Heart rate detection device and heart rate detection program

Abstract

To improve fastness, reduce an effect of disturbance, isolate heartbeat and respiration, prevent multiple counts, and expand a detection range while also making it possible to reduce a burden on nurses and reduce a risk of infection when the heart rate and the respiration rate are calculated on the basis of a radar signal (optionally including an ultrasonic signal) that is reflected off a body surface.SOLUTION: According to the present disclosure, heartbeat detection is performed by extracting a frequency component produced by heartbeat micro-vibrations over a prescribed heartbeat observation time window, extracting amplitude peaks from the frequency component produced by the heartbeat micro-vibrations, and extracting one characteristic micro-vibration from a plurality of characteristic micro-vibrations for one heartbeat. On the other hand, respiration detection is performed by extracting positive and negative frequency components produced by heartbeat micro-vibrations, performing complex multiplication of the positive and negative frequency components, and removing a heartbeat phase change produced by the heartbeat micro-vibrations from the positive and negative frequency components of which the complex multiplication was performed to extract a respiration phase change produced by respiration micro-vibrations.SELECTED DRAWING: Figure 4

Description

[Technical Field]

[0001] This disclosure relates to a technology for calculating heart rate and respiratory rate based on radar or ultrasound signals reflected from the body surface, while reducing the burden on nurses and lowering the risk of infection. [Background technology]

[0002] A technology for calculating heart rate and respiratory rate based on radar signals reflected from the body surface, while reducing the burden on nurses and lowering the risk of infection, is disclosed in Patent Document 1, etc.

[0003] Figure 1 shows an overview of the conventional heart rate and respiration detection process. A spectrogram S, which shows the time evolution of each frequency component, is calculated from the radar signal reflected from the body surface. A DC (Direct Current) component, including extremely low frequency components, is extracted from the spectrogram S. Amplitude peaks due to minute vibrations of respiration are extracted approximately once or twice every 10 seconds. The respiratory rate is calculated based on the time interval of the respiratory amplitude peaks. Amplitude peaks due to minute vibrations of the heartbeat are extracted approximately three times between the respiratory amplitude peaks. The heart rate is calculated based on the time interval of the heartbeat amplitude peaks. [Prior art documents] [Patent Documents]

[0004] [Patent Document 1] Japanese Patent Publication No. 2019-129996 [Overview of the project] [Problems that the invention aims to solve]

[0005] Figure 1 shows the extraction of the DC component, which includes extremely low-frequency components, from the spectrogram S. Disturbances such as air conditioner louvers, curtains swaying in the wind, and the movements of nurses are mainly contained in this extremely low-frequency component and are close in frequency to the frequency components due to minute vibrations of respiration, making it difficult to improve robustness and reduce the effects of disturbances. Furthermore, since the heart rate amplitude peak is smaller than the respiratory amplitude peak, it is difficult to separate the heart rate amplitude peak from the respiratory amplitude peak. Moreover, if the harmonic frequencies of the 3rd and 4th harmonics of the respiratory frequency are approximately equal to the heart rate frequency, it is difficult to separate the heart rate amplitude peak from the respiratory amplitude peak. Even if it is possible to separate the heart rate amplitude peak from the respiratory amplitude peak, there are still problems that will be explained in Figures 2 and 3.

[0006] Figure 2 illustrates the challenges of conventional heart rate detection processing. In a single heartbeat, there are amplitude peaks for the first and second heart sounds. Therefore, it is difficult to prevent the possibility of calculating a heart rate twice the actual heart rate (two heartbeats: (1) and (3) or (2) and (4)), resulting in four heartbeats ((1), (2), (3), and (4)). Furthermore, when heart rate A (bpm) is the target of detection, heart rate 2A (bpm) cannot be detected, making it difficult to expand the heart rate detection range.

[0007] Figure 3 illustrates the challenges of conventional respiratory detection processing. During a single breath, the movements of the chest and abdomen may be out of sync, but in the radar signal, the movements of the chest and abdomen are not distinguished and are combined. Therefore, it is difficult to prevent the possibility of calculating twice the actual number of breaths (two breaths: (1) and (2) of the chest or abdomen) (four breaths: (1), (2), (3), and (4) of the combined breath). Furthermore, when B (bpm) is the target of detection as the respiratory rate, 2B (bpm) cannot be detected as the respiratory rate, making it difficult to expand the detection range of the respiratory rate.

