Method for measuring heart rate intervals, and system for measuring heart rate intervals

Abstract

This technology provides highly accurate estimation of heart rate intervals without requiring prior machine learning. [Solution] The disclosed method includes (a) a step of extracting frequency components of the heart sound band from a heart sound signal including the first heart sound and the second heart sound, and generating a heart sound band waveform SBW by integrating the amplitude components of the heart sound band at each time; (b) a step of determining the time interval TI between the first heart sound and the second heart sound; and (c) a step of detecting the interval between the appearance of peak pairs having a time interval TI in the heart sound band waveform SBW to determine the heartbeat interval.

Description

Technical Field

[0001] The present disclosure relates to a method for measuring an inter-beat interval and a system for measuring an inter-beat interval.

Background Art

[0002] Patent Document 1 discloses a technique for acquiring a biological vibration signal of a sample using a piezoelectric sensor and generating an estimated electrocardiogram signal of an electrocardiogram from the biological vibration signal using a pre-trained estimation model.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] However, in the above prior art, since machine learning of an estimation model composed of a neural network is performed, there is a problem that it is necessary to acquire a large amount of learning data for each person using an electrocardiograph and perform machine learning in advance. Therefore, a technique for estimating an inter-beat interval with high accuracy without performing machine learning in advance is desired.

Means for Solving the Problems

[0005] (1) According to a first aspect of the present disclosure, there is provided a method for measuring an inter-beat interval. This method includes: (a) a step of extracting frequency components in a heart sound band from a heart sound signal including a first heart sound and a second heart sound, and generating a heart sound band waveform obtained by integrating amplitude components in the heart sound band at each time; (b) a step of determining a time interval between the first heart sound and the second heart sound; and (c) a step of detecting an appearance interval of a peak pair having the time interval in the heart sound band waveform and determining the inter-beat interval.

[0006] This method allows for accurate measurement of heart rate intervals using heart sound signals without the need for prior machine learning.

[0007] (2) A second embodiment of the present disclosure provides a system for measuring heart rate intervals. The system includes a heart sound sensor (110) that detects a heart sound signal including a first heart sound and a second heart sound; a heart sound band extraction unit (120) that extracts frequency components of the heart sound band from the heart sound signal and generates a heart sound band waveform by integrating the amplitude components of the heart sound band at each time; a heart sound interval determination unit (130) that determines the time interval between the first heart sound and the second heart sound; and a heart rate interval determination unit (160) that detects the interval between the appearance of peak pairs having the time interval in the heart sound band waveform and determines the heart rate interval.

[0008] This system allows for accurate measurement of heart rate intervals using heart sound signals without the need for prior machine learning. [Brief explanation of the drawing]

[0009] [Figure 1] An explanatory diagram showing examples of heart sound waveforms, template waveforms, and electrocardiogram waveforms. [Figure 2] A block diagram showing the configuration of the heart rate interval measurement system in the first embodiment. [Figure 3] A flowchart illustrating the processing procedure in the first embodiment. [Figure 4] An explanatory diagram showing the processing details in the first embodiment. [Figure 5] A diagram illustrating the detailed steps of step S20. [Figure 6] An explanatory diagram showing an example of cross-correlation processing based on the positional relationship between the heart sound waveform and the template waveform. [Figure 7] A block diagram showing the configuration of the heart rate interval measurement system in the second embodiment. [Figure 8] An explanatory diagram showing the correspondence between heart rate and the time interval of heart sounds in the second embodiment. [Figure 9] A flowchart illustrating the processing procedure in the second embodiment. [Modes for carrying out the invention]

[0010] A. First Embodiment: As shown in Figure 1, the waveform of the heart sound signal HSS includes the first heart sound HS1 and the second heart sound HS2. Hereafter, the first heart sound HS1 and the second heart sound HS2 together will be referred to as the "heart sound pair HSP". The heart sound pair HSP corresponds to one heartbeat. The interval between the two first heart sound HS1s, the interval between the two second heart sound HS2s, and the interval between the two heart sound pair HSPs correspond to the heart rate interval HRI (Heart Rate Interval). The heart rate interval HRI estimated using the heart sound signal HSS corresponds to the heart rate interval RRI (Relative Rate Interval) measured using the electrocardiogram signal.

