JP6604870B2 - Personal identification device, personal identification method, and personal identification program - Google Patents

Description

  The present invention relates to a personal identification device, a personal identification method, and a personal identification program.

  2. Description of the Related Art In order to recognize a user by a computer, a personal identification device for identifying an individual by measuring various biological phenomena such as a personal fingerprint, a face, and voice is known. 2. Description of the Related Art In recent years, various types of personal identification devices that identify an individual using an electrocardiographic waveform signal unique to the individual as a biological phenomenon measurement are also known. As an example of a personal identification device that identifies an individual using an electrocardiogram waveform signal, the pattern of the electrocardiogram waveform signal at rest of the individual is registered in advance as a biometric signature, and is registered with the subject's electrocardiogram waveform 2. Description of the Related Art Personal identification devices that identify individuals by comparing with biometric signatures are known (see, for example, Patent Document 1 and Patent Document 2).

Special table 2008-518709 gazette JP 2013-150806 A

  However, an individual's electrocardiogram waveform is different between a resting state and a physical condition that is different from a resting state such as during exercise and poor physical condition. For example, a difference in an individual electrocardiogram waveform appears due to heart rate fluctuations such as heart rate, heart rate strength, or the shape of one heart beat. Accordingly, when comparing an electrocardiogram waveform signal at rest of an individual with an electrocardiogram waveform during exercise or the like, it may be difficult to identify the individual even though they are the same individual. For example, identifying a person using a biometric signature registered from an electrocardiogram waveform signal at the time of the individual's rest is merely cutting out and comparing a part of the electrocardiogram waveform. May be difficult to identify. In other words, when identifying an individual using a biometric signature registered from an electrocardiogram waveform signal when the individual is at rest, the time-series electrocardiogram waveform signal is used. It is clear that the measured electrocardiographic waveform signal does not match the registered biometric signature. Accordingly, there is room for improvement in providing a personal identification device that can identify an individual in accordance with a change in the physical condition of the individual with respect to a variety of electrocardiographic waveforms.

  The present invention has been made in view of the above facts, and provides a personal identification device, a personal identification method, and a personal identification program that can identify an individual even when the heart rate of the individual fluctuates. Objective.

In order to achieve the above object, the personal identification device of the present invention passes through a storage unit that stores a feature amount related to a time characteristic for identifying an individual and a specific signal in a predetermined frequency band of the electrocardiographic waveform signal of the subject. A band-pass filter to be performed, a time characteristic acquisition unit that acquires a characteristic amount related to a time characteristic based on the specific signal that has passed through the band-pass filter, a characteristic amount related to a time characteristic acquired by the time characteristic acquisition unit, and A determination unit that determines whether the target person is the individual by comparing with a feature value related to time characteristics stored in the storage unit.

  According to the present invention, the time characteristic acquisition unit acquires the feature quantity related to the time characteristic based on the specific signal in the predetermined frequency band of the electrocardiographic waveform signal of the subject who has passed through the bandpass filter. The storage unit stores a feature amount related to a time characteristic for identifying an individual. The determination unit includes a feature amount related to a time characteristic acquired by the time characteristic acquisition unit and a feature amount related to a time characteristic stored in the storage unit. A comparison is made to determine whether the subject is an individual. In this way, since it is determined whether the target person is an individual using the feature amount related to the time characteristic, the individual can be identified even when the individual's heart rate fluctuates.

In the personal identification device, the storage unit further stores a characteristic amount related to a frequency characteristic for identifying an individual, and acquires a characteristic amount related to the frequency characteristic based on the specific signal that has passed through the band-pass filter. And the determination unit compares the feature quantity related to the time characteristic acquired by the time characteristic acquisition unit with the feature quantity related to the time characteristic stored in the storage unit, and acquires the frequency characteristic acquisition unit. It is determined whether the target person is the individual by comparing the feature quantity related to the frequency characteristic with the feature quantity related to the frequency characteristic stored in the storage unit. In this way, since the feature quantity related to the time characteristic and the feature quantity related to the frequency characteristic are used to determine whether the target person is an individual, the person can be identified with higher accuracy than when the feature quantity related to the time characteristic is used. Can be identified.

In the personal identification device, the feature quantity related to the frequency characteristic may be a physical quantity indicating each of a slope of the frequency characteristic related to the intensity of the specific signal and a slope of the frequency characteristic related to the phase of the specific signal. By using physical quantities indicating the slope of the frequency characteristic related to the intensity of the specific signal and the slope of the frequency characteristic related to the phase of the specific signal, the feature quantity related to the frequency characteristic can be easily specified.

In the personal identification device, the characteristic amount related to the time characteristic indicates each of an interval between peaks of the specific signal, an amplitude ratio of adjacent peaks of the specific signal, and an intersection of the specific signal and a predetermined time axis. It can be a physical quantity. Thus, by using the physical quantity indicating each of the interval between the peaks of the specific signal, the amplitude ratio of the adjacent peaks of the specific signal, and the intersection of the specific signal and the predetermined time axis, the characteristic quantity related to the time characteristic can be easily obtained. Can be identified.

In the personal identification device, the electrocardiographic waveform signal of the subject can be a detection signal obtained by detecting the electrocardiographic waveform of the subject seated on the seat of the vehicle. Thereby, it can be easily determined whether the subject person seated on the seat of the vehicle is an individual indicated by the pre-stored feature amount.