[0008] Therefore, in order to solve the aforementioned problems, this disclosure aims to improve robustness, reduce the influence of disturbances, separate heart rate and respiration, prevent multiple counts, and expand the detection range when calculating heart rate and respiratory rate based on radar signals (which may include ultrasound signals) reflected from the body surface, while reducing the burden on nurses and lowering the risk of infection. [Means for solving the problem]

[0009] To solve the aforementioned problems in heart rate detection, frequency components due to minute vibrations of the heart rate are extracted from radar signals or ultrasonic signals reflected from the body surface over a predetermined heart rate observation time window. Then, amplitude peaks are extracted from the frequency components due to minute vibrations of the heart rate, and one characteristic minute vibration is extracted from among several characteristic minute vibrations in a single heartbeat.

[0010] Specifically, the present disclosure is a heart rate and respiration detection device characterized by comprising: a heart rate component extraction unit that extracts frequency components due to minute vibrations of the heartbeat from radar signals or ultrasonic signals reflected from the body surface over a predetermined heart rate observation time window; a heart rate peak extraction unit that extracts amplitude peaks from the frequency components due to minute vibrations of the heartbeat and extracts one characteristic minute vibration from a plurality of characteristic minute vibrations in a single heartbeat; and a heart rate calculation unit that calculates the heart rate based on the time interval of the amplitude peaks or the number of amplitude peaks extracted within a predetermined time.

[0011] This configuration allows for the extraction of frequency components from minute heartbeat oscillations, thereby improving robustness and reducing the impact of external disturbances. Furthermore, by extracting one characteristic minute oscillation from among multiple characteristic minute oscillations in a single heartbeat, it is possible to prevent double counting and expand the detection range. In addition, since the data used to calculate heart rate differs from the data used to calculate respiratory rate (described later), it is possible to separate heartbeat and respiration.

[0012] Furthermore, this disclosure relates to a heart rate and respiration detection device characterized in that the heart rate component extraction unit extracts frequency components due to minute vibrations of the heart rate over a predetermined heart rate observation time window that includes a plurality of characteristic minute vibrations in a single heartbeat.

[0013] With this configuration, when extracting frequency components from minute heartbeat oscillations, multiple characteristic minute oscillations from a single heartbeat are synthesized and output, thus almost completely preventing double counting. However, depending on the time width of a predetermined heartbeat observation window, only one characteristic minute oscillation from a single heartbeat may be included in that window.

[0014] Furthermore, this disclosure relates to a heart rate and respiration detection device characterized in that the heart rate peak extraction unit extracts the maximum amplitude peak from the amplitude peaks among the frequency components due to minute vibrations of the heart rate within a predetermined peak detection time window.

[0015] With this configuration, even when only one characteristic minute oscillation in a single heartbeat is included in a predetermined heart rate observation time window, only the peak with the maximum amplitude is extracted within the predetermined peak detection time window, thus almost completely preventing multiple counting. However, the upper and lower limits of the heart rate will be narrowed to some extent depending on the time width of the predetermined peak detection time window.

[0016] Furthermore, this disclosure provides a heart rate and respiration detection device characterized in that the heart rate peak extraction unit moves the predetermined peak detection time window so as to extract the maximum amplitude peak in a time domain excluding the vicinity of both ends of the predetermined peak detection time window.

[0017] This configuration allows for an expansion of the heart rate upper limit, ensuring that a single characteristic micro-oscillation within a heartbeat does not span adjacent predetermined peak detection time windows.

[0018] Furthermore, this disclosure relates to a heart rate and respiration detection device characterized in that the heart rate calculation unit calculates the heart rate based on the time interval of the maximum amplitude peak extracted in adjacent or non-adjacent predetermined peak detection time windows.

[0019] This configuration allows for an expansion of the lower limit of the heart rate, taking into account that multiple characteristic minute oscillations within a single heartbeat may fall within a predetermined, discrete peak detection time window.

[0020] Furthermore, this disclosure provides a heart rate and respiration detection device characterized in that the heart rate calculation unit increases the weight of the time interval of the amplitude peak as the amplitude value of the amplitude peak increases, and then calculates the average heart rate.

[0021] This configuration allows for highly accurate calculation of heart rate by weighting the time interval of heart rate amplitude peaks according to the magnitude of the heart rate amplitude peaks.

[0022] Furthermore, this disclosure provides a heart rate and respiration detection device characterized in that the heart rate calculation unit increases the weight of the time interval of the amplitude peak as the amplitude value of the amplitude peak increases, and calculates the heart rate based on clustering of two-dimensional data consisting of the time interval of the amplitude peak and the weight of the time interval.

[0023] This configuration allows for highly accurate calculation of heart rate by calculating the weight of the time interval of the heart rate amplitude peak according to the magnitude of the heart rate amplitude peak.