[0011] However, since both the first heart sound HS1 and the second heart sound HS2 have multiple peaks, it is difficult to accurately detect the intervals of the first heart sound HS1 and the second heart sound HS2, which leads to a problem in that the accuracy of measuring the heart rate interval HRI deteriorates. Therefore, in this embodiment, we have devised a way to solve the problem caused by the presence of multiple peaks in the heart sound signal HSS and to measure the heart rate interval HRI with high accuracy.

[0012] According to the inventor's research, the heart rate interval (HRI) tends to fluctuate by about 100 ms. In contrast, the variation in the time interval (TI) between the first heart sound (HS1) and the second heart sound (HS2) was found to be small, at about 15 ms. Furthermore, it was found that the time interval (TI) between the first and second heart sounds (HS2) correlates with heart rate. Moreover, it was found that the relationship between the length of the time interval (TI) and heart rate differed considerably among multiple subjects.

[0013] Therefore, in this embodiment, as shown in Figure 1, a template waveform TPW is created that emphasizes the first heart sound HS1 and the second heart sound HS2 based on the time interval TI between the first heart sound HS1 and the second heart sound HS2, and the heart sound peak corresponding to each heart sound pair HSP is identified using this template waveform TPW. As a result, it is possible to estimate the heart rate interval HRI without false detection of peaks. Furthermore, since the template waveform TPW can be automatically created from the heart sound signal HSS, it has the advantage of not requiring prior training.

[0014] As shown in Figure 2, the heart rate interval measurement system 100 of the first embodiment includes a heart sound sensor 110, a heart sound band extraction unit 120, a heart sound interval determination unit 130, a template creation unit 140, a cross-correlation processing unit 150, a heart rate interval determination unit 160, and a display unit 170.

[0015] The heart sound sensor 110 is a sensor that detects the heart sound signal HSS of an occupant sitting in a seat of a vehicle, such as an automobile. The heart sound sensor 110 can be a piezoelectric sensor installed on the seat or a wearable sensor attached to the occupant's wrist, etc. The heart sound band extraction unit 120 uses the heart sound signal HSS detected by the heart sound sensor 110 to extract the frequency components of the heart sound band and generates a heart sound band waveform SBW by integrating the amplitude components of the heart sound band at each time point. The heart sound interval determination unit 130 uses the heart sound band waveform SBW to determine the time interval TI between the first heart sound HS1 and the second heart sound HS2 included in the heart sound signal HSS. The template creation unit 140 creates a template waveform TPW that includes multiple peaks occurring at the time interval TI. The cross-correlation processing unit 150 performs cross-correlation processing between the heart sound band waveform SBW and the template waveform TPW. The heart rate interval determination unit 160 uses the results of cross-correlation processing to detect the interval between peak pairs with time intervals TI in the heart sound band waveform SBW and determines the heart rate interval HRI. The display unit 170 displays the determined heart rate interval HRI.

[0016] The heart rate interval measurement system 100 can be realized as an ECU (Electronic Control Unit) having a processor and a memory. Also, the functions of each part of the heart rate interval measurement system 100 can be realized by the processor executing a computer program stored in the memory. Also, a part or all of each part may be realized by a hardware circuit.

[0017] As shown in FIGS. and

[0017] , in step S10, a heart sound signal HSS including the first heart sound HS1 and the second heart sound HS2 is detected using the heart sound sensor 110. In step S20, the heart sound band extraction unit 120 extracts the frequency components of the heart sound band from the heart sound signal HSS and generates a heart sound band waveform SBW obtained by integrating the amplitude components of the heart sound band at each time. The heart sound band is, for example, in the range of 25 Hz to 50 Hz.