In the personal identification method, a feature amount related to a time characteristic is acquired based on a specific signal that has passed a predetermined frequency band in an electrocardiographic waveform signal of the subject, and the feature amount related to the acquired time characteristic is identified from the individual. It is determined whether or not the target person is the individual by comparing with a feature quantity related to time characteristics stored in a storage unit that previously stores feature quantities related to time characteristics.

In the personal identification method, the storage unit further stores a feature amount related to a frequency characteristic for identifying an individual, further acquires a feature amount related to a frequency characteristic based on the specific signal, and a feature related to the acquired time characteristic The target is compared with the feature quantity related to the time characteristic stored in the storage unit and the feature quantity related to the acquired frequency characteristic is compared with the feature quantity related to the frequency characteristic stored in the storage unit. It is determined whether it is the individual.

Personal identification program causes the computer to function as the personal identification device.

  Even with such a personal identification method and personal identification program, it is possible to identify an individual even when the heart rate of the individual fluctuates.

  As described above, according to the present invention, it is possible to identify an individual even when the heart rate of the individual fluctuates.

It is a block diagram which shows an example of the personal identification device which concerns on embodiment. It is a block diagram which shows an example which mounted the personal identification device which concerns on embodiment in the vehicle. It is a characteristic view which shows an example of the normalized electrocardiogram waveform signal. It is a characteristic view which shows an example of the electrocardiogram waveform signal at the time of rest and exercise. It is a characteristic view which shows an example of the spectrum about the frequency component contained in an electrocardiogram waveform signal. It is a characteristic view which shows an example of the heartbeat waveform signal at the time of rest and exercise. It is a characteristic view which shows an example of the time characteristic of an electrocardiogram waveform signal, and its feature-value. It is a characteristic view which shows an example of the frequency characteristic of a spectrum, and its feature-value. It is a characteristic view which shows the signal and the feature-value which are related to individual OP1 and OP2. It is a characteristic view which shows the signal and the feature-value which are related with individual OP3 and OP4. It is a characteristic view which shows the signal and the feature-value which are related to individual OP5 and OP6. It is a characteristic view which shows the signal and the feature-value which are related to individual OP7 and OP8. It is a characteristic view which shows the signal and feature-value relevant to a subject. It is an image figure which shows the spatial distance of a subject's feature-value and an individual's feature-value. It is a block diagram which shows an example of a structure of the personal identification device realizable by a computer. It is a flowchart which shows an example of the flow of the personal identification process performed with a computer.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

  In this embodiment, an individual is identified using an electrocardiographic waveform signal unique to the individual as a measurement of the biological phenomenon. The electrocardiographic waveform signal is a signal in which the excitation of the stimulus transmission system, which is a command of myocardial activity for the heart to send blood to the whole body, and the contraction of the myocardium are detected as changes in human body potential.

  FIG. 3 shows, as an example of the electrocardiogram waveform signal, a signal obtained by averaging and normalizing the electrocardiogram waveform signals of a plurality of (here, 10 persons) individuals. As shown in FIG. 3, an electrocardiographic waveform signal with the horizontal axis representing time and the vertical axis representing voltage has a peak called a PQRST wave.

  The P wave is a waveform detected when the atrium contracts and sends blood to the ventricle when blood is sent from the heart to the whole body, and the QRS wave is a waveform detected as the ventricle contracts. The T wave is a waveform detected when the ventricle returns to its original state. In the present embodiment, an individual is identified using an electrocardiographic waveform signal having a peak called a PQRST wave.

  FIG. 1 shows an example of the configuration of a personal identification device 10 according to this embodiment. FIG. 2 shows an example of a mode in which the personal identification device according to the embodiment of the present invention is mounted on a vehicle to identify a vehicle occupant whose heart rate fluctuates.

  As shown in FIG. 1, the personal identification device 10 includes an electrocardiogram waveform detection unit 12, a bandpass filter 14, a heartbeat waveform cutout unit 16, a registration / identification selection unit 18, a mode instruction unit 20, a waveform analysis unit 22, and features. A quantity extraction unit 24, a determination unit 30, a machine learning unit 32, and an output unit 36 are provided.

  The electrocardiogram waveform detection unit 12 acquires an electrocardiogram waveform of the subject to be identified. For example, as shown in FIG. 2, an electrocardiographic waveform is detected by acquiring an electrocardiographic waveform signal detected by a potential sensor 12S embedded in at least one of a vehicle steering, a seat back portion and a seat portion.

  The band-pass filter 14 performs filtering that passes a specific signal including a frequency component for acquiring a characteristic amount related to the characteristics of the electrocardiographic waveform from the electrocardiographic waveform signal. That is, filtering is performed to obtain an electrocardiographic waveform signal including a frequency component necessary for extracting a feature amount for identifying the subject. A band of a signal passing through the bandpass filter 14 will be described.

  FIG. 4 shows an example of ECG waveform signals at rest (when the heart rate is low) and exercise (when the heart rate is high), and both ECG waveform signals at rest and when exercising. FIG. 4 (1) shows an electrocardiogram waveform signal at rest, and FIG. 4 (2) shows an electrocardiogram waveform signal during exercise. FIG. 4 (3) shows a resting time with a dotted line for one heartbeat in the electrocardiographic waveform signal, and a solid line with an exercise time. As shown in FIG. 4, the amplitudes of the P wave and the T wave, the time interval between the P wave and the R wave, and the time interval between the R wave and the T wave are different. Therefore, the ECG waveform signal as it is cannot be used for personal identification. The inventor pays attention to the frequency component included in the electrocardiogram waveform signal, and considers using it for personal identification in consideration of the frequency component.