[0024] To solve the aforementioned problems in respiration detection, positive and negative frequency components due to minute heartbeat vibrations are extracted from radar or ultrasonic signals reflected from the body surface, and these positive and negative frequency components are complex multiplied. Then, from the complex multiplied positive and negative frequency components, the respiratory phase change due to minute respiration vibrations is extracted, with the heartbeat phase change due to minute heartbeat vibrations removed.

[0025] Specifically, the present disclosure is a heart rate and respiration detection device characterized by comprising: a heart rate component multiplication unit that extracts positive and negative frequency components due to minute vibrations of the heartbeat from a radar signal or ultrasonic signal reflected from the body surface and complex multiplies the positive and negative frequency components; a respiratory phase extraction unit that extracts respiratory phase changes due to minute vibrations of respiration from the complex multiplied positive and negative frequency components, from which heart rate phase changes due to minute vibrations of the heartbeat have been removed; and a respiratory rate calculation unit that calculates the respiratory rate based on the frequency components of the respiratory phase change.

[0026] This configuration allows for the extraction of frequency components due to minute heart rate oscillations, thereby improving robustness and reducing the influence of external disturbances. Furthermore, by calculating the respiratory rate based on the frequency components of respiratory phase changes with heart rate phase changes removed, it is possible to prevent multiple counting and expand the detection range by considering the reflected signal from the chest without considering the reflected signal from the abdomen. In addition, since the data used to calculate the respiratory rate is different from the data used to calculate the heart rate (as described above), respiratory heart rate can be separated.

[0027] Furthermore, this disclosure provides a heart rate and respiration detection device characterized in that the respiratory phase extraction unit extracts amplitude peaks due to minute vibrations of the heartbeat from the complex multiplied positive and negative frequency components, extracts the respiratory phase change at the amplitude peaks, while applying zero padding between the amplitude peaks without extracting the respiratory phase change.

[0028] This configuration allows for highly accurate calculation of respiratory rate by taking into account heart rate variability and utilizing information on respiratory phase changes at the peak of heart rate amplitude, while not utilizing this information between peaks.

[0029] Furthermore, this disclosure relates to a heart rate and respiration detection device characterized in that the respiratory rate calculation unit increases the weight of the respiratory rate as the maximum peak amplitude of the frequency component of the respiratory phase change increases, and then calculates the average respiratory rate.

[0030] This configuration allows for highly accurate calculation of respiratory rate by weighting the respiratory rate according to the magnitude of the maximum peak amplitude of the frequency components of the respiratory phase change.

[0031] Furthermore, this disclosure is a heart rate and respiration detection program that causes a computer to execute each processing step corresponding to each processing unit of the heart rate and respiration detection device described above.

[0032] This configuration makes it possible to provide a program that has the aforementioned effects. [Effects of the Invention]

[0033] Thus, this disclosure enables improved robustness, reduced impact of disturbances, separation of heart rate and respiratory rate, prevention of multiple counts, and expansion of the detection range when calculating heart rate and respiratory rate based on radar signals (which may include ultrasound signals) reflected from the body surface, while reducing the burden on nurses and lowering the risk of infection. [Brief explanation of the drawing]

[0034] [Figure 1] This diagram shows an overview of conventional heart rate and respiration detection processing. [Figure 2] This figure shows the challenges of conventional heart rate detection processing. [Figure 3] This figure shows the challenges of conventional respiratory detection processing techniques. [Figure 4] This figure shows the configuration of the heart rate and respiration detection device disclosed herein. [Figure 5] This figure shows an overview of the heart rate and respiration detection process in this disclosure. [Figure 6] This figure shows the procedure for the heart rate detection process described herein. [Figure 7] This figure shows a specific example of the heart rate component extraction process described herein. [Figure 8] This figure shows a specific example of the first heart rate detection process of this disclosure. [Figure 9] This figure shows a specific example of the first heart rate detection process of this disclosure. [Figure 10]This figure shows a specific example of the first heart rate detection process of this disclosure. [Figure 11] This figure shows a specific example of the second heart rate detection process of this disclosure. [Figure 12] This figure shows a specific example of the third heart rate detection process of this disclosure. [Figure 13] This figure shows a specific example of the third heart rate detection process of this disclosure. [Figure 14] This figure shows the procedure for the respiratory detection process described herein. [Figure 15] This figure shows the principle of the respiration detection process of this disclosure. [Figure 16] This figure shows the principle of the respiration detection process of this disclosure. [Figure 17] This figure shows the principle of the respiration detection process of this disclosure. [Figure 18] This figure shows a specific example of the respiration detection process of this disclosure. [Figure 19] This figure shows the results of the heart rate detection process in this disclosure. [Figure 20] This figure shows the results of the heart rate detection process in this disclosure. [Figure 21] This figure shows the results of the respiratory detection process in this disclosure. [Figure 22] This figure shows the results of the respiratory detection process in this disclosure. [Figure 23] This figure shows the results of the respiratory detection process in this disclosure. [Modes for carrying out the invention]

[0035] Embodiments of the present disclosure will be described with reference to the attached drawings. The embodiments described below are examples of the implementation of the present disclosure, and the present disclosure is not limited to these embodiments.