[0018] As shown in FIG. , step S20 includes steps S21 to S23. In step S21, by performing frequency analysis of the heart sound signal HSS, complex number information (amplitude and phase) of a plurality of frequency bands is obtained for each sampling time. As a frequency analysis method, for example, wavelet transform can be used. In step S22, for each frequency band, the amplitude component is extracted by obtaining the absolute value of the complex number information. In step S23, the heart sound band waveform SBW is generated by adding up the amplitude components of the heart sound band at each sampling time. Thus, the heart sound band waveform SBW is a waveform obtained by integrating the amplitude components of the heart sound band at each time.

[0019] In step S30 of FIG. , the heart sound interval determination unit 130 determines the time interval TI between the first heart sound HS1 and the second heart sound HS2 using the heart sound band waveform SBW. Specifically, first, the heart sound interval determination unit 130 sequentially selects two adjacent peaks in the heart sound band waveform SBW, obtains the peak interval PIT therebetween, and creates a histogram HGM of the peak interval PIT. The heart sound interval determination unit 130 determines the peak interval PIT having the highest frequency in the histogram HGM as the time interval TI between the first heart sound HS1 and the second heart sound HS2.

[0020] In step S40, the template creation unit 140 creates a template waveform TPW for emphasizing heart sounds based on the time interval TI between the first heart sound HS1 and the second heart sound HS2. The template waveform TPW is a waveform including a plurality of peaks generated at the time interval TI.

[0021] The template waveform TPW shown in FIG. 4 has four peaks P1 to P4 generated at the time interval TI. The first peak P1 and the fourth peak P4 are minimum peaks with negative amplitudes, and the second peak P2 and the third peak P3 are maximum peaks with positive amplitudes. That is, this template waveform TPW has a waveform in which two maximum peaks P2 and P3 are provided in the center, and minimum peaks P1 and P4 are provided before and after them, respectively. The half-value width and amplitude of each peak can be arbitrarily set. However, it is preferable to set the half-value width and amplitude to the same value for a plurality of peaks.

[0022] The first peak P1 is a waveform in which the amplitude value monotonically increases from the minimum value to 0, and the fourth peak P4 is a waveform in which the amplitude value monotonically decreases from 0 to the minimum value. However, each of the minimum peaks P1 and P4 may be formed so as to have a first half portion in which the amplitude value monotonically decreases from 0 to the minimum value and a second half portion in which the amplitude value monotonically increases from the minimum value to 0. Also, the minimum peak may be omitted. That is, the template waveform TPW preferably has at least two maximum peaks P2 and P3.

[0023] In step S50, the cross-correlation processing unit 150 performs cross-correlation processing on the heart sound band waveform SBW and the template waveform TPW to detect a heart sound peak corresponding to the heart sound pair HSP. In step S50, first, while gradually changing the time of the heart sound band waveform SBW to which the template waveform TPW is applied, cross-correlation processing of the heart sound band waveform SBW and the template waveform TPW is performed at each time to generate a processing result CC shown in FIG. 4.

[0024] In the heart sound peak detection process in step S50, heart sound peaks whose amplitude value is greater than or equal to the amplitude threshold are detected in the cross-correlation processing result CC. It is also preferable to detect multiple heart sound peaks such that the interval between them is greater than or equal to the interval threshold. As the amplitude threshold, for example, the average amplitude value in the cross-correlation processing result CC can be used. As the interval threshold, any value can be used, for example, it can be set to a value between 0.3 seconds and 0.6 seconds. However, after monitoring the heart rate interval for a certain period of time, the heart rate interval will not increase or decrease significantly instantaneously, so the interval threshold may be readjusted according to the history of the heart rate interval.

[0025] In the cross-correlation processed result CC shown in Figure 4, white circles are placed at the tips of the heart sound peaks. Each heart sound peak corresponds to two peaks that appear at time interval TI in the heart sound band waveform SBW. In other words, in the cross-correlation processed result CC, each heart sound pair HSP, consisting of the first heart sound HS1 and the second heart sound HS2, appears as a single clear heart sound peak. Furthermore, since noise often does not produce peaks at the time interval TI between the first heart sound HS1 and the second heart sound HS2, the noise can be removed by cross-correlation processing.

