CN102184416A - Method and device for registering biometric sample - Google Patents

Abstract

The invention discloses a method for registering a biometric sample, which comprises the following steps of: registering a plurality of groups of biometric samples; obtaining quality evaluation parameters of the plurality of groups of biometric samples; judging the quality of the samples according to the quality evaluation parameters; and if the samples are qualified, storing the samples as the biometric samples. The method further comprises the step of prompting the sample quality information according to the obtained quality evaluation parameters. Through adding a step of evaluating the quality of the registered biometric samples in the method for registering the biometric sample provided by the invention, whether the collected biometric samples are uniformly distributed can be effectively evaluated; and the quality of the registered biometric samples is fed back to a user, so that the user can choose to re-register the biometric sample when the registered biometric sample is not qualified, therefore the quality of the registered biometric sample can be guaranteed, and the recognition efficiency is improved.

Description

A kind of method and device of registering biometric sample

Technical field

The invention belongs to the living things feature recognition field, relate to a kind of method of registering biometric sample.

Background technology

In the current information age, how accurately to identify a people's identity, protection information security, become a crucial social concern that must solve.Traditional authentication more and more is difficult to satisfy the demand of society owing to very easily forging and losing, and at present convenient with safe solution is exactly biometrics identification technology undoubtedly.It is not only fast succinct, and utilizes it to carry out the identification of identity, safety, reliable, accurate.Be easier to matching computer and safety, monitoring, management system integration simultaneously, realize automatic management.Because its wide application prospect, huge social benefit and economic benefit, biometrics identification technology obtains swift and violent development recent years, has caused extensive concern and great attention.So-called biological identification technology (Biometric IdentificationTechnology) is meant a kind of technology of utilizing human body biological characteristics to carry out authentication.More more specifically, biometrics identification technology is exactly that high-tech means is close combines by computing machine and optics, acoustics, biology sensor and biostatistics principle etc., utilizes intrinsic physiological property of human body (as fingerprint, palmmprint, appearance, iris etc.) and behavioural characteristic (as person's handwriting, sound, gait etc.) to carry out the evaluation of personal identification.

Since the characteristics of human body have human body intrinsic not reproducible uniqueness, this biological secret key can't duplicate, and is stolen or pass into silence, and utilizes biological identification technology to carry out the identity identification, safety, reliable, accurately.This technology can be widely used in government, army, bank, social welfare guarantee, ecommerce, safe defence etc.Occurred many biological identification technologies at present, as fingerprint recognition, palmmprint identification, iris recognition, face recognition, signature identification, voice recognition etc., wherein everybody is familiar with the most with fingerprint recognition, recognition of face, iris recognition.

Yet no matter be which kind of biological identification technology, its general workflow all is divided into registration and discerns two parts.The process of registration will be extracted the biometric sample of colony to be identified and be stored, as the reference frame of identifying.Biological recognition system is taken a sample to biological characteristic, and extract its unique feature and change into digital code, and further with the synthetic feature templates of these code-group.When people carry out authentication alternately with recognition system, recognition system is obtained its feature and is compared with feature templates, determining whether coupling, thus whether success of decision authentication, and therefore the quality of the feature templates of registration directly influences the accuracy and the speed of User Recognition.

Consider that condition variation, dynamic and randomness in the identifying are more intense, and the requirement that reduces the User Recognition process, requirement to the biometric sample of registration just relatively wants strict, need the more attitude of registered user such as recognition of face equipment, abundanter expression is to obtain to try one's best comprehensive biometric sample.Present method generally is to point out the user initiatively to adjust suggestion means such as attitude and distance by sound or interface in enrollment process, sample quality does not possess objectivity, and the feature samples to registration lacks a basic criteria of quality evaluation, cause the sample of a lot of registrations too single, increase difficulty to identifying, caused misclassification rate or reject rate all to rise.