[0036] (Configuration of the heart rate and respiration detection device in this disclosure) The configuration of the heart rate and respiration detection device of this disclosure is shown in Figure 4. An overview of the heart rate and respiration detection process of this disclosure is shown in Figure 5. The heart rate and respiration detection device M comprises a heart rate component extraction unit 1, a heart rate component multiplication unit 2, a heart rate peak extraction unit 3, a heart rate calculation unit 4, a respiration phase extraction unit 5, and a respiration rate calculation unit 6, and can be implemented by installing the heart rate and respiration detection program shown in Figures 6 and 14 onto a computer.

[0037] The radar transceiver R or ultrasound transceiver R transmits a radar signal or ultrasound signal (carrier band) to irradiate the surface of the patient P's body, receives the radar signal or ultrasound signal (carrier band) reflected from the surface of the patient P's body, and converts the received radar signal or ultrasound signal to the baseband band for output. The radar or ultrasound system may be CW, FMCW, standing wave, or any other system. The radar signal or ultrasound signal (carrier band) has a wavelength on the order of 1 to 10 mm, which is equal to the order of the minute vibration amplitude on the surface of the patient P's body.

[0038] The heart rate component extraction unit 1 extracts positive and negative frequency components (±10 or 10) from the radar signal or ultrasound signal (baseband I / Q complex signal) reflected from the body surface of the patient P, due to minute vibrations of the heartbeat. 2 From the order of Hz, one or both frequency components are extracted. Alternatively, the heart rate component extraction unit 1 extracts frequency components (+10 or 10) due to minute vibrations of the heart from the radar signal or ultrasound signal (baseband real signal) reflected from the body surface of the patient P. 2 Extracts the frequency components (on the order of Hz). The heart rate component multiplication unit 2 extracts the positive and negative frequency components (±10 or 10) due to minute vibrations of the heart rate. 2 Multiply by a complex number (of the order of Hz).

[0039] Here, the heart rate component extraction unit 1 calculates a spectrogram S showing the time evolution of each frequency component from the radar signal or ultrasound signal (baseband I / Q complex signal) reflected from the body surface of the patient P. Then, from the spectrogram S, the DC (Direct Current) component, which includes extremely low frequency components, is removed, and the positive and negative frequency components (±10 or 10) due to minute vibrations of the heartbeat are extracted. 2Extracts frequencies (on the order of Hz). Alternatively, the heart rate component extraction unit 1 calculates the bandpass filter result B for the frequency components due to minute vibrations of the heartbeat from the radar signal or ultrasound signal (baseband real signal) reflected from the body surface of patient P. Then, in the bandpass filter result B, the frequency components due to minute vibrations of the heartbeat (+10 or 10) are extracted, similar to the spectralgram S. 2 A frequency (on the order of Hz) is extracted.

[0040] In Figure 5, small amplitude peaks due to minute heartbeat oscillations are extracted approximately once or twice every two seconds, along with similar amplitude peaks for the first and second heart sounds in a single heartbeat. Then, large amplitude peaks due to minute respiratory oscillations are extracted approximately once or twice every four seconds, although the harmonic frequencies of the third and fourth times the respiratory frequency may be nearly equal to the heartbeat frequency.

[0041] Here, the heart rate component extraction unit 1 extracts (ri[n], rq[n]) (where r is the positive frequency, i and q are the i and q components, and n is time) as the positive frequency component due to minute vibrations of the heartbeat from the radar signal or ultrasound signal (baseband I / Q complex signal) reflected from the surface of the patient P's body, and extracts (li[n], lq[n]) (where l is the negative frequency, i and q are the i and q components, and n is time) as the negative frequency component due to minute vibrations of the heartbeat. Alternatively, the heart rate component extraction unit 1 extracts b[n] (where b is the passband frequency band of the bandpass filter result B, and n is time) as the frequency component due to minute vibrations of the heartbeat from the radar signal or ultrasound signal (baseband real signal) reflected from the surface of the patient P's body, Then, the heart rate component multiplication unit 2 calculates (mi[n], mq[n]) = (ri[n]·li[n]-rq[n]·lq[n], rq[n]·li[n]+ri[n]·lq[n]) as the complex multiplied positive and negative frequency components.