  FIG. 5 shows an example of a spectrum (frequency characteristic) for a frequency component included in the electrocardiographic waveform signal. FIG. 5 (1) shows the frequency characteristic regarding the magnitude | size (intensity or absolute value) of the spectrum about the frequency component contained in the ECG waveform signal, and FIG. 5 (2) shows the frequency component contained in the ECG waveform signal. The frequency characteristic regarding the phase of the spectrum about is shown.

  As shown in FIG. 5 (1), with respect to the absolute value of the spectrum, in the band BD1, the characteristics cross at rest and at the time of movement, but in the band BD2, the characteristics are almost translated on the logarithmic axis (that is, changed to a constant multiple). is doing. Further, as shown in FIG. 5 (2), with respect to the phase of the spectrum, the value of the phase is different in the band BD1 depending on the conditions at rest and in motion, and is almost the same value in the band BD2. In the time waveform having the spectral characteristics of the band BD2, as shown in FIG. 6, the shape of the time waveform does not change and only the interval changes. Thus, if the signal of the band BD1 is cut by the band pass filter 14 with respect to the electrocardiogram waveform signal, an electrocardiogram waveform signal independent of the heart rate can be obtained.

  In the example shown in FIG. 5, 0 to 10 Hz is shown as the band BD1. In addition, various experiments are performed on the band BD1, and a bandpass filter 14 that cuts below 10 Hz can be obtained by using the bandpass filter 14 that cuts below 10 Hz, and a bandpass filter 14 that cuts below 15 Hz is used. The experimental result that it was further preferable was obtained. Thereby, the cut frequency on the low frequency side is preferably set to 10 Hz, and more preferably set to 15 Hz. Moreover, since the spectrum of 80 Hz or more was at the same level as the noise floor, an experimental result was obtained that it is preferable to set the cut frequency on the high frequency side to 80 Hz. Therefore, in the present embodiment, the band of the electrocardiographic waveform signal that is passed by the bandpass filter 14 is set to 15 to 80 Hz.

  The one heartbeat waveform cutout unit 16 shown in FIG. 1 detects the peak for each heartbeat of the electrocardiogram waveform signal (specific signal) after passing through the bandpass filter 14 and cuts out one heartbeat waveform.

  The registration / identification selection unit 18 is a feature quantity indicating an individual to the personal identification device 10 in accordance with an instruction (instruction of either the registration mode or the identification mode) input by the operation of the mode instruction unit 20 by the individual or the subject. Or a process for determining whether the target person is an individual stored in advance. That is, the registration / identification selection unit 18 uses the one heartbeat waveform signal output from the one heartbeat waveform cutout unit 16 to obtain and store a feature amount indicating an individual, or the subject is an individual stored in advance. It is selected whether to use for determining whether or not there is.

  The waveform analysis unit 22 performs processing such as removal of hum noise caused by the commercial power supply on the one heartbeat waveform signal output from the one heartbeat waveform cutout unit 16. In addition, the waveform analysis unit 22 can perform, for example, discrete Fourier transform on the one heartbeat waveform signal output from the one heartbeat waveform cutout unit 16 to perform spectrum analysis to obtain frequency characteristics related to the absolute value and phase of the spectrum. .

  The feature quantity extraction unit 24 includes a time characteristic acquisition unit 26 and a frequency characteristic acquisition unit 28, and extracts feature quantities of the time characteristic of the electrocardiogram waveform signal and the frequency characteristic of the spectrum.

  FIG. 7 shows an example of the time characteristic of the electrocardiogram waveform signal (one heartbeat waveform signal) after passing through the bandpass filter 14 and the feature amount thereof. This electrocardiographic waveform signal is a waveform cut out by detecting a peak for each heartbeat by the one heartbeat waveform cutout unit 16, and a positive peak and a negative peak appearing before and after the peak used for detection of the one heartbeat. And N1p, N2p, N3p, N4p, and N5p. Here, N3p indicates a peak used for detection of one heartbeat. FIG. 7 (1) shows the time value of the peak of the electrocardiogram waveform signal, FIG. 7 (2) shows the amplitude value of the peak, and FIG. 7 (3) shows the time axis when the electrocardiogram waveform signal is 0 voltage. Indicates the intersection of

  In the present embodiment, the first feature amount to the third feature amount are used as an example of the feature amount related to the time characteristic of the electrocardiogram waveform.

  The first feature value is a time value indicating the waveform peak interval in the time characteristics of the electrocardiogram waveform. In the present embodiment, a time value indicating four types of peak intervals is set as the first feature amount. That is, the time value g2 between the first peak N1p and the second peak N2p, the time value g2 between the first peak N2p and the third peak N3p, the third peak N3p and the fourth peak Each of the time value g3 between the peaks with the peak N4p and the time value g4 between the peaks with the fourth peak N4p and the fifth peak N5p is defined as a first feature amount. Note that when the time values g1 to g4 between peaks are used as the first feature values, the first feature values can be uniquely expressed and specified in a four-dimensional space with the time values g1 to g4 as axes.

