)}=\sqrt{{\mathrm{\&rho;}}_{1,l}}{a}_l^{i_1}+\sqrt{{\mathrm{\&rho;}}_{2,l}}{a}_l^{i_1,i_2}+\mathrm{\&Lambda;}+\sqrt{{\mathrm{\&rho;}}_{l,l}}{a}_l^{i_1,\mathrm{\&Lambda;},i_l}---\left(14\right)$

Safe transmission for ease of fingerprint characteristic information, by DSP the transform domain that fingerprint characteristic data carries out Information hiding is handled, at transmitting terminal, fingerprint secret information and user cipher are carried out mould two computings by the pseudo-random sequence that conversion generates, carry out the transmission error control Error Correction of Coding again, transmit in conjunction with chnnel coding again, receiving end carries out information extraction according to extraction algorithm in without the data of channel decoding, the data that propose are passed through error-correcting decoding again, carry out mould two computings with pseudo-random sequence and can get finger print data.

For the further security of enhanced system, also further comprise detection and identification step to collusion attack person, the detection of system and identification are divided into three processes: the range of nodes of at first adding up collusion attack person; Point out the tree node that collusion attack person is taken place by the correlativity vector again; At last, find the leaf at collusion attack person place, wherein, the correlativity vector that detects at i level subrange is:

${\mathrm{<figure-callout\; id="1"\; label="T\; i"\; filenames="A20071002753100241.png"\; state="\{\{state\}\}">T</figure-callout>}}_i\left(i_m\right)=\frac{{(Y-X)}^T\&CenterDot;a_{\mathrm{<figure-callout\; id="1"\; label="i"\; filenames="A20071002753100241.png"\; state="\{\{state\}\}">i</figure-callout>}1,.,i_m}}{\sqrt{{\left|\right|S\left|\right|}^2}}---\left(15\right)$

Its probability distribution when K user's appearance is:

$P\left({\mathrm{<figure-callout\; id="1"\; label="T\; i"\; filenames="A20071002753100241.png"\; state="\{\{state\}\}">T</figure-callout>}}_i\right|K,S_c,{\mathrm{\&sigma;}}_d^2)=N\left({\mathrm{\&mu;}}_{1,..,i_m}{\mathrm{\&sigma;}}_d^2\right)---\left(16\right)$

In the formula, σ _d ²Be the variance of received signal Y, and

Form be:

${\mathrm{\&mu;}}_{i1,...,i_{m-1}{(i}_m})=\frac{{\mathrm{<figure-callout\; id="1"\; label="k\; i"\; filenames="A20071002753100241.png"\; state="\{\{state\}\}">k</figure-callout>}}_i\&CenterDot;\sqrt{{\mathrm{\&rho;}}_m}}{K}\left|\right|S\left|\right|---\left(17\right)$

Wherein,

Be among K the user some survivor of i level subrange, by formula (17), we search for the place in i level subrange be that leaf is an index:

$\widehat{j_m}={\arg }_{i_m}^L\{{\mathrm{<figure-callout\; id="1"\; label="T\; i"\; filenames="A20071002753100241.png"\; state="\{\{state\}\}">T</figure-callout>}}_i\left(i_m\right)\&GreaterEqual;h_m\}---\left(18\right)$

Wherein, h _mDetermine by the false-alarm probability of detection of packets and the variance of received signal, when needing design system, take all factors into consideration the threshold value adjusted of setting.

With reference to Fig. 9, Figure 10, a kind of fingerprint payment system is provided, comprise fingerprint capturer 1, peripheral control circuit 2, also comprise fingerprint image processor and fingerprint memory 3, be DSP digital signal processor, outlet terminal 4, entry terminal 5, fingerprint image processor is to utilize said method that fingerprint image is handled and discerned.

Server end fingerprint detection to user's confirmation turn back to the fingerprint terminal user: in this fingerprint identification system, in order to guarantee that this important information of fingerprint characteristic value is not obtained by the disabled user, adopt specific digital protocol technology to solve, for convenience of description, the symbol that adopts in the agreement is done as giving a definition: A is the user, AS is a certificate server, KU _ASBe certificate server PKI, T _ASBe the time limit of certificate server, N _ABe the current data of A, F _ABe the fingerprint characteristic value of A, ID _ABe the sign of A, what also it should be noted that employing here is the unilateral authentication agreement.

Basic agreement is described below:

(1) A uses the signature that oneself identifies to certificate server AS request authentication, use signature technology can stop a could comprise bogus authentication server that the duplicity of user A is connected effectively, because have only legal certificate server just to preserve user's PKI, thereby can verify that this is signed obtains ID _ACome to use for following verification process.