[0042] The heart rate peak extraction unit 3 and the heart rate calculation unit 4 calculate the heart rate based on frequency components (ri[n], rq[n]), (li[n], lq[n]), b[n], or (mi[n], mq[n]), as shown in Figures 6 to 13. The respiratory phase extraction unit 5 and the respiratory rate calculation unit 6 calculate the respiratory rate based on frequency components (mi[n], mq[n]), as shown in Figures 14 to 18. First, the calculation of the heart rate will be explained, and then the calculation of the respiratory rate will be explained.

[0043] (Procedure for heart rate detection processing in this disclosure: Heart rate component extraction process) The procedure for heart rate detection processing according to this disclosure is shown in Figure 6. The heart rate component extraction unit 1 extracts frequency components (ri[n], rq[n]), (li[n], lq[n]), or b[n] over a predetermined heart rate observation time window (step S1). The heart rate peak extraction unit 3 extracts amplitude peaks from the frequency components (ri[n], rq[n]), (li[n], lq[n]), b[n], or (mi[n], mq[n]) and extracts one characteristic minute oscillation from among several characteristic minute oscillations in a single heartbeat (step S2). The heart rate calculation unit 4 calculates the heart rate based on the time interval of the amplitude peaks (in this embodiment) or the number of amplitude peaks extracted within a predetermined time (modified example) (step S3). Specifically, the processes shown in Figures 7 to 13 are executed.

[0044] In this way, by extracting frequency components from minute vibrations of the heartbeat, robustness can be improved and the influence of external disturbances can be reduced. Furthermore, by extracting one characteristic minute vibration from among several characteristic minute vibrations in a single heartbeat, it is possible to prevent double counting and expand the detection range. In addition, since the data used to calculate heart rate is different from the data used to calculate respiratory rate (described later), heart rate and respiration can be separated.

[0045] In this embodiment, the first and second sounds of a single heartbeat are used as multiple characteristic micro-vibrations within a single heartbeat. However, as a modification, three or more characteristic micro-vibrations within a single heartbeat may be used depending on individual differences or animal species of the subject being detected for heart rate and respiration.

[0046] A specific example of the heart rate component extraction process of this disclosure is shown in Figure 7. The heart rate component extraction unit 1 extracts frequency components (ri[n], rq[n]), (li[n], lq[n]), or b[n] over a predetermined heart rate observation time window that includes multiple characteristic minute oscillations in a single heartbeat (step S1).

[0047] In Figure 7, the time widths of the heart rate observation time windows Wa, Wb, Wc, and Wd are the effective data length of the spectrogram S or 1 / (passbandwidth of the bandpass filter result B).

[0048] In the heart rate observation time window Wa, the first and second tones of a single heartbeat are included in phase (the phase relationship varies depending on the individual or animal species). Then, in the spectrogram S, the first and second tones of a single heartbeat are combined in a constructive interference state, and the frequency components (ri[n], rq[n]) and (li[n], lq[n]) are extracted. Alternatively, in the bandpass filter result B, the first and second tones of a single heartbeat are combined in a constructive interference state, and the frequency component b[n] is extracted. Therefore, at the point of extracting frequency components due to minute vibrations of the heartbeat, it is possible to prevent multiple counts almost completely.

[0049] In the heart rate observation time window Wb, the first and second tones of a single heartbeat are included in an out-of-phase state (the phase relationship varies depending on the individual or animal species). Then, in the spectrogram S, the first and second tones of a single heartbeat are combined in a destructive canceling state, and the frequency components (ri[n], rq[n]) and (li[n], lq[n]) are extracted. Alternatively, in the bandpass filter result B, the first and second tones of a single heartbeat are combined in a destructive canceling state, and the frequency component b[n] is extracted. However, as shown in Figures 9, 10, and 13, τ a , τ b , τ c By taking these factors into consideration, it is possible to almost completely prevent the miscounting of multiple entries.

[0050] In the heart rate observation time window Wc, only the first heartbeat is included (the number of minute vibrations varies depending on the heart rate and respiration conditions). Then, in the spectrogram S, the frequency components (ri[n], rq[n]) and (li[n], lq[n]) are extracted only for the first heartbeat. Alternatively, in the bandpass filter result B, the frequency component b[n] is extracted only for the first heartbeat. However, by applying a peak detection time window as shown in Figures 8 to 13, it is possible to almost completely prevent multiple counts.

