KR102022510B1 - Method and apparatus for personal authentication using electrocardiogram signals - Google Patents

Abstract

According to one embodiment of the present invention, provided is a personal authentication method using an elecgtrocardiogram signal performed by a personal authentication apparatus. The personal authentication method using an elecgtrocardiogram signal comprises the steps of: extracting heartbeats from received elecgtrocardiogram signals; extracting features from each of the extracted heartbeats; extracting functions by inputting the extracted features into an artificial neural network; generating pattern distributions for each of the extracted functions, based on output values obtained by inputting the extracted features to each of the extracted functions; and determining whether to be same, based on a result of matching the pre-extracted authentication features with the generated pattern distributions.

Description

METHOD AND APPARATUS FOR PERSONAL AUTHENTICATION USING ELECTROCARDIOGRAM SIGNALS}

The present invention relates to a personal authentication method and apparatus using an electrocardiogram signal, and to a personal authentication method and apparatus using an electrocardiogram signal that can determine whether the same using the electrocardiogram signal.

Traditional authentication methods are based on numbers, letters and graphical passwords. However, in the non-biometric method, there is a risk that the password may be lost, lost, or stolen. Therefore, the use of biometric information for authentication is limited depending on the situation.

Electrocardiogram (ECG) signals have a biometric function of the individual's body and heart and can be used as an authentication method. (Non-Patent Document 1)

In the related art, there are many studies that show that an ECG signal can be used for personal authentication in a personal authentication method using an ECG signal. (Non-Patent Documents 2 to 5)

However, the conventional personal authentication method using the ECG signal still has a disadvantage. The method of Non-Patent Document 2 has a problem of measuring and using an ECG signal for a long time for authentication, and the method of Non-Patent Document 3 has a problem of requiring a long-term ECG signal for authentication, and the method of Non-Patent Document 4 There is a problem that requires an additional biometric database to obtain better results.

In addition, the methods of Non-Patent Documents 2 to 5 are based on Fiducial Features extracted from an ECG signal. There is a problem that there is a lot of noise in the extraction and calculation that affects the authentication result because it is not guaranteed to accurately extract the reference point and the feature in the actual operating environment.

In order to solve the problems of the related arts, it is necessary to develop a technology for personal authentication method that reduces noise of an electrocardiogram signal using non-fiducial features and is more suitable for a real environment.

 Biel, L., Pettersson, O., Philipson, L., Wide, P., 2001. ECG analysis: A new approach in human identification. IEEE Trans. Instrum. Meas. 50 (3), 808-812.  G. W *? * Bbeler, M. Stavridis, D. Kreiseler, R.-D. Bousseljot, and C. Elster, “Verification of humans using the electrocardiogram,” Pattern Recognit. Lett., Vol. 28, no. 10, pp. 1172-1175, Jul. 2007.  Y. N. Singh and S. K. Singh, “Evaluation of electrocardiogram for biometric authentication,” J. Inf. Secur., Vol. 3, no. 1, pp. 39-48, Jan. 2012.  Y. N. Singh and S. K. Singh, “Identifying individuals using eigenbeat features of electrocardiogram,” J. Eng., Vol. 2013, Feb. 2013, Art. ID 539284.  Juan Sebastian Arteaga-Falconi, Hussein Al Osman, and Abdulmotaleb El Saddik, “ECG Authentication for Mobile Devices”, IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 65, NO. 3, MARCH 2016.

The present invention has been invented to solve the above problems, using non-fiducial features, and using a multi-layered weighted fuzzy membership function based neural network with Weighted Fuzzy Membership Function (NEWFM). The present invention relates to a personal authentication method and apparatus using an electrocardiogram signal for performing personal authentication.

According to an embodiment of the present invention, a personal authentication method using an ECG signal may include extracting heartbeats from received ECG signals; Extracting features from each of the extracted heart rates; Extracting functions by inputting the extracted features into an artificial neural network; Generating pattern distributions for each of the extracted functions based on output values obtained by inputting the extracted features to each of the extracted functions; And determining whether they are the same, based on a result of matching the pre-extracted authentication features with the generated pattern distributions.