  The second feature amount is a ratio of peak amplitude values in the time characteristics of the electrocardiogram waveform. In the present embodiment, the ratio of the amplitude values of adjacent peaks is set as the second feature amount. That is, the ratio r2 between the amplitude value N1y of the first peak N1p and the amplitude value N2y of the second peak N2p, the ratio r2 of the amplitude value N2y of the first peak N2p and the amplitude value N3y of the third peak N3p, Each of the ratio r3 of the amplitude value N3y of the third peak N3p and the amplitude value N4y of the fourth peak N4p, and the ratio r4 of the amplitude value N4y of the fourth peak N4p and the amplitude value N5y of the fifth peak N5p Is the second feature amount. Note that when the ratio r1 to r4 of the amplitude values of adjacent peaks is used as the second feature quantity, the second feature quantity can be uniquely expressed and specified in a four-dimensional space with the ratios r1 to r4 as axes. .

  The third feature amount is an intersection with the time axis in the time characteristic of the electrocardiogram waveform. In the present embodiment, the time value indicating the intersection of the electrocardiogram waveform and the time axis is set as the third feature amount. That is, a time value d1 indicating the intersection of the electrocardiogram waveform connecting the first peak N1p and the second peak N2p and the time axis with the voltage 0 as the axis, and the heart connecting the second peak N2p and the third peak N3p. Time value d2 indicating the intersection between the radio wave shape and the time axis with voltage 0 as the axis, and the intersection between the electrocardiogram waveform connecting the third peak N3p and the fourth peak N4p with the time axis with voltage 0 as the axis Each of the time value d3 and the time value d4 indicating the intersection of the electrocardiogram waveform connecting the fourth peak N4p and the fifth peak N5p and the time axis with the voltage 0 as an axis is defined as a third feature amount. When time values d1 to d4 indicating intersections are used as the third feature values, the third feature values may be uniquely expressed and specified in a four-dimensional space with the time values d1 to d4 indicating the intersections as axes. it can.

  The time characteristic acquisition unit 26 included in the feature quantity extraction unit 24 according to the present embodiment acquires each of the first feature quantity to the third feature quantity described above as the feature quantity related to the time characteristic of the electrocardiogram waveform.

  FIG. 8 shows an example of the frequency characteristic of the spectrum and its feature amount. FIG. 8 (1) shows the distribution of the absolute value of the spectrum among the frequency characteristics of the spectrum, and FIG. 8 (2) shows the distribution of the phase of the spectrum.

  In the present embodiment, the fourth feature amount and the fifth feature amount are used as an example of the feature amount related to the frequency characteristic of the spectrum.

  The fourth feature amount is a slope of the distribution of the absolute value of the spectrum in the frequency characteristic of the spectrum. In the present embodiment, the value of the slope at the frequency of interest predetermined for the absolute value of the spectrum is set as the fourth feature amount. That is, the slope value SLa at the frequency of interest predetermined for the distribution of the absolute value of the spectrum in the frequency characteristics of the spectrum is set as the fourth feature amount. The example of FIG. 8A shows the slope SLa of the absolute value distribution of the spectrum when 40 Hz is set as the frequency of interest. Note that, when the gradient SLa is used as the fourth feature amount, the fourth feature amount can be uniquely expressed and specified in a one-dimensional space with the gradient SLa as an axis.

  The fifth feature amount is the slope of the spectrum phase distribution in the spectrum frequency characteristics. In the present embodiment, the value of the gradient at the frequency of interest predetermined for the phase of the spectrum is set as the fifth feature amount. That is, the slope value SLb at the frequency of interest predetermined for the distribution of the spectrum phase in the spectrum frequency characteristic is set as the fifth feature amount. The example of FIG. 8B shows the slope SLb of the phase distribution of the spectrum when 40 Hz is set as the frequency of interest. Note that, when the gradient SLb is used as the fifth feature amount, the fifth feature amount can be uniquely expressed and specified in a one-dimensional space with the gradient SLb as an axis.

  The frequency characteristic acquisition unit 28 included in the feature amount extraction unit 24 according to the present embodiment acquires each of the above-described fourth feature amount and fifth feature amount as the feature amount related to the frequency characteristic of the spectrum.

  The feature amount extraction unit 24 according to the present embodiment uses the feature amounts acquired by the time characteristic acquisition unit 26 and the frequency characteristic acquisition unit 28 to extract the feature amount included in the electrocardiographic waveform of the individual or the subject to be identified. can do. That is, when the first feature value to the fifth feature value are used as the feature value related to the time characteristic of the electrocardiogram waveform and the feature value related to the frequency characteristic of the spectrum, the values of the first feature value to the fifth feature value are used as axes. In other words, the feature quantity related to the time characteristic of the electrocardiogram waveform and the feature quantity related to the frequency characteristic of the spectrum are uniquely expressed and specified in a 14-dimensional space obtained by summing the dimensionality of each of the first feature quantity to the fifth feature quantity. it can.

  Accordingly, the feature amount extraction unit 24 extracts the feature amount included in the electrocardiographic waveform of the individual or the subject to be identified by specifying the position in the 14-dimensional space by each of the first feature amount to the fifth feature amount. can do.

  In the present embodiment, a case is described in which a feature amount is uniquely specified in a 14-dimensional space using all of the first feature amount to the fifth feature amount. The method of specifying the feature amount related to the frequency characteristic is not limited to this. For example, for each of the first to fifth feature amounts, a feature amount used for personal identification can be selected by principal component analysis, and the number of dimensions can be reduced or the feature amount can be weighted.