[0026] By using a template waveform TPW with four peaks P1 to P4 as shown in Figure 4, a single clear peak corresponding to the first heart sound HS1 and the second heart sound HS2 can be easily detected by cross-correlation processing of the heart sound band waveform SBW.

[0027] As shown in Figure 6, in cross-correlation processing, the following three processing examples can be considered regarding the positional relationship between the heart sound band waveform (SBW) and the template waveform (TPW).

[0028] <Processing Example A> Processing Example A is an example of processing when the timing of the occurrence of peak Phs1, corresponding to the first heart sound HS1, and peak Phs2, corresponding to the second heart sound HS2, in the heart sound band waveform SBW coincides with the timing of the occurrence of the two maximum peaks P2 and P3 in the template waveform TPW. In this processing example A, a large peak is obtained as the cross-correlation processing result CC.

[0029] <Processing Example B> Processing Example B is an example of processing when the timing of the occurrence of peak Phs2, which corresponds to the second heart sound HS2 in the heart sound band waveform SBW, coincides with the timing of the occurrence of the first maximum peak P2 in the template waveform TPW. In this processing example B, the timing of the occurrence of peak Phs1, which corresponds to the first heart sound HS1 in the heart sound band waveform SBW, coincides with the timing of the occurrence of the first minimum peak P1 in the template waveform TPW, so the cross-correlation processing result CC will be close to zero.

[0030] <Processing Example C> Processing example C is an example of processing when the timing of the occurrence of peak Phs1, which corresponds to the first heart sound HS1 in the heart sound band waveform SBW, coincides with the timing of the occurrence of the second maximum peak P3 in the template waveform TPW. In this processing example C, the timing of the occurrence of peak Phs2, which corresponds to the second heart sound HS2 in the heart sound band waveform SBW, coincides with the timing of the occurrence of the last minimum peak P4 in the template waveform TPW, so the cross-correlation processing result CC will be close to zero.

[0031] As can be seen from the processing examples A to C above, by using a template waveform TPW with four peaks P1 to P4, one heart sound peak corresponding to one heart sound pair HSP, which consists of the first heart sound HS1 and the second heart sound HS2, can be easily detected through cross-correlation processing.

[0032] If we consider the case where a template waveform is used that has only two maximum peaks P2 and P3 and no minimum peaks P1 and P4, then an intermediate-sized peak will appear between the single heart sound peak corresponding to the first heart sound HS1 and the second heart sound HS2. This is because, since there are no minimum peaks P1 and P4, the cross-correlation processing result CC in the processing examples B and C described above will not be close to zero, but will be a somewhat large value. Therefore, in order to improve the detection accuracy of the heart rate interval HRI, it is preferable to use a template waveform TPW that has four peaks P1 to P4 as shown in Figure 4.

[0033] In step S60, the heart rate interval determination unit 160 determines the heart rate interval HRI from the intervals of the heart sound peaks detected in step S50. Since each heart sound peak in the cross-correlation processing result CC corresponds to one heart sound pair HSP consisting of the first heart sound HS1 and the second heart sound HS2, the interval between adjacent heart sound peaks can be determined as the heart rate interval HRI. Thus, it can be understood that the processing in steps S50 and S60 is a process of detecting the appearance interval of peak pairs having a time interval TI in the heart sound band waveform SBW and determining the heart rate interval HRI.

[0034] The determined heart rate interval (HRI) is displayed on the display unit 170, as illustrated in Figure 4. Alternatively, instead of displaying the HRI itself, the "instantaneous heart rate," calculated by dividing 60 seconds by the HRI, may be displayed. Furthermore, the HRI and instantaneous heart rate may be output to an external device other than the display unit 170. The process shown in Figure 3 is preferably repeated at regular intervals.