[0051] In the heart rate observation time window Wd, only the second heart tone from a single heartbeat is included (the number of minute vibrations varies depending on the heart rate and respiration conditions). Then, in the spectrogram S, the frequency components (ri[n], rq[n]) and (li[n], lq[n]) are extracted only from the second heart tone from a single heartbeat. Alternatively, in the bandpass filter result B, the frequency component b[n] is extracted only from the second heart tone from a single heartbeat. However, by applying a peak detection time window as shown in Figures 8 to 13, it is possible to almost completely prevent multiple counts.

[0052] (Procedure for heart rate detection processing in this disclosure: First heart rate detection process) Specific examples of the first heart rate detection process of this disclosure are shown in Figures 8 to 10. The heart rate peak extraction unit 3 extracts the maximum amplitude peak from the frequency components (ri[n], rq[n]), (li[n], lq[n]), b[n] or (mi[n], mq[n]) within predetermined peak detection time windows W0, W1, W2, W3 (step S2).

[0053] In FIG. 8, within peak detection time windows W0, W1, W2, and W3 of a predetermined time width t, amplitude peaks of the first heart sound and the second heart sound in one heartbeat exist (the peak detection time window is slightly longer than the heartbeat observation time window). Within peak detection time window W0, the amplitude peak of the first heart sound has a peak time n0 and a maximum amplitude value p0. Within peak detection time window W1, the amplitude peak of the second heart sound has a peak time n1 and a maximum amplitude value p1. Within peak detection time window W2, the amplitude peak of the first heart sound has a peak time n2 and a maximum amplitude value p2. Within peak detection time window W3, the amplitude peak of the second heart sound has a peak time n3 and a maximum amplitude value p3.

[0054] The heart rate calculation unit 4 calculates an average heart rate (step S3) after increasing the weight of the time interval of the amplitude peak as the amplitude value of the amplitude peak increases. Alternatively, the heart rate calculation unit 4 calculates the heart rate based on the clustering of two-dimensional data consisting of the time interval of the amplitude peak and the weight of the time interval after increasing the weight of the time interval of the amplitude peak as the amplitude value of the amplitude peak increases (steps S3, FIG. 20).

[0055] The heart rate calculation unit 4 calculates the heart rate based on the time intervals of the maximum amplitude peaks extracted in adjacent or non - adjacent predetermined peak detection time windows (step S3). In FIG. 8, between peak detection time windows W0 and W1, the heart rate interval τ a is n1 - n0, and the weighting coefficient is √(p0p1). Between peak detection time windows W0 and W2, the heart rate interval τ b is n2 - n0, and the weighting coefficient is √(p0p2). Between peak detection time windows W0 and W3, the heart rate interval τ c is n3 - n0, and the weighting coefficient is √(p0p3).

[0056] In FIG. 9, the average heart rate intervals τ a、ave 、τ b、ave 、τ c、ave are calculated as shown in Formulas 1 - 3. Here, when the peak detection time window Wm (m = 0 to M - 1) is used as a reference, τ a =τ a[m], τ b =τ b [m], τ c =τ c [m], p0=p0[m], p1=p1[m], p2=p2[m], p3=p3[m]. And the average heart rate interval τ a、ave , τ b、ave , τ c、ave This is the average value from the beginning (m=0) to the end (m=M-1) of the heart rate detection period.

number

number

number

[0057] In Figure 10, amplitude peaks for the first and second heart sounds exist at the boundaries of peak detection time windows W0, W1, W2, and W3, each with a predetermined time width t (the peak detection time windows are slightly longer than the heart rate observation time windows). Within peak detection time window W0, the amplitude peak of the second heart sound has a peak time n0 and a maximum amplitude value p0. Within peak detection time window W1, the amplitude peak of the second heart sound has a peak time n1 (=t) and a maximum amplitude value p1. Within peak detection time window W2, the amplitude peak of the first first heart sound has a peak time n2 and a maximum amplitude value p2, and the amplitude peak of the second second heart sound has a peak time n2' and a second amplitude value p2'. Within peak detection time window W3, the amplitude peak of the second heart sound has a peak time n3 and a maximum amplitude value p3.

[0058] In Figure 10, the heart rate interval τ is measured between peak detection time windows W0 and W1. a n1-n0, and the weighting coefficient is √(p0p1) (heart rate interval τ a (This is short in duration and is therefore ignored, as described below). Between peak detection time windows W0 and W2, the heart rate interval τ b n²-n₀, the weighting coefficient is √(p₀p₂), and the heart rate interval τ b ' is n²'-n0, and the weighting coefficient is √(p0p²'). Between peak detection time windows W0 and W3, the heart rate interval τ c The formula is n³-n₀, and the weighting coefficient is √(p₀p₀).