  The machine learning unit 32 illustrated in FIG. 1 includes a feature amount storage unit 34, and after performing machine learning on the feature amount of the electrocardiogram waveform extracted by the feature amount extraction unit 24 using the electrocardiogram waveform signal, the feature amount is provided. Store in the storage unit 34. The feature quantity of the electrocardiographic waveform stored in the feature quantity storage unit 34 is output from the machine learning unit 32 as a determination criterion used for personal identification.

  Specifically, the machine learning unit 32 is extracted by the feature amount extraction unit 24 when a registration mode is instructed and the registration / identification selection unit 18 selects a process for registering a feature amount indicating an individual. A feature value indicating an individual is generated using the feature value, and the generated feature value is stored in the feature value storage unit 34 as a criterion for personal identification. That is, a criterion for determining whether or not the subject of the electrocardiographic waveform signal detected by the electrocardiographic waveform detection unit 12 is an individual registered in advance is generated. Then, the generated determination criterion is stored in the feature amount storage unit 34. An example of generation of a determination criterion is a feature quantity indicating an individual using, as input data, feature quantities (feature quantities extracted by the feature quantity extraction unit 24) of a plurality of electrocardiogram waveforms (for example, each of a plurality of one heartbeat waveform signals). A feature value as a result of machine learning that is machine-learned so that one is output is a feature value indicating an individual.

  The determination unit 30 determines whether the target person based on the input feature amount is an individual registered in advance using a determination criterion. Then, the output unit 36 outputs the determination result of the determination unit 30. When the identification mode is instructed and the registration / identification selection unit 18 selects the process of identifying the target person, the determination unit 30 uses the feature amount extracted by the feature amount extraction unit 24 and the machine learning unit 32. It is determined whether or not the subject of the detected electrocardiographic waveform signal is an individual registered in advance by comparing with a determination criterion that is an individual feature amount stored in the feature amount storage unit 34.

  Here, the feature quantity of the electrocardiographic waveform of the individual or the subject can be uniquely specified in a 14-dimensional space using all of the first feature quantity to the fifth feature quantity. Therefore, the determination unit 30 of the present embodiment determines whether the target person is an individual registered in advance using the spatial distance of the feature amount for each of the individual (determination criterion) and the target person. In the present embodiment, the Euclidean distance is used as the spatial distance of the feature amount.

  Therefore, in the 14-dimensional space, the determination unit 30 has a spatial distance between the feature amount of the individual (determination criterion) stored in the feature amount storage unit 34 and the feature amount of the target person is 0 or within a predetermined spatial distance. In this case, it is determined that the target person is an individual.

  The personal identification device 10 is an example of a personal identification device according to the present invention. The bandpass filter 14 is an example of a bandpass filter according to the present invention. The time characteristic acquisition unit 26 included in the feature amount extraction unit 24 is an example of a time characteristic acquisition unit according to the present invention. The frequency characteristic acquisition unit 28 included in the feature amount extraction unit 24 is an example of a frequency characteristic acquisition unit according to the present invention. The determination unit 30 is an example of a determination unit according to the present invention. The feature amount storage unit 34 is an example of a storage unit according to the present invention.

  Next, the operation of the personal identification device 10 according to this embodiment will be described. The personal identification device 10 may operate in a registration mode or in an identification mode.

  First, a case where the personal identification device 10 operates in the registration mode in accordance with an instruction input through an operation by the mode instruction unit 20 will be described.

  The electrocardiogram waveform detection unit 12 acquires an individual electrocardiogram waveform (electrocardiogram waveform signal) registered in the personal identification device 10 and outputs it to the band-pass filter 14. The band-pass filter 14 performs filtering that passes a specific signal including a frequency component for acquiring a characteristic amount related to the characteristics of the electrocardiographic waveform from the electrocardiographic waveform signal output from the electrocardiographic waveform detection unit 12. Next, the one heartbeat waveform cutout unit 16 detects a peak for each heartbeat of the electrocardiogram waveform signal (specific signal) after passing through the bandpass filter 14 and cuts out one heartbeat waveform.

  Next, since the registration mode is instructed here by the mode instruction unit 20, the registration / identification selection unit 18 selects a process for registering a feature quantity indicating an individual in the personal identification device 10 and cuts out one heartbeat waveform. The one heartbeat waveform signal output from the unit 16 is output to the waveform analysis unit 22. The waveform analysis unit 22 performs processing such as removal of hum noise caused by the commercial power supply and discrete Fourier transform on the one heartbeat waveform signal output from the one heartbeat waveform cutout unit 16, and relates to the absolute value and phase of the spectrum. Perform spectrum analysis to obtain frequency characteristics.

  The signal output from the waveform analysis unit 22 is output to the feature amount extraction unit 24. The feature amount extraction unit 24 acquires the feature amount of the time characteristic of the electrocardiogram waveform signal in the time characteristic acquisition unit 26, and acquires the feature amount of the frequency characteristic of the spectrum in the frequency characteristic acquisition unit 28. The characteristic amount of each of the time characteristic and the frequency characteristic of the spectrum is extracted.

  The time characteristic acquisition unit 26, as feature amounts related to the time characteristic of the electrocardiogram waveform, ratios r1 to r4 of time values g1 to g4 between peaks that are first feature amounts and amplitude values of adjacent peaks that are second feature amounts. The time values d1 to d4 indicating the intersections of the electrocardiogram waveform and the time axis as the third feature amount are acquired. In addition, the frequency characteristic acquisition unit 28 acquires each of the slope SLa of the absolute value distribution of the spectrum, which is the fourth feature quantity, and the slope SLb of the phase distribution of the spectrum, which is the fifth feature quantity, as the feature quantities related to the frequency characteristics of the spectrum. To do.