Summary of the invention

Defective at the prior art existence, the present invention is directed to above situation and proposed a kind of method of registering biometric sample, improve the quality of the biometric sample of registration by the operation that increases the quality of assessing biometric sample, effectively improve recognition efficiency, reduce misclassification rate and reject rate.

A kind of method of registering biometric sample of the present invention comprises:

The many groups of registration biometric sample;

Obtain the quality assessment parameter of described many group biometric sample;

According to quality assessment parameter judgement sample quality,, preserve feature samples if qualified.

Also comprise: according to the quality assessment parameter prompting sample quality information that obtains.

Wherein, described quality assessment parameter prompting sample quality information according to acquisition comprises by sound or interface prompt.

Described registration is organized biometric sample more and is comprised many groups biometric sample of registering respectively under the multiple environment, and described multiple environment comprises: multiple attitude and/or multiple angles and/or multiple illumination condition.

The quality assessment parameter of the described many group biometric sample of described acquisition further comprises: the mean value that calculates many group biometric sample; Calculate the eigenwert of the covariance matrix of many group biometric sample; Obtain the quality assessment parameter of sample according to the eigenwert that calculates.

Described quality assessment parameter comprises: whether biometric sample scattering degree score, quality mark of conformity.

The present invention also provides a kind of device of registering biometric sample, and this device comprises:

Registration unit is used to register many group biometric sample;

Computing unit is used to calculate the quality assessment parameter of described many stack features sample;

Judging unit is used for according to quality assessment parameter judgement sample quality, if qualified, preserves feature samples.

Described device also comprises: Tip element is used for according to quality assessment parameter prompting sample quality information.

Described registration unit further comprises a plurality of current feature registration subelements, is used to register the biometric sample of multiple environment, and described multiple environment comprises multiple attitude and/or multiple angles and/or multiple illumination condition.

The mean value calculation subelement is used to calculate the mean values of organizing biometric sample more;

The variance computation subunit is used to calculate the eigenwert of the covariance matrix of many stack features sample;

Quality assessment calculation of parameter subelement is used for obtaining the quality assessment parameter according to the many groups biological characteristic that obtains covariance matrix eigenwert and preset threshold value originally.

A kind of method and device of registering biometric sample of the present invention, by in the process of registration feature samples, increasing the assessment of feature samples quality, and the assessment result of sample quality offered the user, allow the user close under the situation of requirement at the feature samples non-conformity of quality of registration, select to register again, improve the total quality of the feature samples of registration, thereby improve the effect of living things feature recognition.

Description of drawings

Fig. 1 is the method flow diagram of registration biometric sample in the prior art;

Fig. 2 is the method flow diagram of the registration biometric sample of the embodiment of the invention one;

Fig. 3 is the process flow diagram of many stack features of registration sample in the method for the embodiment of the invention one;

Fig. 4 is the process flow diagram of a posture feature sample of registration in the method for the embodiment of the invention one;

Fig. 5 is the method flow diagram of assessment sample quality in the embodiment of the invention one method;

Fig. 6 is the structure drawing of device of the embodiment of the invention two registration biometric sample;

Fig. 7 is the device registration unit structural drawing of the embodiment of the invention two registration biological characteristics;

Fig. 8 is the device characteristic registration sub-unit structure figure of the embodiment of the invention two registration biological characteristics;

Fig. 9 is the device computing unit structural drawing of the embodiment of the invention two registration biological characteristics.

Embodiment

In the biometric sample enrollment process commonly used at present, the process flow diagram of registration biometric sample comprises as shown in Figure 1: registration to be organized biometric sample and preserves described many group biometric sample.In the many groups of registration biometric sample processes, initiatively adjust attitude or typing angle so that the abundant as far as possible feature samples of registration according to the demand of registration sample by the user.Whether but the user can't assess the action of self carrying out has reached the effect of improving sample quality, and the quality of the biometric sample of registration still can not get ensureing.