[0059] Figure 10 shows the average heart rate interval τ. a、ave =Στ a √(p0p1) / Σ√(p0p1), τ b、ave =Στ b √(p0p2) / Σ√(p0p2), τ b、ave '=Στ b '√(p0p2') / Σ√(p0p2'), τ c、ave =Στ c √(p0p3) / Σ√(p0p3) is calculated (the sum is taken over the heart rate detection period). When the average heart rate is 60 / t (bpm), the heart rate interval τ is calculated. a , τ b , τb ', τ c The weight of the average heart rate interval τ a、ave , τ b、ave , τ b、ave ', τ c、ave It has a peak at the average heart rate interval τ b、ave Only the selected option is chosen. Alternatively, composite two-dimensional data may be created, as shown in the lower part of Figure 20 described later.

[0060] Thus, even when only one characteristic minute oscillation in a single heartbeat is included in a predetermined heart rate observation time window, by extracting only the maximum amplitude peak in a predetermined peak detection time window, it is possible to almost completely prevent multiple counting. However, the upper limit of 60 / t (bpm) and the lower limit of 60 / 3t (bpm) of the heart rate will be narrowed to some extent depending on the time width of the predetermined peak detection time window. Furthermore, by weighting the time interval of the heart rate amplitude peak according to the magnitude of the amplitude value of the heart rate amplitude peak, the heart rate can be calculated with high accuracy.

[0061] (Procedure for heart rate detection processing in this disclosure: Second heart rate detection process) A specific example of the second heart rate detection process of this disclosure is shown in Figure 11. The heart rate peak extraction unit 3 extracts amplitude peaks within a predetermined number or with an amplitude greater than or equal to a predetermined amount from frequency components (ri[n], rq[n]), (li[n], lq[n]), b[n], or (mi[n], mq[n]) in a predetermined peak detection time window W that is longer than a predetermined peak detection time window W0 to W3 (step S2).

[0062] In Figure 11, within a peak detection time window W (with a time width of 4t), amplitude peaks for the first and second heart sounds are present in four heartbeats. In chronological order within the peak detection time window W (with a time width of 4t), the amplitude peak of the first heart sound has a peak time n0 and a first amplitude value p0, the amplitude peak of the second heart sound has a peak time n1 and a third amplitude value p1, the amplitude peak of the first heart sound has a peak time n2 and a second amplitude value p2, and the amplitude peak of the second heart sound has a peak time n3 and a fourth amplitude value p3. When extracting amplitude peaks within a predetermined number or with a predetermined amplitude or greater, known QRS detection algorithms such as those by Engelse and Zeelenberg may be used.

[0063] The heart rate calculation unit 4 calculates the average heart rate by increasing the weight of the time interval of the amplitude peak as the amplitude value of the amplitude peak increases (step S3). Alternatively, the heart rate calculation unit 4 calculates the heart rate based on clustering of two-dimensional data consisting of the time interval of the amplitude peak and the weight of the time interval, by increasing the weight of the time interval of the amplitude peak as the amplitude value of the amplitude peak increases (step S3, Figure 20).

[0064] In Figure 11, in the peak detection time window W (time width 4t), the first adjacent heart rate interval τ is n1-n0, with a weighting coefficient of √(p0p1); the second adjacent heart rate interval τ is n2-n1, with a weighting coefficient of √(p1p2); and the third adjacent heart rate interval τ is n3-n2, with a weighting coefficient of √(p2p3).

[0065] Figure 11 shows the average heart rate interval τ. ave =Στ√(p n p n+1 ) / Σ√(p n p n+1 The following is calculated (the sum is taken over the heart rate detection period). When the average heart rate is 60 / t (bpm), the weight of the heart rate interval τ is the average heart rate interval τ ave It has a peak at [location]. Alternatively, composite two-dimensional data may be created as shown in the lower part of Figure 20, which will be described later.

[0066] Thus, even when only one characteristic minute oscillation in a single heartbeat is included in a predetermined heart rate observation time window, the upper and lower limits of the heart rate can be sufficiently expanded compared to Figures 8 to 10 by extracting amplitude peaks within a predetermined number or with an amplitude greater than or equal to a predetermined amplitude within a predetermined peak detection time window. However, depending on the predetermined number and amplitude, some degree of multiplicative counting will occur. Furthermore, by weighting the time interval of the heart rate amplitude peaks according to the magnitude of the amplitude value of the heart rate amplitude peaks, the heart rate can be calculated with high accuracy.