Before extracting the heartbeats, the method may further include normalizing an amplitude and a sampling rate of the received ECG signals.

Extracting the heartbeats may include extracting the heartbeats from the normalized ECG signals.

The extracting the heartbeats may include: selecting a predetermined number of ECG signal values from the ECG signals; And extracting the selected signal values into one heartbeat.

In the extracting of the functions, the artificial neural network has a multi-layer structure, and output values of the artificial neural networks included in the first layer may be input to the artificial neural network included in the second layer.

Extracting the functions comprises: inputting the extracted features into first artificial neural networks; Inputting values output from each of the first artificial neural networks into a second artificial neural network; And extracting functions output from the second artificial neural network.

After generating the pattern distribution, the method may further include calculating a first pattern value based on the generated pattern distributions.

The determining of the same may include: calculating a second pattern value by matching the authentication features with each of the generated pattern distributions; And comparing the calculated first pattern value with the size of the second pattern value and determining whether the same is the same.

The calculating of the first pattern value may include: calculating the first pattern value based on an area ratio of the generated pattern distributions and a difference between function values of the extracted functions and function values of previously extracted functions for each of others; Calculating a value may be included.

Computing the first pattern value,

의 수식에 의해 상기 제1 패턴값인 T_i를 산출하는 단계를 포함하고,

Calculating T _i , the first pattern value, by a formula of

Where n is the number of generated pattern distributions and extracted functions, m is the number of extracted features, PV _{n, j} is the area ratio of the j-th interval in the generated nth pattern distribution, D _{n, j} is the difference between the j th function value of the n th function extracted from the artificial neural network and the average of the j th function values of the n th functions among the extracted functions for each of the others, and the bias is the pattern value This is a preset value for calculation.

Computing the first pattern value,

의 수식에 의해 상기 제1 패턴값인 T_i를 산출하는 단계를 포함하고,

Calculating T _i , the first pattern value, by a formula of

Where n is the number of generated pattern distributions and extracted functions, m is the number of extracted features, PV _{n, j} is the area ratio of the j-th interval in the generated nth pattern distribution, D _{n, j} is the difference between the j th function value of the n th function extracted from the artificial neural network and the average of the j th function values of the n th functions among the extracted functions for each of the others, and the bias is the pattern value This is a preset value for calculation.

The determining of whether or not to be the same may include determining that the second pattern value belongs to an EKG signal of the same person when the second pattern value falls within a threshold range set based on the first pattern value.

According to another embodiment of the present invention, a personal authentication device using an electrocardiogram may include a communication unit configured to receive electrocardiogram signals; And extracting heartbeats from the received ECG signals, extracting features from each of the extracted heartbeats, inputting the extracted features into an artificial neural network, extracting functions, and applying the extracted functions to the extracted functions. Based on the output values obtained by inputting the extracted features, generating pattern distributions for each of the extracted functions, and based on a result of matching the extracted feature for authentication with the generated pattern distributions, It may include a control unit for determining.

The controller may normalize an amplitude and a sampling rate of the received ECG signals and extract the heartbeats from the normalized ECG signals before extracting the heartbeats.

The controller may select a predetermined number of ECG signal values from the ECG signals and extract the selected signal values as one heartbeat.

The artificial neural network has a multilayer structure, and output values of the artificial neural networks included in the first layer may be input to the artificial neural network included in the second layer.

The control unit may input the extracted features to first artificial neural networks, input values output from each of the first artificial neural networks to a second artificial neural network, and extract functions output from the second artificial neural network. Can be.

After generating the pattern distribution, the controller calculates a first pattern value based on the generated pattern distributions, and calculates a second pattern value by matching the authentication features to each of the generated pattern distributions. The size of the calculated first pattern value and the second pattern value may be compared to determine whether they are the same.

The controller may calculate the first pattern value based on an area ratio of the generated pattern distributions and a difference between the function values of the extracted functions and the function values of the extracted functions for each of the others.