(2) certificate server produces time limit T _AS, current data N _A, and with oneself PKI KU _AS, N _AWith time limit T _ASPKI KU with user A _AReturn to the party A-subscriber of client after the encryption.

(3) customer end A receives certificate server PKI, time limit and current data N _A, read user's fingerprint gray level image at the fingerprint sensor of client simultaneously, and obtain fingerprint characteristic F _A, tuple { T _AS, N _A, F _AWith the PKI KU of certificate server _ASSend to certificate server after the encryption.

The integrality of security, data, the anti-property denied and information privacy are the essential problems of considering of data transmission network identity verification scheme designing institute, based on the authentication agreement of fingerprint secret information: guarantee that communication authentication can prevent that third-party storming from hitting.Authentication based on the fingerprint biological characteristic: can solve password and spy on and problems such as key management difficulty, but be difficult to stop third-party storming to hit, thereby, the solution of the fingerprint identity validation system that the present invention proposes combines foregoing fingerprint identification technology and Information Hiding Techniques and obtains afterwards.

Server end is arranged on the inner PC server of trade center or bank, and its function of finishing is:

(a) search of the finger print data of registered user input and user registration number, password or ID number (cell-phone number etc. all can), as the off-line part of Fig. 9, its processing procedure is handled identical with the client fingerprint);

(b) set up customer data base and corresponding data search engine mechanism based on the tree construction of linux system;

(c) user profile that client transmissions is come is extracted decoding, and carries out fingerprint comparison according to the online treatment process of Figure 10, obtains the affirmation information of a user identity;

(d) server end returns confirmation, and carries out user's processing of deducting fees and transfer accounts.

Therefore, the gordian technique of server end technology realization is: user fingerprints database that the tree construction classification is set up and the quick technology of confirming user identity of searching for.

The working specification of finger print identification verification equipment: connect service terminal by the RS-232/485 serial ports, can carry out read operation to fingerprint scanner automatically, obtain the information such as fingerprint template that need, equipment is positioned over teller's working top, convenient fingerprint collecting checking at any time.Each shopping outlet is provided with the platform number that needs according to its teller window number correspondence.When the shopping clearing, the user only need confirm to consume the amount of money, and gathers fingerprint with the fingerprint read head that finger touches on the payment terminal, imports user's search number (normally user's mobile phone or home phone number) then.

The fingerprint identity of server and the affirmation identity of dealing money and dealing money return a confirmation code by server end according to protocol information and can finish processing after determining.The fingerprint payment process gets the Green Light and confirms.

Fingerprint payment system does not use actual all fingerprints of user, just extracts small unique point from user's fingerprint, and they are encrypted the unique fingerprint characteristic data that changes into the user with fingerprint algorithm of the present invention.These fingerprint characteristic datas enough come the unique and every other people's difference of user's identity, therefore in fact, do not keep and store user's fingerprint on the server of system, using the fingerprint characteristic data of being gathered is to copy user's physics fingerprint image or complete finger print information by backstepping, and all fingerprint characteristic datas can be stored in the financial settlement grade server system of highest security and confidentiality certainly.

Claims (9)

1. the fingerprint identification method based on DSP is characterized in that, may further comprise the steps:

A, employing DSP TMS320VC5000 carry out extreme value filtering, smothing filtering, Laplce's sharpening processing to the fingerprint of gathering;

B, extract through pretreated fingerprint image characteristics point by the iteration threshold binaryzation;

C, the fingerprint image characteristics point data and the Account Data of user's registration that will extract through the iteration threshold binaryzation are bound and are stored in the tree construction database on the server;

D, DSP are connected with server by the RS-232/485 interface, and server can be connected with a plurality of DSP service terminals.

2. the fingerprint identification method based on DSP according to claim 1 is characterized in that, the process that described fingerprint to collection carries out the extreme value Filtering Processing is as follows:

A, to be provided with a pending pixel region be s0, and 8 neighborhood territory pixels are arranged as follows around it:

$\left[\begin{array}{ccc}s1& s2& s3\\ s4& s0& s5\\ s6& s7& s8\end{array}\right]$

Get the average Ai of neighborhood related pixel earlier, i ∈ 1,2 ... 8} is one group of processing unit with four pixels, can improve correlation technique document extreme value filtering algorithm, is expressed as follows:

If b is A0＞max (Ai), i ∈ 1,2 ... 8} then

s1＝s2＝s4＝s0＝max(A1，A2，A4)

s2＝s3＝s5＝s0＝max(A2，A3，A5)

s4＝s6＝s7＝s0＝max(A4，A6，A7)

s5＝s7＝s8＝s0＝max(A5，A7，A8)

If c is A0＜min (Ai), i ∈ 1,2 ... 8} then

s1＝s2＝s4＝s0＝min(A1，A2，A4)

s2＝s3＝s5＝s0＝min(A2，A3，A5)

s4＝s6＝s7＝s0＝min(A4，A6，A7)

s5＝s7＝s8＝s0＝min(A5，A7，A8)

If d is min (Ai)＜=Ai＜=max (Ai), i ∈ 1,2 ... 8} then with the output of pixel initial value, does not process.