[0035] According to the first embodiment described above, the frequency components of the heart sound band are extracted from the heart sound signal HSS to generate a heart sound band waveform SBW, the time interval TI between the first heart sound HS1 and the second heart sound HS2 is determined, and the heart rate interval HRI is determined using this time interval TI and the heart sound band waveform SBW. Therefore, the heart rate interval HRI can be measured accurately without prior machine learning.

[0036] B. Second Embodiment: The heart rate interval measurement system 100 of the second embodiment shown in Figure 7 differs from the heart rate interval measurement system 100 of the first embodiment shown in Figure 2 in the following respects, while other configurations are the same as those of the first embodiment. (1) The heart rate sensor 180 has been added. (2) The heart sound interval determination unit 130a estimates the time interval TI between the first heart sound HS1 and the second heart sound HS2 using the heart rate signal HBS detected by the heart rate sensor 180, rather than the heart sound signal HSS.

[0037] The heart rate sensor 180 can be a wearable sensor, a camera, radar, or the like. For example, an electrical heart sensor attached to the subject's arm can be used as the heart rate sensor 180. Alternatively, a camera or radar can be used to photograph the part of the subject where pulsation due to the heartbeat appears, and the heart rate can be detected by analyzing the resulting image. Or, the heart sound sensor 110 can be repurposed as the heart rate sensor 180, and the heart rate can be estimated from the heart sound signal HSS.

[0038] The heart sound interval determination unit 130a estimates the time interval TI between the first heart sound HS1 and the second heart sound HS2 from the heart rate using a pre-set table TB. Table TB shows the relationship between the heart rate detected by the heart rate sensor 180 and the time interval TI between the first heart sound HS1 and the second heart sound HS2. Table TB can be experimentally set in advance for each subject. Furthermore, for example, if the system is configured to recognize subjects with a camera, it is possible to accumulate data for subjects recognized by the camera in Table TB without having to experimentally set it in advance.

[0039] In the example shown in Figure 8, the relationship between heart rate and time interval TI is set in tables TB_1 and TB_2, respectively, for two subjects. In these tables TB_1 and TB_2, the relationship between heart rate and time interval TI is represented by a straight line, but it could also be represented by a curve. Alternatively, the correspondence between heart rate and time interval TI could be set using other forms such as functions or maps, instead of tables TB. In this way, using heart rate information to determine the time interval TI can reduce processing load.

[0040] As shown in Figure 9, the processing procedure of the second embodiment is the same as that of the first embodiment, except that step S30 of the processing procedure of the first embodiment shown in Figure 3 is replaced with steps S110 and S120, and the other steps are the same as those of the first embodiment.

[0041] In step S110, the heart rate sensor 180 detects the heart rate. In step S120, the heart sound interval determination unit 130a estimates the time interval TI from the heart rate using the correspondence between the heart rate and the time interval TI. Steps S40 onward are the same as in the first embodiment.

[0042] While it is possible to directly calculate the heart rate interval from the heart rate detected in step S110, such an estimate may be inaccurate. For example, if a heartbeat occurs at four different times, 0, 1, 2, and 3 seconds, the heart rate will be 60 beats / minute, and the heart rate interval will be 1 second. However, if a heartbeat occurs at four different times, 0, 0.7, 2.3, and 3.0 seconds, the heart rate will still be 60 beats / minute, but the heart rate intervals will be 0.7 seconds, 1.6 seconds, and 0.7 seconds, respectively. Therefore, an error will occur in the heart rate interval calculated directly from the heart rate (=1 second). Consequently, instead of directly calculating the heart rate interval from the heart rate, it is preferable to estimate the time interval TI from the heart rate using the correspondence between heart rate and time interval TI, and then use this time interval TI to determine the heart rate interval HRI.