The control unit,

의 수식에 의해 상기 제1 패턴값인 T_i를 산출하고,

T _i , which is the first pattern value, is calculated by a formula of

Where n is the number of generated pattern distributions and extracted functions, m is the number of extracted features, PV _{n, j} is the area ratio of the j-th interval in the generated nth pattern distribution, D _{n, j} is the difference between the j th function value of the n th function extracted from the artificial neural network and the average of the j th function values of the n th functions among the extracted functions for each of the others, and the bias is the pattern value This is a preset value for calculation.

The control unit,

의 수식에 의해 상기 제1 패턴값인 T_i를 산출하고,

T _i , which is the first pattern value, is calculated by a formula of

Where n is the number of generated pattern distributions and extracted functions, m is the number of extracted features, PV _{n, j} is the area ratio of the j-th interval in the generated nth pattern distribution, D _{n, j} is the difference between the j th function value of the n th function extracted from the artificial neural network and the average of the j th function values of the n th functions among the extracted functions for each of the others, and the bias is the pattern value This is a preset value for calculation.

The controller may set a threshold range based on the first pattern value, and if the second pattern value falls within the set threshold range, may determine that the cardiograph signal is the same.

According to the method and apparatus for personal authentication using an electrocardiogram signal according to an embodiment of the present invention, there is an advantage of performing personal authentication using an ECG signal that is more reliable than biometric data such as iris or fingerprint.

In addition, since features are extracted using non-reference features, there is an advantage that is suitable for application in a real environment over conventional techniques of extracting features using reference features.

In addition, since the use of a multilayer artificial neural network, there is an advantage that the personal authentication can be performed with high accuracy even if fewer features, that is, using samples.

BRIEF DESCRIPTION OF THE DRAWINGS The accompanying drawings, included as part of the detailed description in order to provide a thorough understanding of the present invention, provide embodiments of the present invention and together with the description, describe the technical features of the present invention.
1 is a flow chart briefly illustrating a personal authentication method using an electrocardiogram signal according to an embodiment of the present invention.
2 is a block diagram schematically illustrating a personal authentication device using an ECG signal according to an exemplary embodiment of the present invention.
3 is a view briefly illustrating a process of outputting a function using a multilayer artificial neural network in a personal authentication method using an electrocardiogram signal according to an embodiment of the present invention.
4 is a diagram briefly illustrating a process of generating pattern distributions in a personal authentication method using an electrocardiogram signal according to an embodiment of the present invention.
5 is a diagram illustrating an example of calculating an area ratio in a personal authentication method using an electrocardiogram signal according to an embodiment of the present invention.

In this specification, terms such as first and / or second are used only for the purpose of distinguishing one component from another component. In other words, it is not intended to limit the components by the above terms.

Components, features, and steps that are referred to herein as "comprising" mean that such components, features, and steps exist and are intended to exclude one or more other components, features, steps, and equivalents. This is not it.

Unless otherwise specified and stated in the singular, the plural forms are included. That is, the components and the like mentioned herein may mean the presence or addition of one or more other components.

Unless defined otherwise, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. to be.

In other words, terms such as those defined in the commonly used dictionaries should be construed as meanings consistent with the meanings in the context of the related art, and, unless expressly defined herein, are construed in ideal or overly formal meanings. It doesn't work.

Hereinafter, a personal authentication method and apparatus using an ECG signal according to an embodiment of the present invention will be described in detail with reference to the accompanying drawings.

1 is a flowchart schematically illustrating a personal authentication method using an ECG signal according to an embodiment of the present invention, and FIG. 2 is a block diagram briefly illustrating a personal authentication device using an ECG signal according to an embodiment of the present invention. .

1 and 2, a personal authentication method using an electrocardiogram signal according to an embodiment of the present invention is an operation method performed by the personal authentication device 100 using an electrocardiogram signal according to an embodiment of the present invention.

1 and 2, in the personal authentication method using an ECG signal according to an embodiment of the present invention, normalizing ECG signals (S101), extracting heartbeats (S103), and extracting features. Step S105, extracting functions S107, generating pattern distributions S109, calculating a first pattern value S111, calculating a second pattern value S113, and whether they are the same It may include the step (S115).