In order to solve the defective of above-mentioned existing method, the invention provides a kind of method of registering biometric sample, as shown in Figure 2, in the process of registration feature samples, the step that increase is carried out quality evaluation to the biometric sample of registration, the sample quality of registration is fed back to the user, thereby guarantee the quality of the feature samples of registration.Below be the embodiment one of the method for the invention, with helping reader understanding's method provided by the present invention.

S1, the many groups of registration biometric sample.

At first, each user is before carrying out living things feature recognition, all to register the biometric sample that is used as the identification foundation in advance, for example: face characteristic sample, fingerprint characteristic sample, palm print characteristics sample, iris feature sample etc., in the process of registration, need register the many groups biometric sample under the various environment in advance as basis of characterization, the quantity of sample is many more, the unique point that sample covers is extensive more, and the identification accuracy is high more.Described various environment comprises multiple attitude and/or multiple angles and/or different distance and/or multiple illumination condition etc.Present embodiment is an example with registrant's face feature samples, adjust different attitudes by the prompting user, register to many facial images of all angles as far as possible, by the recognition of face program described many facial images are carried out feature extraction respectively again, obtain the face characteristic sample of registration, to satisfy registration sample distribution demand widely.Usually each user need register 10 to 30 stack features samples, and every stack features sample is made of multidimensional characteristic vectors, and the dimension of proper vector is many more, and unique point is many more, and recognition accuracy is high more, and common proper vector is 200 to 1000 dimensions.The registration biometric sample concrete operations of the embodiment of the invention one as shown in Figure 3, wherein, the biometric sample of the multiple environment of described registration comprises many stack features sample of registering multiple attitude respectively.Described multiple attitude comprises: watches camera attentively, slightly bows, and slight the new line, slight rotary head, perhaps other attitudes can obtain better effect.As in registration palm print characteristics sample process, then prompting: the five fingers close up, the five fingers open etc.Difference according to the feature of gathering is provided with different attitudes.And in enrollment process, change illumination condition and simulate actual environment for use, to adapt to actual environment-identification needs.Quantity according to the biometric templates of each user's needs registration, the feature samples quantity that pre-defined every kind of attitude need be registered, if each user need register 16 stack features samples, need four kinds of attitudes of registration, can preestablish every kind of attitude so and register four stack features samples, register these four kinds of attitudes four stack features samples separately respectively.

Register certain attitude feature samples idiographic flow as shown in Figure 4: at first, keep certain attitude by sound or interface prompt user after (for example: first kind, " please watch camera attentively "); Register a facial image, judge the validity of described facial image, and therefrom extract lineup's face feature; Judge whether current attitude face characteristic registration is finished, for example, whether reached the quantity that sets in advance, if registration is not finished, continue next group face characteristic of registration, if the face characteristic quantity of the current attitude of registration reaches predefined four groups, then finish the face characteristic registration of current attitude, according to current attitude and predefined attitude, determine the attitude that next step need be registered, enter down a kind of enrollment process of face characteristic template of attitude, finish up to all attitude registrations.

The quality assessment parameter of S2, the described many group biometric sample of acquisition.

Variance can accurately reflect the distribution and the average degree of one group of data, and variance is more little, illustrates that distribution is concentrated more, and variance is big more, illustrates that each opposite sex of data is big more.The present invention adopts the evaluating of determining the biometric sample quality based on the thought of variance, estimates the quality of the biometric sample of registration.Concrete steps as shown in Figure 5.

The first step: the mean value that calculates many group biometric sample;

If the feature samples number of registration is the N group, N group sample is respectively X ₁, X ₂..., X _N,, if

Be the mean value of described N stack features sample, then

Computing formula as follows:

$\stackrel{\‾}{X}=\frac{1}{N}\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{X}_i---\left(1\right)$

During concrete enforcement, if every stack features sample comprises the M dimensional feature vector, feature samples X _NCan be expressed as the matrix of M * 1

The mean value of many group biological characteristics

The mean value that comprises the M dimensional feature vector also can be expressed as the matrix of M * 1

Second step: the eigenwert of calculating the covariance matrix of many stack features sample.