[0043] The second embodiment achieves the same effects as the first embodiment. Furthermore, in the second embodiment, since heart rate is used to determine the time interval TI, the processing load can be reduced. In addition, even when it is difficult to detect the time interval TI from the heart sound signal HSS, the heart rate interval HRI can be accurately determined using the time interval TI estimated from the heart rate. Moreover, in the second embodiment, the determination of the time interval TI is faster than in the first embodiment, thus improving real-time performance.

[0044] This disclosure is not limited to the embodiments described above or their variations, and can be implemented in various forms without departing from its essence. [Explanation of symbols]

[0045] 100... Heart rate interval measurement system, 110... Heart sound sensor, 120... Heart sound band extraction unit, 130, 130a... Heart sound interval determination unit, 140... Template creation unit, 150... Cross-correlation processing unit, 160... Heart rate interval determination unit, 170... Display unit, 180... Heart rate sensor

Claims

1. A method for measuring heart rate intervals, (a) A step of extracting frequency components of the heart sound band from a heart sound signal including the first heart sound and the second heart sound, and generating a heart sound band waveform by integrating the amplitude components of the heart sound band at each time point, (b) A step of determining the time interval between the first heart sound and the second heart sound, (c) A step of determining the heart rate interval by detecting the interval between the appearance of peak pairs having the time interval in the heart sound band waveform, A method that includes this.

2. The method according to claim 1, The above step (a) is, The process involves performing frequency analysis on the aforementioned heart sound signal to obtain complex number information representing the amplitude and phase of multiple frequency bands for each sampling time, For each frequency band, the process involves extracting the amplitude component by determining the absolute value of the complex number information, A step of generating the heart sound band waveform by summing the amplitude components of the heart sound band at each sampling time, A method that includes this.

3. The method according to claim 1, The aforementioned step (c) is, A step of creating a template waveform that includes multiple peaks occurring at the aforementioned time intervals, A step of performing a cross-correlation process between the heart sound band waveform and the template waveform, and determining the heart rate interval using the interval between peaks in the result of the cross-correlation process, A method that includes this.

4. The method according to claim 1, The above step (b) is, A method comprising the step of determining the time interval using a histogram of peak intervals in the heart sound band waveform.

5. The method according to claim 1, The above step (b) is, The process of detecting heart rate, A step of estimating the time interval from the heart rate using a predetermined correspondence between the heart rate and the time interval, A method that includes this.

6. A system for measuring heart rate intervals, A heart sound sensor (110) that detects heart sound signals including the first and second heart sounds, A heart sound band extraction unit (120) extracts frequency components of the heart sound band from the heart sound signal and generates a heart sound band waveform by integrating the amplitude components of the heart sound band at each time point, A heart sound interval determination unit (130) that determines the time interval between the first heart sound and the second heart sound, A heart rate interval determination unit (160) detects the interval between the appearance of peak pairs having the time interval in the heart sound band waveform and determines the heart rate interval, A system equipped with these features.

7. The system according to claim 6, The aforementioned heart sound band extraction unit is The process involves performing frequency analysis on the aforementioned heart sound signal to obtain complex number information representing the amplitude and phase of multiple frequency bands for each sampling time, For each frequency band, the process involves extracting the amplitude component by calculating the absolute value of the complex number information, A process to generate the heart sound band waveform by summing the amplitude components of the heart sound band at each sampling time, A system configured to perform the following actions.

8. The system according to claim 6, A template creation unit (140) creates a template waveform that includes multiple peaks occurring at the aforementioned time intervals, A cross-correlation processing unit (1150) performs cross-correlation processing between the heart sound band waveform and the template waveform, Equipped with, The system is configured such that the heart rate interval determination unit determines the heart rate interval using the interval between peaks in the results of the cross-correlation processing.

9. The system according to claim 6, The system is configured such that the heart sound interval determination unit determines the time interval using a histogram of peak intervals in the heart sound band waveform.

10. The system according to claim 6, Equipped with a heart rate sensor (180) that detects heart rate, The system is configured such that the heart sound interval determination unit estimates the time interval from the heart rate using a pre-set correspondence between the heart rate and the time interval.

Images