3. the fingerprint identification method based on DSP according to claim 1 is characterized in that,

The process that described fingerprint to collection carries out The disposal of gentle filter is as follows:

Each pixel in the fingerprint and M are carried out convolution, and level and smooth convolution kernel is:

$M=\frac{1}{15}\left[\begin{array}{ccc}1& 2& 1\\ 2& 5& 2\\ 1& 2& 1\end{array}\right].$

4. the fingerprint identification method based on DSP according to claim 1 is characterized in that, the process that described fingerprint to collection carries out Laplce's sharpening processing is as follows:

The sharpening convolution kernel adopts the La Pulashi operator, and is as follows,

$h=\left[\begin{array}{ccc}-1& -1& -1\\ -1& 9& -1\\ -1& -1& -1\end{array}\right]$

By this convolution kernel image pixel is carried out the convolution budget, realize high-pass filtering, thereby make Laplace operator be used on the fingerprint image, and obtain the fingerprint ridge line after the sharpening.

5. the fingerprint identification method based on DSP according to claim 1 is characterized in that, and is described as follows through the process of pretreated fingerprint image characteristics point by the extraction of iteration threshold binaryzation:

Take the method for iteration threshold, in the iteration threshold computing:

A, setting Initial Hurdle T, as make T=127 (gray level), the average gray value of fingerprint image is divided into two groups of R1, R2;

B, calculating two groups average gray value u1, u2;

C and then reset new gray scale threshold values T, new T is defined as: T=(u1+u2)/2;

D, fingerprint image is carried out threshold segmentation according to this T.

6. the fingerprint identification method based on DSP according to claim 1 is characterized in that, also comprises a step, by DSP the transform domain that fingerprint characteristic data carries out Information hiding is handled.At transmitting terminal, fingerprint characteristic information and ID users are carried out mould two computings by the pseudo-random sequence that conversion generates, carry out the transmission error control Error Correction of Coding again, transmit in conjunction with chnnel coding again, receiving end carries out information extraction according to extraction algorithm in without the data of channel decoding, the data that propose are passed through error-correcting decoding again, carry out mould two computings with pseudo-random sequence and can get fingerprint characteristic data.

7. the fingerprint identification method based on DSP according to claim 1, it is characterized in that, database on the described server is based on the tree construction fingerprint database, be divided into a certain classification according to the fingerprint characteristic fingerprint of naming a person for a particular job, the affirmation user identity is searched in classification according to fingerprint fast, in fingerprint base design based on tree construction, unique fingerprint vector a ^l ₁ ^{.., l} ₁At tree node [i ₁...., i _l] go up according to Gaussian distribution N (0, σ _W ²) produce, wherein l is the degree of depth of node, and i is certain user node, and { a} is separate, for user u for the fingerprint vector collection simultaneously ⁽ⁱ⁾Its index is i=[i ₁, i ₂...., i _l], distribute to u ⁽ⁱ⁾Fingerprint copy X ⁽ⁱ⁾In, the L of a total embedding fingerprint is a ^l ₁, a ^l ₁ ^{, l} ₂..., a ^l ₁ ^{., l} _LAlso can carry out the modulating transformation of orthogonal vector to the finger print data sequence in the database: conversion coefficient is divided into set of vectors according to fingerprint ID key, make the vector quantization (COVQ of channel optimization by combination, channel-optimized VQ) the Euclid vector quantization index i* of acquisition standard, to given input information source vector v, from coded vector y _iAnd size is N _CbThe quantization index i* that obtains of information source code book can be expressed as

$i*=\arg \min (\underset{j=1}{\overset{N_{\mathrm{cb}}}{\mathrm{\&Sigma;}}}{P}_{j/i}\&CenterDot;{\left|\right|v-y_i\left|\right|}^2),$ i∈{0，1，Λ，N _cb-1} (1)