With the variance of D (X) expression N group sample, the computing formula of variance is as follows:

$D\left(X\right)=\frac{1}{N-1}\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{(X_i-\stackrel{\‾}{X})}^2---\left(2\right)$

I is a natural number in the formula.

With the example that is characterized as of N group M dimension, represent all features with the form of the matrix of a M * N, for

Adopt formula (2) can obtain the variance of the m dimensional feature vector of described N stack features sample

Wherein

The average of the m dimensional feature vector of expression N stack features sample, by

Calculate.But, only can accurately not estimate out the distribution dispersion of sample with the variance of a certain dimensional feature vector because the eigenwert of each dimension of feature samples is a random arrangement.Formula (2) is launched and can obtain:

$D\left(X\right)=\frac{1}{N-1}({(X_1-\stackrel{\‾}{X})}^2+{(X_2-\stackrel{\‾}{X})}^2+...+{(X_N-\stackrel{\‾}{X})}^2)---\left(3\right)$

Use for reference formula (3), can obtain the covariance computing formula of many group samples, as shown in the formula (4):

$D\left(X\right)=\frac{1}{N-1}((X_1-\stackrel{\‾}{X})\×{(X_1-\stackrel{\‾}{X})}^{\′}+(X_2-\stackrel{\‾}{X})\×{(X_2-\stackrel{\‾}{X})}^{\′}+...+(X_N-\stackrel{\‾}{X}){\×(X_N-\stackrel{\‾}{X})}^{\′})$

(4) in the above-mentioned formula (4)

Be the matrix of M * 1 forming of the mean value of each dimension of N stack features sample,

Be the matrix of a M * M, concrete operation is as follows:

$(X_1-\stackrel{\‾}{X})\×{(X_1-\stackrel{\‾}{X})}^{\′}=\left[\begin{array}{cccc}(a_{11}-\stackrel{\‾}{a_1})(a_{11}-\stackrel{\‾}{a_1})& (a_{11}-\stackrel{\‾}{a_1})(a_{21}-\stackrel{\‾}{a_1})& ...& (a_{11}-\stackrel{\‾}{a_1})(a_{M1}-\stackrel{\‾}{a_1})\\ (a_{21}-\stackrel{\‾}{a_2})(a_{11}-\stackrel{\‾}{a_2})& (a_{21}-\stackrel{\‾}{a_2})(a_{21}-\stackrel{\‾}{a_2})& ...& (a_{21}-\stackrel{\‾}{a_2})(a_{M1}-\stackrel{\‾}{a_2})\\ ...& ...& ...& ...\\ (a_{M1}-\stackrel{\‾}{a_M})(a_{M1}-\stackrel{\‾}{a_M})& (a_{M1}-\stackrel{\‾}{a_M})(a_{M1}-\stackrel{\‾}{a_M})& ...& (a_{M1}-\stackrel{\‾}{a_M})(a_{M1}-\stackrel{\‾}{a_M})\end{array}\right]$

Obtaining D (X) thus is the matrix of a M * M, and above-mentioned D (X) is asked eigenwert.With reference to by the built-in function eig () in the Matlab tool box, can obtain the eigen vector of D (X) very easily.I.e. [v, d]=eig (D (X)); Wherein, v represents the proper vector of D (X), and d represents the eigenwert of D (X).As calculated, M the eigenwert that obtains D (X) is respectively: d1, d2 ..., dM.

The 3rd step: the quality assessment parameter that obtains sample according to the eigenwert that calculates.