  Then, the feature quantity extraction unit 24 uses the feature quantities acquired by the time characteristic acquisition unit 26 and the frequency characteristic acquisition unit 28 to extract the feature quantities included in the registered electrocardiogram waveform. That is, the first feature value to the fifth feature value are obtained. The first feature amount to the fifth feature amount are feature amounts that can be uniquely specified in the 14-dimensional space.

  Next, the machine learning unit 32 stores the feature quantity of the electrocardiographic waveform extracted by the feature quantity extraction unit 24 in the feature quantity storage unit 34 as a criterion after machine learning. That is, the machine learning unit 32 generates a feature amount indicating an individual using the feature amount extracted by the feature amount extraction unit 24 and stores it as a determination criterion. The machine learning unit 32 uses a plurality of feature values such as a plurality of average values and distribution ranges as determination criteria for feature values (feature values that can be specified in a 14-dimensional space) included in an individual's electrocardiogram waveform obtained by machine learning. It is stored in the feature amount storage unit 34.

  In this way, the feature quantity of the personal electrocardiographic waveform is registered. In this embodiment, an electrocardiographic waveform for five heartbeats is detected and registered for each of eight individuals OP1 to OP8 as an example of a feature quantity stored (registered) in the feature quantity storage unit 34 as a determination criterion.

  FIG. 9 shows signals and feature quantities related to the registered individuals OP1 and OP2. For the individual OP1, FIG. 9 (1) shows an electrocardiogram waveform signal after passing through the bandpass filter 14, FIG. 9 (2) shows the frequency characteristic of the absolute value of the spectrum, and FIG. 9 (3) shows the spectrum of the spectrum. The frequency characteristics of the phase are shown, and the 14-dimensional feature value is shown in FIG. In the feature amount shown in FIG. 9 (4), “1” indicates four values of the ratio r1 to r4 of the amplitude values of adjacent peaks as the second feature amount, and “2” indicates the first feature amount. Four values of time values g1 to g4 between certain peaks are shown, and “3” shows time values d1 to d4 indicating the intersections of the electrocardiogram waveform and the time axis as the third feature amount. Further, “4” indicates the slope SLb of the phase distribution of the spectrum as the fifth feature quantity and the slope SLa of the absolute value distribution of the spectrum as the fourth feature quantity.

  For the individual OP2, FIG. 9 (5) shows an electrocardiogram waveform signal after passing through the bandpass filter 14, FIG. 9 (6) shows the frequency characteristic of the absolute value of the spectrum, and FIG. 9 (7). The frequency characteristics of the phase of the spectrum are shown, and the 14-dimensional feature quantity is shown in FIG.

  Similarly, FIG. 10 shows signals and feature quantities related to individuals OP3 and OP4, FIG. 11 shows signals and feature quantities related to individuals OP5 and OP6, and FIG. 12 shows signals and features related to individuals OP7 and OP8. Amount indicated.

  Next, a case where the personal identification device 10 operates in the identification mode according to an instruction input by an operation by the mode instruction unit 20 will be described.

  When the personal identification device 10 operates in the identification mode, the feature quantity of the individual's electrocardiographic waveform is stored (registered) in advance in the feature quantity storage unit 34 as a criterion (FIGS. 9 to 12). .

  Further, when the personal identification device 10 operates in the identification mode, the electrocardiogram waveform detection unit 12 acquires the electrocardiogram waveform (electrocardiogram waveform signal) of the subject in the same manner as in the registration mode, and the bandpass filter In step 14, filtering is performed to pass a specific signal including a frequency component for acquiring a characteristic amount related to the characteristics of the electrocardiographic waveform. Then, the one heartbeat waveform cutout unit 16 cuts out one heartbeat waveform from the electrocardiogram waveform signal (specific signal) after passing through the bandpass filter 14.

  Next, since the identification mode is instructed by the mode instructing unit 20, the one heart rate waveform signal output from the one heart rate waveform cutout unit 16 is subjected to processing such as hum noise removal and spectrum analysis by the waveform analysis unit 22. And output to the feature amount extraction unit 24. The feature quantity extraction unit 24 uses the feature quantities acquired by the time characteristic acquisition unit 26 and the frequency characteristic acquisition unit 28 to extract the feature quantities included in the electrocardiographic waveform of the subject to be identified. The extracted feature amount is output to the determination unit 30.

  In the determination unit 30, the nearest neighbor search is performed in the 14-dimensional space with respect to the feature amount of the identification target subject as compared to the determination criterion, and it is determined whether the subject is one of the eight registered in advance. judge. In this way, the target person is identified.

  FIG. 13 shows signals and feature quantities related to the subject. FIG. 13 shows the ECG waveform signal in FIG. 13 (1), FIG. 13 (2) shows the frequency characteristics of the absolute value of the spectrum, and FIG. The frequency characteristics of the phase are shown, and the 14-dimensional feature quantity is shown in FIG.

  FIG. 14 shows a result of obtaining a spatial distance between the feature amount of the subject to be identified and the feature amounts of each of eight individuals registered in advance. As shown in FIG. 14, the feature amount of the subject to be identified has the shortest spatial distance from the feature amount of the registered individual OP1, and the subject can be determined to be the individual OP1.