Here, P _J/iReceive the conditional probability of j vector signal when being known index i, hiding finger print data at first carries out information source coding (depending on data character) and produces the symbol sebolic addressing { s of Q direction ₁, s ₂, Λ, s _Q, at the k dimension space, to each vector coefficients v of Q direction _jSymbol of middle embedding obtains to upset vector

For the constraint condition of satisfied transmission, The scope of value should be at radius Circle in, like this, size just can be formed channel codebook for the k of Q dimension upset value; This channel codebook can be adjusted by the embedment strength α of decision transmission performance, promptly

Can be expressed as $\widehat{v_j}=v_j=\mathrm{\&alpha;}\&CenterDot;C\left(s_i\right),$ I=1,2, Λ, Q, wherein C (s _i) be exactly that size is the channel codebook of Q, before the transmission, the coefficient of upset is changed to host signal by contravariant; The data S that fingerprint is hidden carries out transform domain and decomposes after the pre-service of ranks direction, obtains transmission signals at last

Again will

(x y) is divided into different fingerprint embedded blocks, i user u according to decomposition layer i and direction θ ⁽ⁱ⁾Fingerprint W ⁽ⁱ⁾Produce by formula (2):

$W^{\left(i\right)}=\sqrt{{\mathrm{\&rho;}}_1}{a}^{i_1}+\sqrt{{\mathrm{\&rho;}}_2}{a}^{i_1,i_2}+\mathrm{\&Lambda;}+\sqrt{{\mathrm{\&rho;}}_L}{a}^{i_1,i_2,i_L}---\left(2\right)$

Wherein, { ρ _lBe subjected to the probability decision of collusion attack by the different branches of corresponding fingerprint tree, and 0≤ρ is arranged ₁, Λ, ρ _L,≤1, ${\mathrm{\&Sigma;}}_{j=1}^L{\mathrm{\&rho;}}_j=1$ Relation, adjust between the different user fingerprint that assignment is distributed according to the capacity of each grade fingerprint of embedment strength α control tree and self-adaptation; a ⁱ ₁ ^{.., i} ₁For the value of fingerprint vector on the different nodes of tree, therefore, distribute to u ⁽ⁱ⁾The signal form of finger print data content be:

$X^{\left(i\right)}=\hat{S}+{\mathrm{\&alpha;W}}^{\left(i\right)}---\left(3\right)$

And in the modulation of quadrature fingerprint, host signal S can be divided into the part S of L non-overlapping copies ₁,, S _L, like this at S _lIn the coefficient that can embed satisfy relation:

And satisfy ${\mathrm{\&Sigma;}}_{l=1}^L{N}_l=N---\left(4\right)$

In the fingerprint modulation of this proposition, definition matrix P is one and goes up the triangle square formation, at P ^JIn, in order to realize effective bandwidth,, make p for k＞1 _{K, l}=0, for k≤l, make 0＜p _{K, l}≤ 1 to realize robustness, definition E _{K, l}For being embedded into X _l ⁽ⁱ⁾Middle k level fingerprint vector a ⁱ ₁ ^{, Λ, i} _kCapacity, then $E_l\stackrel{\mathrm{\&Delta;}}{=}{\mathrm{\&Sigma;}}_{k=1}^L{E}_{k,l}$ Be W _l ⁽ⁱ⁾All told,

At the 1st grade, p _1,1=1, at the 2nd≤l≤L level, known p _{1, l}, seek { p _{K, 1}} _K＜lRelation value, satisfying a kind of condition is E _{1, l}: E _{2, l}: Λ: E _L-1, l=ρ ₁: ρ ₂: Λ, ρ _L-1So, but for k＜1 relation is arranged:

$p_{k,l}=\frac{{\mathrm{\&rho;}}_k(1-p_{l,l})}{{\mathrm{\&rho;}}_1+\mathrm{\&Lambda;}+{\mathrm{\&rho;}}_{l-1}}---\left(5\right)$

With known { p _{L, l}} _{L=1, Λ, L}The time matrix arranged

$P^J={\left[\begin{array}{cccc}1& 1-p_{\mathrm{2,2}}& \mathrm{\&Lambda;}& (1-p_{L,L})\frac{{\mathrm{\&rho;}}_1}{1-{\mathrm{\&rho;}}_L}\\ 0& p_{\mathrm{2,2}}& \mathrm{\&Lambda;}& (1-p_{L,L})\frac{{\mathrm{\&rho;}}_2}{1-{\mathrm{\&rho;}}_L}\\ M& M& O& M\\ 0& 0& 0& p_{L,L}\end{array}\right]}_{L\&times;L}---\left(6\right)$