Normalizing the ECG signals (S101) is a step performed by the control unit 103 and normalizing the amplitude and sampling rate of the ECG signals received by the communication unit 101.

Normalizing the ECG signals (S101) is a step of normalizing the ECG signals when the ECG signals received by the communicator 101 are measured by different ECG measuring apparatuses. It can be used to normalize the amplitude of the received ECG signals.

Where E (i) is the i-th ECG signal value included in one ECG record, L is the length of one ECG record, and Emax is the maximum ECG signal included in one ECG record. Emin is the minimum ECG signal value included in one ECG record.

Extracting heartbeats (S103) is a step performed by the controller 103 and extracts heartbeats from the received ECG signals.

Extracting heartbeats (S103) may include selecting specific signal values from ECG signals whose amplitude and sampling rate are normalized, and extracting the selected specific signal values into one heartbeat.

For example, the controller 103 may extract an R point for each ECG signal from normalized ECG signals using a QRS search algorithm. In addition, the controller 103 may select 40 ECG signals at the same interval before the R point around the R point extracted from each ECG signal, and select 80 ECG signals at the same interval after the R point. In addition, the controller 103 may extract a total of 120 selected ECG signals as one heart rate.

Extracting the features (S105) is a step performed by the controller 103, and extracting a feature from each of the extracted heartbeats based on the feature extraction algorithm.

For example, the extracting of the features (S105) may extract the Har Wavelet Features using a Har Wavelet Transform.

The Har wavelet transform processes signals of different frequency bands by converting an input sampling frequency into another sampling frequency, and scale coefficients (a) and transition elements (d) from wavelet generation functions (ψ) ) And may be expressed through Equation 2.

The scale factor a and the transition element d in Equation 2 may be transformed for wavelet transform of the ECG signal through Equations 3 and 4.

는 고주파 신호이고,

는 저주파 신호이고, j는 레벨이고, 심전도 신호는

과

의 합을 통해 나타내어질 수 있다.here

Is a high frequency signal,

Is the low frequency signal, j is the level, and the ECG signal is

and

It can be represented by the sum of.

Referring to Equations 3 and 4, the scaling factor a and the transition element d are divided by two as the level j decreases.

Some of the extracted Har wavelet features may be classified as features for registration, and some of the other may be classified as features for authentication.

For example, if 34 features were extracted through the Har wavelet transform, 30 features could be classified as features for registration, and 4 features could be classified as features for authentication.

Extracting the functions (S107) is performed in the control unit 103, and inputs the extracted features to the artificial neural network to output the functions.

That is, the step of extracting the functions (S107) inputs the extracted features to the first artificial neural networks, inputs values output from each of the first artificial neural networks into the second artificial neural network, and outputs from the second artificial neural network. Extracting the functions.

For example, the artificial neural network includes a Neural Network with a Weighted Fuzzy Membership (NEWFM), and the artificial neural network may have a multi-layered structure.

Extracting the functions (S107) will be described in detail with reference to FIG. 3 is a view briefly illustrating a process of outputting a function using a multilayer artificial neural network in a personal authentication method using an electrocardiogram signal according to an embodiment of the present invention. Specifically, FIG. 3A is a diagram illustrating a process of outputting a function, FIG. 3B is a diagram illustrating an artificial neural network having a multilayer structure, and FIG. 3C is a diagram illustrating an example of calculating a function value of an output function.

3A and 3B, the first artificial neural networks 201 and the second artificial neural network 203 are configured to have a hierarchical structure. For example, a two-layer neuro fuzzy network may be established.

The extracted Har wavelet features may be equally divided according to the number of first neural networks 201 and input to the first neural networks 201. That is, the artificial neural network has a multilayer structure, and output values of the artificial neural networks included in the first layer may be input to the artificial neural network included in the second layer.

Each of the first artificial neural networks 201 may output a Sugeno value S, and the output Segino values may be input to the second artificial neural network 203.

The second neural network 203 may output the same number of functions as the number of the first neural networks 201. For example, the output function may include a bounded sum of weighted fuzzy memberships (BSWFM).