  As described above, in the present embodiment, the frequency component for acquiring the characteristic amount related to the characteristics of the electrocardiographic waveform by the band pass filter 14 from the electrocardiographic waveform (electrocardiographic waveform signal) of the individual to be registered is included. Perform filtering to pass a specific signal. In the electrocardiogram waveform signal (specific signal) after passing through the band pass filter 14, the peak for each heartbeat is detected, and after one heartbeat waveform is cut out, the feature amount extraction unit 24 performs time characteristics of the electrocardiogram waveform signal. And the characteristic amount of the frequency characteristic of the spectrum are acquired. The acquired feature amount is machine-learned by the machine learning unit 32 and then stored in the feature amount storage unit 34 as a determination criterion.

  In addition, the electrocardiogram waveform (electrocardiogram waveform signal) of the subject to be identified is also filtered by the bandpass filter 14, and one heartbeat waveform is cut out from the electrocardiogram waveform signal (specific signal) after passing through the bandpass filter 14. After that, the feature amount extraction unit 24 acquires the feature amount of the time characteristic of the electrocardiogram waveform signal and the feature amount of the frequency characteristic of the spectrum. The acquired feature amount of the target person is compared with the feature value of the determination criterion stored in the feature amount storage unit 34 after machine learning in the machine learning unit 32 by the determination unit 30, and the target person is an individual registered in advance. It is determined whether there is any.

  Thus, according to the personal identification device using the electrocardiographic waveform of the present embodiment, it is possible to identify an individual even if the heart rate fluctuates. That is, if the electrocardiogram waveform at rest is registered, an individual can be identified using the electrocardiogram waveform acquired in that state even when the heart rate during exercise is increased.

  In the personal identification device 10 according to the present embodiment, each constituent element may be configured by hardware such as an electronic circuit having each function described above, and at least a part of each constituent element may be configured. The computer may be configured to realize the function.

  FIG. 15 shows an example of a configuration in which each component constituting the personal identification device 10 is realized by a computer.

  The personal identification device 10 shown in FIG. 15 includes a control device 40 configured by a computer having a central processing unit (CPU) 42, a random access memory (RAM) 44, and a read only memory (ROM) 46. The ROM 46 includes a personal identification program 48 for realizing the functions described above. The control device 40 includes an input / output interface (I / O) 50. The CPU 42, the RAM 44, the ROM 46, and the I / O 50 are connected via a bus 52 so that commands and data can be exchanged. The I / O 50 is connected to an electrocardiographic waveform detection sensor 54, a mode instruction unit 20 such as a keyboard, a nonvolatile memory 58, and an output unit 36.

  The control device 40 functions as the personal identification device 10 shown in FIG. 1 when the personal identification program 48 is read from the ROM 46 and expanded in the RAM 44 and the personal identification program 48 expanded in the RAM 44 is executed by the CPU 42. . The personal identification program 48 includes a bandpass filter 14, a heartbeat waveform cutout unit 16, a registration / identification selection unit 18, a waveform analysis unit 22, a feature amount extraction unit 24, and a determination in the personal identification device 10 illustrated in FIG. 1. A process for realizing any of the functions of the unit 30 and the machine learning unit 32 is included.

  FIG. 16 shows an example of the flow of processing in the control device 40 of the personal identification device 10 realized including a computer.

  In step 100, the control device 40 executes a process of acquiring an electrocardiogram waveform (electrocardiogram waveform signal) from the electrocardiogram waveform detection unit 12 that is the electrocardiogram waveform detection sensor 54. In step 102, the band-pass filter 14 passes a specific signal including a frequency component for acquiring a feature quantity related to the characteristics of the electrocardiographic waveform from the electrocardiographic waveform signal output from the electrocardiographic waveform detection unit 12. Perform filtering. In step 104, the one heartbeat waveform cutout unit 16 detects a peak for each heartbeat of the electrocardiogram waveform signal (specific signal) after the bandpass filter process, and executes a process of cutting out one heartbeat waveform. In step 106, the registration / identification selection unit 18 executes a process of acquiring information indicating the instruction mode (registration mode or identification mode) instructed from the mode instruction unit 20. In step 108, the registration / identification selection unit 18 determines whether the instructed instruction mode is the identification mode. If the instructed instruction mode is the registration mode, a negative determination is made in step 108, and the process proceeds to step 118. In step 118, the waveform analysis unit 22 performs processing such as hum noise removal and spectrum analysis on the one heartbeat waveform signal output from the one heartbeat waveform cutout unit 16. Next, in step 120, the feature amount extraction unit 24 executes a feature amount extraction process for acquiring the feature amount of the time characteristic of the electrocardiogram waveform signal and the feature amount of the frequency characteristic of the spectrum. Next, the machine learning unit 32 stores the feature quantity of the electrocardiogram waveform extracted by the feature quantity extraction unit 24 in step 122 in the nonvolatile memory 58 as the feature quantity storage unit 34 as a determination criterion after machine learning. To do. In this way, the feature quantity of the personal electrocardiographic waveform is registered.

  On the other hand, when the instruction mode instructed from the mode instruction unit 20 is the identification mode, an affirmative determination is made in step 108, and the waveform analysis unit 22 is output from the one heartbeat waveform cutout unit 16 in step 110 as in step 118. Further, processing such as removal of hum noise and spectrum analysis are performed on one heartbeat waveform signal. Next, similarly to step 120, the feature amount extraction unit 24 executes a feature amount extraction process in step 112 to acquire the feature amount of the time characteristic of the electrocardiogram waveform signal and the feature amount of the frequency characteristic of the spectrum.