Variance can accurately reflect the distribution and the average degree of one group of data, and variance is more little, illustrates that distribution is concentrated more, and variance is big more, illustrates that each opposite sex of data is big more; For covariance matrix, its eigenwert has reflected the scattering degree of data, and promptly eigenwert is big more, illustrates that the otherness of data is big more.The present invention wishes the sample that obtains difference to some extent, and therefore the eigenwert of the covariance matrix of asking is the bigger the better.

The quality assessment parameter can be many stack features sample scattering degree score or quality mark of conformity whether.By a large amount of sample trainings, can draw a minimum empirical value of covariance matrix eigenwert, be made as threshold value.By above-mentioned second step, can obtain a plurality of eigenwerts of covariance matrix, many stack features sample scattering degree score can obtain according to described a plurality of eigenwerts and threshold calculations.As: the scattering degree that the threshold value correspondence is set must be divided into 60 fens, when the average of a plurality of eigenwerts of the covariance matrix that obtains during all greater than threshold value, can pass through to calculate the ratio of the average of described a plurality of eigenwerts, calculate many stack features sample scattering degree score greater than threshold value.For the ease of subsequent operation, can whether mark of conformity be as the sample quality evaluating with quality, in the time of can being set to eigenwert when the covariance matrix that calculates all greater than this threshold value, judge that then sample quality is qualified, mark of conformity is set; Otherwise judge that this sample quality is defective, set defective sign.

S3, according to quality assessment parameter judgement sample quality, if qualified, preserve feature samples;

Quality assessment parameter judgement sample quality according to step S2 acquisition, if described quality assessment parameter is many stack features sample scattering degree score, when the many stack features sample scattering degree score that obtains is higher than the setting score, think that then the biometric sample of registration is qualified sample, and preserve and to organize biometric sample more; If described quality assessment parameter is a whether mark of conformity of quality, when judging that this is masked as mark of conformity, think that then the biometric sample of registration is qualified sample, and preserve should many group biometric sample.

When being defective, do not preserve many groups biometric sample of registration according to quality assessment parameter judgement sample quality.

S4, according to quality assessment parameter prompting sample quality information.

The quality assessment parameter can be many stack features sample scattering degree score or quality mark of conformity whether.If the quality assessment parameter is many stack features sample scattering degree score, can be after the biometric sample of register setting quantity, by the interface or/and the score of the biometric sample that the voice suggestion active user registers; If the quality assessment parameter is a whether mark of conformity of quality, when the quality assessment parameter that obtains is mark of conformity, by the interface or/and the biometric sample of voice suggestion active user registration is up-to-standard, otherwise by the interface or/and the biometric sample of voice suggestion active user registration is off quality.

By method provided by the invention, in the process of registration biometric sample, increase the evaluation operation of biometric sample quality, and the quality of sample is dedicates the user to, allow the user close under the situation of requirement, select to register again, improve the total quality of the feature samples of registration at the biometric sample non-conformity of quality of registration, thereby reduce reject rate and misclassification rate effectively, improve the efficient of living things feature recognition.

The present invention also provides a kind of device of registering biometric sample, and as shown in Figure 6, this device comprises:

Registration unit 61 is used to register many group biometric sample; Described registration is organized biometric sample more and is included in to register to organize biometric sample under the multiple environment respectively more, and described multiple environment comprises: multiple attitude and/or multiple angles and/or multiple illumination condition.

Computing unit 62 is used to calculate the quality assessment parameter of described many stack features sample;

Judging unit 63 is used for according to quality assessment parameter judgement sample quality, if qualified, preserves feature samples;

Tip element 64 is used for according to quality assessment parameter prompting sample quality information.

Wherein, as shown in Figure 7, described registration unit 61 further comprises a plurality of current feature registration subelements 610, as: first feature registration subelement, second feature registration subelement ... N feature registration subelement, be used to register predefined a certain attitude or angle, and/or at least one stack features sample under the illumination condition.