In this example, the artificial neural network is constructed with four first neural networks 201 and one second neural network 203, but the artificial neural network is not limited to the present example, but depends on the number of neural neural networks. The constructed neural network can be expanded.

Referring to FIG. 3C, the boundary sum of the output k-th function, for example, the k-th weighted fuzzy-proportional function, may represent function values of the i-th person and others.

P _{k, j} is the function value of the i th person for the k th function, O _{k, j} is the mean of the function values of others for the k th function, and D _{k, j} is the i th person for the k th function Is the difference between the function value of and the average of the function values of others.

Generating Pattern Distributions (S109) A step performed by the control unit 103, and generating pattern distributions for each of the extracted functions based on output values obtained by inputting extracted features to each of the extracted functions. Step.

Based on the function values as many as the number of Har wavelet features for registration, a pattern distribution for each of the people can be generated. A process of generating a pattern distribution will be described in detail with reference to FIG. 4.

4 is a diagram briefly illustrating a process of generating pattern distributions 205 in a personal authentication method using an electrocardiogram signal according to an embodiment of the present invention.

Referring to FIG. 4, the extracted Har wavelet features may be input to a two-layer neuro fuzzy network, and pattern distributions 205 may be generated based on a result output from the neuro fuzzy network. have. For example, the third pattern distribution generated for the first person may be P _3,1 and the second pattern distribution generated for the fifth person may be expressed as P _2,5 .

The number of generated pattern distributions 205 is the same as the number of first artificial neural networks 201.

The calculating of the first pattern value (S111) is performed by the controller 103, and calculating the first pattern value 207 based on the generated pattern distributions 205.

Specifically, the step S111 of calculating the first pattern value 207 includes the area ratios of the generated pattern distributions 205 and the function values of the extracted functions and the function values of the functions extracted for each of the others. Computing the first pattern value 207 based on the difference between the two.

The area ratio of the pattern distributions 205 will be described with reference to FIG. 5. 5 is a diagram illustrating an example of calculating an area ratio in a personal authentication method using an electrocardiogram signal according to an embodiment of the present invention.

Referring to FIG. 5, the area ratio A _i is an area ratio occupied by some regions of the entire pattern distribution. Each of the pattern distributions 205 may be divided into as many parts as the number of Har wavelet features input to each of the functions, and an area ratio of each part may be calculated.

The first pattern value 207 may be calculated using Equation 5 below.

Where n is the number of generated pattern distributions 205 and extracted functions, m is the number of extracted features, A _{n, j} is the area ratio of the jth interval in the generated nth pattern distribution, D _{n, j} is the difference between the j-th function value of the n-th function extracted from the artificial neural network and the j-th function value of the n-th function among the extracted functions for each of the others.

The first pattern value 207 may be calculated using Equation 6 in addition to Equation 5.

That is, the first pattern value 207 may be calculated for all regions of the pattern distribution as shown in Equation 5, or may be calculated only for the middle region of the pattern distribution as shown in Equation 6.

The calculating of the second pattern value (S113) is performed by the control unit 103 and may include calculating the second pattern value by matching the authentication features to each of the generated pattern distributions 205. have.

The features for authentication are classified into features for authentication among the features extracted through the Har wavelet transform.

For example, if 34 features were extracted through the Har wavelet transform, 30 features could be classified as features for registration, and 4 features could be classified as features for authentication.

The second pattern value V _j may be calculated using Equation 7 below.

Where A _{n, j} is the area ratio of the j-th interval in the generated n-th pattern distribution, and D _{n, j} is extracted from the j-th function value of the n-th function extracted from the artificial neural network and each of others. Difference in the average of the j th function values of the n th function among the functions.

Determining whether or not it is the same (S115) is performed by the control unit 103, and it is a step of determining whether or not it is the same based on a result of matching the extracted authentication features to the generated pattern distributions.

Determining whether or not it is the same (S115) may include comparing the calculated first pattern value 207 and the size of the second pattern value to determine whether it is the same.