  When the feature amount extraction processing is completed, in step 114, the determination unit 30 performs a nearest neighbor search in the 14-dimensional space for the feature amount of the target person to be identified, compared with the determination criterion. It is determined whether any of the eight individuals registered in advance. The determination result is output at the output unit 36 at step 116. In this way, the target person is identified.

  In the present embodiment, the case where the characteristic quantities of the time characteristics and the spectrum frequency characteristics of the ECG waveform signal are extracted from the ECG waveform signal has been described. However, the present invention is not limited to extracting both. Absent. For example, the same effect can be obtained by extracting and using only the feature quantity indicating the time characteristic of the electrocardiographic waveform signal.

  Moreover, although this invention was demonstrated using embodiment, the technical scope of this invention is not limited to the range as described in the said embodiment. Various modifications or improvements can be added to the above-described embodiment without departing from the gist of the invention, and embodiments to which such modifications or improvements are added are also included in the technical scope of the present invention.

DESCRIPTION OF SYMBOLS 10 Personal identification apparatus 12 ECG waveform detection part 14 Band pass filter 24 Feature-value extraction part 26 Time characteristic acquisition part 28 Frequency characteristic acquisition part 30 Determination part 32 Machine learning part 34 Feature-value storage part

Claims (6)

An interval between peaks of signals in a predetermined frequency band of the electrocardiogram waveform signal, an amplitude ratio of adjacent peaks of the signal in the predetermined frequency band, and a predetermined amplitude value with the signal in the predetermined frequency band A storage unit that stores a feature amount relating to a time characteristic for identifying an individual indicating each of the intersection positions with respect to the time axis .
A bandpass filter for passing a signal of the predetermined frequency band of the subject's electrocardiographic waveform signal as a specific signal,
Based on the specific signal that has passed through the band-pass filter, each of an interval between peaks of the specific signal, an amplitude ratio of adjacent peaks of the specific signal, and an intersection of the specific signal and the predetermined time axis A time characteristic acquisition unit for acquiring a characteristic amount related to the time characteristic to be shown
A determination unit that determines whether the target person is the individual by comparing a feature amount related to the time characteristic acquired by the time characteristic acquisition unit and a feature amount related to the time characteristic stored in the storage unit;
A personal identification device. The storage unit includes individuals indicating each of a slope of a frequency characteristic related to the intensity of a signal in a predetermined frequency band of the ECG waveform signal and a slope of a frequency characteristic related to a phase of the signal in the predetermined frequency band. Further memorizing the characteristic amount related to the frequency characteristic to be identified,
Based on the specific signal that has passed through the band-pass filter, a frequency characteristic that obtains a characteristic amount related to the frequency characteristic indicating each of the slope of the frequency characteristic related to the intensity of the specific signal and the slope of the frequency characteristic related to the phase of the specific signal. An acquisition unit,
The determination unit compares the characteristic amount related to the time characteristic acquired by the time characteristic acquisition unit with the characteristic amount related to the time characteristic stored in the storage unit, and the characteristic amount related to the frequency characteristic acquired by the frequency characteristic acquisition unit. The personal identification device according to claim 1, wherein a determination is made as to whether or not the target person is the individual by comparing the characteristic amount relating to the frequency characteristic stored in the storage unit. The personal identification device according to claim 1 or 2 , wherein the electrocardiographic waveform signal of the subject is a detection signal obtained by detecting an electrocardiographic waveform of the subject seated on a seat of a vehicle. Based on the particular signal that has passed through a predetermined frequency band range of the subject's electrocardiographic waveform signal, the interval of the peak of the specific signal, the amplitude ratio of adjacent peaks of the specific signal, and a predetermined and said specific signal To obtain a characteristic value related to the time characteristic indicating each intersection position with the time axis with the amplitude value as an axis ,
The acquired characteristic amount related to the time characteristic, the peak interval of the signal of the predetermined frequency band in the electrocardiographic waveform signal, the amplitude ratio of adjacent peaks of the signal of the predetermined frequency band, and the predetermined A characteristic amount relating to a time characteristic stored in a storage unit preliminarily storing a characteristic amount relating to a time characteristic for identifying an individual indicating each intersection position between a signal in a predetermined frequency band and a time axis having a predetermined amplitude value as an axis ; A personal identification method for determining whether or not the subject is the individual. The storage unit includes individuals indicating each of a slope of a frequency characteristic related to the intensity of a signal in a predetermined frequency band of the ECG waveform signal and a slope of a frequency characteristic related to a phase of the signal in the predetermined frequency band. Further memorizing the characteristic amount related to the frequency characteristic to be identified,
Based on the specific signal, further acquiring a characteristic amount relating to the frequency characteristic indicating each of the slope of the frequency characteristic related to the intensity of the specific signal and the slope of the frequency characteristic related to the phase of the specific signal ,
The feature quantity related to the acquired time characteristic is compared with the feature quantity related to the time characteristic stored in the storage unit, and the feature quantity related to the acquired frequency characteristic is compared with the feature quantity related to the frequency characteristic stored in the storage unit. The personal identification method according to claim 4, wherein it is determined whether the target person is the individual. The personal identification program for functioning a computer as a personal identification device as described in any one of Claims 1-3 .

Images