In (6), work as $\underset{l=1}{\overset{L}{\mathrm{\&Sigma;}}}{N}_l=N,$ 0≤N _lDuring≤N, seek N ₁, N ₂, Λ, N _LSatisfy condition

P ^J[N ₁?N ₂ΛN _L] ^T＝N[ρ ₁?ρ ₂Λρ _L] ^T (7)

In (7) formula, work as p _{L, L}=p _LThe time, just be based on a kind of method that the quadrature fingerprint is modulated, so,, realize that the method for the multistage imbedding problem of fingerprint is according to above definition:

Condition and relation according to (7) formula $N_L=N-\underset{l=1}{\overset{L-1}{\mathrm{\&Sigma;}}}{N}_l,$ Each fingerprint vector a for tree 1≤l≤L level ⁱ ₁ ^{, Λ, i} _lHave:

$a^{i_1,\mathrm{\&Lambda;},i_l}=a_l^{i_1,\mathrm{\&Lambda;},i_l}{{\mathrm{\&Psi;a}}_{l+1}}^{i_1,\mathrm{\&Lambda;},i_l}\mathrm{\&Psi;\&Lambda;}{{\mathrm{\&Psi;a}}_L}^{i_1,\mathrm{\&Lambda;},i_l}---\left(8\right)$

Here, ${\left\{{a_k}^{i_1,\mathrm{\&Lambda;},i_l}\right\}}_{k=l,\mathrm{\&Lambda;},L}$ According to Gaussian distribution N (0, σ _W ²) and separate.For k 〉=1, length is N _kFingerprint vector a _k ⁱ ₁ ^{, Λ, i} ₁Be embedded into S _kIn.Symbol " Ψ " has been represented the cascade factor.To u ⁽ⁱ⁾The i part finger print data content that the user receives is $X_l^{(i_1,\mathrm{\&Lambda;},i_l)}=S_l+{\mathrm{\&alpha;W}}_l^{(i_1,\mathrm{\&Lambda;},i_l)},$ Have

$W_l^{(i_1,{\mathrm{\&Lambda;},i}_l)}=\sqrt{p_{1,l}}{a}_l^{i_1}+\sqrt{p_{2,l}}{a}_l^{i_1,i_2}+\mathrm{\&Lambda;}+\sqrt{p_{l,l}}{a}_l^{i_1,\mathrm{\&Lambda;},i_l}---\left(9\right).$

8. the fingerprint identification method based on DSP according to claim 1 is characterized in that, also further comprises detection and identification step to collusion attack person, and the detection of system and identification are divided into three processes: the range of nodes of at first adding up collusion attack person; Point out the tree node that collusion attack person is taken place by the correlativity vector again; At last, find the leaf at collusion attack person place, wherein, the correlativity vector that detects at i level subrange is:

$T_{i1,K,i_{L-1}}\left(i_m\right)=\frac{{(Y-X)}^T\&CenterDot;a_{i1,....,i_m}}{\sqrt{{\left|\right|S\left|\right|}^2}}---\left(10\right)$

Its probability distribution when K user's appearance is:

$P\left(T_{i1,..,i_{m-1}}\right|K,S_c,{\mathrm{\&sigma;}}_d^2)=N\left({\mathrm{\&mu;}}_{1,..,i_m,}{\mathrm{\&sigma;}}_d^2\right)---\left(11\right)$

In the formula, σ _d ²Be the variance of received signal Y, and μ _{I1., im}Form be:

${\mathrm{\&mu;}}_{i1,..,i_{m-1}(i_m})=\frac{k_{i1,...,{}_m}\&CenterDot;\sqrt{{\mathrm{\&rho;}}_m}}{K}|\left|S\right||---\left(12\right)$

Wherein, k _{I1, m}Be among K the user some survivor of i level subrange, by formula (12), we search for the place in i level subrange be that leaf is an index:

$\widehat{j_m}={\arg }_{i_m}^L\{T_{i1,.,i_{m-1}}\left(i_m\right)\&GreaterEqual;h_m\}---\left(13\right)$

Wherein, h _mDetermine by the false-alarm probability of detection of packets and the variance of received signal, when needing the design fingerprint recognition system, take all factors into consideration the threshold value adjusted that corresponding tree node is set.

9. fingerprint recognition system, it is characterized in that, comprise fingerprint capturer (1), peripheral control circuit (2), also comprise fingerprint image processor and fingerprint memory (3), outlet terminal (4), entry terminal (5), fingerprint image processor is to utilize the described method of claim 1-9 that fingerprint image is handled and discerned.

Images