If the second pattern value is within the range of preset thresholds, it may be determined to be the same person, and if the second pattern value is out of the range of preset thresholds, it may be determined to be not the same person.

For example, in the determining of whether or not it is the same (S115) it may be determined whether or not the same using the following relationship.

That is, the second pattern value (V _j) is a group belongs to the range of the threshold set (-α T _i to T _i + β) is determined by the same person, the second pattern value (V _j) is more than T _i -α small or it may be determined as not the same person is greater than T _i + β. Ti is the pattern value of the i th person.

The final result is voted on as a result of authentication for all features, so that it can be determined whether they are the same person.

Table 1 is a table showing the experimental results of the non-patent documents 2 to 5 and the personal authentication method using an electrocardiogram according to an embodiment of the present invention.

Algorithm Subjects TAR (%) FAR (%) Authent. length Enroll Length Non Patent Literature 2 74 97.00 3.00 10 s 10 s Non-Patent Document 3 73 82.00 7.00 1/2 of record 1/2 of record Non-Patent Document 4 73 95.55 3.00 10 heart beats 10 heart beats Non Patent Literature 5 73 84.93 1.29 4s 30 s Proposed Algorithm 73 89.04 1.27 4 heart beats 30 heart beats

First, the TAR (True Acceptance Rate) and the FAR (False Acceptance Rate) were calculated based on the same data set as the non-patent documents 2 to 5 using data sets of 73 subjects in the Physionet Database.

TAR is the probability of recognizing the registrant and succeeded in personal authentication, and FAR is the probability of misrecognizing the non-registrant as the registrant. TAR using the algorithm according to the invention is 89.04%, FAR is 1.27%.

 In the algorithm according to the present invention, the heart rate used for authentication is 4, the heart rate used for registration is 30, and the number of heart rates required for registration and authentication is lower than that of non-patent literatures, but the FAR is the lowest, so that the acceptance rate is remarkably low. Found to be low.

In other words, considering the number of heart rates required for registration and authentication, the algorithm according to the present invention yielded better results than non-patent literature in the same data set.

Although the description herein has been shown in some illustrative aspects, various modifications or changes may be made from the scope defined by the claims that follow, and the technical protection scope of the invention is set forth in the following claims. It must be decided by.

100: personal authentication device
101: communication unit
103: control unit
201: first neural networks
203: second artificial neural network
205 pattern distributions
207: first pattern value

Claims (20)

Extracting heartbeats from the received ECG signals;
Extracting features from each of the extracted heart rates;
Extracting functions by inputting the extracted features into an artificial neural network;
Generating pattern distributions for each of the extracted functions based on output values obtained by inputting the extracted features to each of the extracted functions; And
And determining whether they are the same, based on a result of matching the pre-extracted authentication features with the generated pattern distributions.
Operation method of personal authentication device. The method of claim 1,
Prior to extracting the heartbeats, normalizing the amplitude and sampling rate of the received electrocardiogram signals,
Extracting the heartbeats,
Extracting the heartbeats from the normalized electrocardiogram signals,
Operation method of personal authentication device. The method of claim 1,
Extracting the heartbeats,
Selecting a predetermined number of electrocardiogram signal values from the electrocardiogram signals; And
Extracting the selected signal values into a single heartbeat,
Operation method of personal authentication device. The method of claim 1,
In the step of extracting the functions,
The artificial neural network has a multi-layer structure, and output values of the artificial neural networks included in the first layer are input to the artificial neural network included in the second layer.
Operation method of personal authentication device. The method of claim 1,
Extracting the functions,
Inputting the extracted features into first artificial neural networks;
Inputting values output from each of the first artificial neural networks into a second artificial neural network; And
Extracting functions output from the second artificial neural network;
Operation method of personal authentication device. The method of claim 1,
After generating the pattern distribution,
Calculating a first pattern value based on the generated pattern distributions,
Determining whether the same is the same,
Calculating a second pattern value by matching the authentication features with each of the generated pattern distributions; And
Comparing the size of the calculated first pattern value and the second pattern value to determine whether the same,
Operation method of personal authentication device. The method of claim 6,
Computing the first pattern value,
Calculating the first pattern value based on an area ratio of the generated pattern distributions and a difference between function values of the extracted functions and function values of previously extracted functions for each of others.
Operation method of personal authentication device. The method of claim 6,
Computing the first pattern value,