As shown in Figure 8, described current feature registration subelement further comprises: attitude prompting subelement 6101 is used for carrying out current attitude by sound or interface prompt user; Feature extraction subelement 6102 is used to take the image that comprises biological information, and therefrom extracts one group of biometric sample; Quantity judgment sub-unit 6103 is used to judge whether to register the biometric sample of finishing predetermined number.

Described computing unit 62 further comprises: mean value calculation subelement 621 is used to calculate the mean values of organizing biometric sample more; Variance computation subunit 622 is used to calculate the eigenwert of the covariance matrix of many stack features sample; Quality assessment calculation of parameter subelement 623 is used for covariance matrix eigenwert and preset threshold value according to the many groups biometric sample that obtains, obtains the quality assessment parameter.Described threshold value is the empirical value that arrives that the living things feature recognition method that adopted is trained in the process that reality is used.Described quality assessment parameter comprises: organize whether mark of conformity of the score of biometric sample scattering degree and quality more.

By device provided by the invention, in the process of registration biometric sample, increase the appraisal procedure of biometric sample quality, and the quality of sample is dedicates the user to, allow the user close under the situation of requirement, select to register again, improve the total quality of the feature samples of registration at the feature samples non-conformity of quality of registration, thereby reduce reject rate and misclassification rate effectively, improve the efficient of living things feature recognition.

More than to a kind of method of registering biometric sample provided by the present invention and device row detailed introduction, used specific case herein principle of the present invention and embodiment are set forth, the explanation of above embodiment just is used for helping to understand method of the present invention and core concept thereof; Simultaneously, for one of ordinary skill in the art, according to thought of the present invention, the part that all can change in specific embodiments and applications, in sum, this description should not be construed as limitation of the present invention.

Claims (10)

1. a method of registering biometric sample is characterized in that, comprising:

The many groups of registration biometric sample;

Obtain the quality assessment parameter of described many group biometric sample;

According to quality assessment parameter judgement sample quality,, preserve feature samples if qualified.

2. method according to claim 1 is characterized in that, also comprises: according to the quality assessment parameter prompting sample quality information that obtains.

3. method according to claim 2 is characterized in that, described quality assessment parameter prompting sample quality information according to acquisition comprises by sound or interface prompt.

4. according to claim 1 or 3 described methods, it is characterized in that, described registration is organized biometric sample more and is comprised many groups biometric sample of registering respectively under the multiple environment, and described multiple environment comprises: multiple attitude and/or multiple angles and/or multiple illumination condition.

5. method according to claim 4 is characterized in that, the quality assessment parameter of the described many group biometric sample of described acquisition further comprises: the mean value that calculates many group biometric sample; Calculate the eigenwert of the covariance matrix of many group biometric sample; Obtain the quality assessment parameter of sample according to the eigenwert that calculates.

6. according to claim 2 or 5 described methods, it is characterized in that described quality assessment parameter comprises: whether biometric sample scattering degree score, quality mark of conformity.

7. a device of registering biological characteristic is characterized in that, comprising:

Registration unit is used to register many group biometric sample;

Computing unit is used to calculate the quality assessment parameter of described many stack features sample;

Judging unit is used for according to quality assessment parameter judgement sample quality, if qualified, preserves feature samples.

8. device according to claim 7 is characterized in that, also comprises:

Tip element is used for according to quality assessment parameter prompting sample quality information.

9. according to claim 7 or 8 described devices, it is characterized in that, described registration unit further comprises a plurality of current feature registration subelements, is used to register the biometric sample of multiple environment, and described multiple environment comprises multiple attitude and/or multiple angles and/or multiple illumination condition.

10. device according to claim 9 is characterized in that, described computing unit further comprises:

The mean value calculation subelement is used to calculate the mean values of organizing biometric sample more;

The variance computation subunit is used to calculate the eigenwert of the covariance matrix of many stack features sample;

Quality assessment calculation of parameter subelement is used for obtaining the quality assessment parameter according to the many groups biological characteristic that obtains covariance matrix eigenwert and preset threshold value originally.

Images