Calculating T _i , the first pattern value, by a formula of
N is the number of the generated pattern distributions and the extracted functions, m is the number of the extracted features, A _{n, j} is the area ratio of the j-th interval in the generated nth pattern distribution, D _{n, j} is the difference between the j th function value of the n th function extracted from the artificial neural network and the average of the j th function values of the n th functions among the extracted functions for each of others.
Operation method of personal authentication device. The method of claim 6,
Computing the first pattern value,

Calculating T _i , the first pattern value, by a formula of
N is the number of the generated pattern distributions and the extracted functions, m is the number of the extracted features, A _{n, j} is the area ratio of the j-th interval in the generated nth pattern distribution, D _{n, j} is the difference between the j th function value of the n th function extracted from the artificial neural network and the average of the j th function values of the n th functions among the extracted functions for each of others.
Operation method of personal authentication device. The method of claim 6,
Determining whether the same is the same,
If the second pattern value falls within the threshold range set based on the first pattern value, determining that the card signal of the same person;
Operation method of personal authentication device. A communication unit for receiving electrocardiogram signals; And
Extract heartbeats from the received ECG signals,
Extract features from each of the extracted heart rates,
Extract the functions by inputting the extracted features into the artificial neural network,
Generating pattern distributions for each of the extracted functions based on output values obtained by inputting the extracted features to each of the extracted functions,
And a controller configured to determine whether they are the same based on a result of matching the extracted authentication features to the generated pattern distributions.
Personal authentication device. The method of claim 11,
The control unit,
Prior to extracting the heartbeats, normalizing the amplitude and sampling rate of the received ECG signals,
Extracting the heartbeats from the normalized electrocardiogram signals,
Personal authentication device. The method of claim 11,
The control unit,
Selecting a predetermined number of ECG signal values from the ECG signals,
Extracting the selected signal values into one heart rate,
Personal authentication device. The method of claim 11,
The artificial neural network,
Comprising a multi-layer structure, the output value of the artificial neural network included in the first layer is input to the artificial neural network included in the second layer,
Personal authentication device. The method of claim 11,
The control unit,
Input the extracted features into first artificial neural networks,
Input values output from each of the first artificial neural networks into a second artificial neural network,
Extracting the functions output from the second artificial neural network,
Personal authentication device. The method of claim 11,
The control unit,
After generating the pattern distribution,
Calculating a first pattern value based on the generated pattern distributions,
Calculating a second pattern value by matching the authentication features with each of the generated pattern distributions;
Comparing the magnitudes of the calculated first and second pattern values to determine whether they are the same;
Personal authentication device. The method of claim 16,
The control unit,
Calculating the first pattern value based on an area ratio of the generated pattern distributions and a difference between function values of the extracted functions and function values of previously extracted functions for each of others;
Personal authentication device. The method of claim 16,
The control unit,

T _i , which is the first pattern value, is calculated by a formula of
N is the number of the generated pattern distributions and the extracted functions, m is the number of the extracted features, A _{n, j} is the area ratio of the j-th interval in the generated nth pattern distribution, D _{n, j} is the difference between the j th function value of the n th function extracted from the artificial neural network and the average of the j th function values of the n th functions among the extracted functions for each of others.
Personal authentication device. The method of claim 16,
The control unit,

T _i , which is the first pattern value, is calculated by a formula of
N is the number of the generated pattern distributions and the extracted functions, m is the number of the extracted features, A _{n, j} is the area ratio of the j-th interval in the generated nth pattern distribution, D _{n, j} is the difference between the j th function value of the n th function extracted from the artificial neural network and the average of the j th function values of the n th functions among the extracted functions for each of others.
Personal authentication device. The method of claim 16,
The control unit,
Set a threshold range based on the first pattern value,
If the second pattern value falls within the set threshold range, it is determined that the card signal of the same person;
Personal authentication device.

Images