[0067] (Procedure for heart rate detection processing as disclosed herein: Third heart rate detection process) Specific examples of the third heart rate detection process of this disclosure are shown in Figures 12 and 13. The heart rate peak extraction unit 3 moves the predetermined peak detection time windows W0, W1, W2, and W3 so as to extract the maximum amplitude peak in the time domain excluding the vicinity of both ends of the predetermined peak detection time windows W0, W1, W2, and W3 (step S2). Short heart rate interval τ in Figure 10 a This is to prevent the occurrence of [unspecified problem].

[0068] In Figure 12, a peak detection time window W0 with a predetermined time width t is moved according to the amplitude peaks of the first and second heart sounds in a single heartbeat (the peak detection time window is slightly longer than the heart rate observation time window). In the first stage of Figure 12, the amplitude peak of the first heart sound has an amplitude value greater than a predetermined noise threshold and is selected as a candidate for the maximum amplitude peak. The peak detection time window W0 is then set so that the time n0 of the candidate for the maximum amplitude peak is at the center of the peak detection time window W0.

[0069] In the second stage of Figure 12, the amplitude peak of the second sound has an amplitude value greater than a predetermined noise threshold and an amplitude value greater than the previous candidate for the maximum amplitude peak, and is selected as a new candidate for the maximum amplitude peak. Then, the peak detection time window W0 is shifted by time n1-n0 so that the time n1 of the new candidate for the maximum amplitude peak is at the center of the peak detection time window W0.

[0070] In the third stage of Figure 12, the amplitude peak of the first sound has an amplitude value greater than a predetermined noise threshold and an amplitude value smaller than the candidate beyond the maximum amplitude peak, and is therefore not selected as a new candidate for the maximum amplitude peak. The peak detection time window W0 is then fixed so that the time n1 of the candidate beyond the maximum amplitude peak falls at the center of the peak detection time window W0.

[0071] In the fourth stage of Figure 12, the amplitude peak of the second tone has an amplitude value greater than a predetermined noise threshold, but it is not included in the peak detection time window W0. Therefore, a new peak detection time window W1 is shifted according to the amplitude peaks of the first and second tones in a single heartbeat. In other words, the amplitude peak of the second tone has an amplitude value greater than a predetermined noise threshold and is selected as a candidate for the maximum amplitude peak. Then, the peak detection time window W1 is set so that the time n3 of the candidate for the maximum amplitude peak is the time in the center of the peak detection time window W1. The same process is repeated thereafter.

[0072] The heart rate calculation unit 4 calculates the average heart rate by increasing the weight of the time interval of the amplitude peak as the amplitude value of the amplitude peak increases (step S3). Alternatively, the heart rate calculation unit 4 calculates the heart rate based on clustering of two-dimensional data consisting of the time interval of the amplitude peak and the weight of the time interval, by increasing the weight of the time interval of the amplitude peak as the amplitude value of the amplitude peak increases (step S3, Figure 20).

[0073] The heart rate calculation unit 4 calculates the heart rate based on the time interval of the maximum amplitude peak extracted in adjacent or non-adjacent predetermined peak detection time windows (step S3). In Figure 13, the heart rate interval τ is calculated between peak detection time windows W0 and W1.a n1-n0, and the weighting coefficient is √(p0p1). Between peak detection time windows W0 and W2, the heart rate interval τ b The formula is n²-n0, and the weighting coefficient is √(p0p²). Between the peak detection time windows W0 and W3, the heart rate interval τ c The formula is n³-n₀, and the weighting coefficient is √(p₀p₀).

[0074] Figure 13 shows the average heart rate interval τ. a、ave =Στ a √(p0p1) / Σ√(p0p1), τ b、ave =Στ b √(p0p2) / Σ√(p0p2), τ c、ave =Στ c √(p0p3) / Σ√(p0p3) is calculated (the sum is taken over the heart rate detection period). When the average heart rate is 60 / t (bpm), the heart rate interval τ is calculated. a , τ b , τ c The weight of the average heart rate interval τ a、ave , τ b、ave , τ c、ave It has a peak at the average heart rate interval τ a、ave Only the selected option is chosen. Alternatively, composite two-dimensional data may be created, as shown in the lower part of Figure 20 described later.

[0075] Thus, in order to prevent a single characteristic minute oscillation in a single heartbeat from spanning adjacent predetermined peak detection time windows, the short heartbeat interval τ in Figure 10 is used. a This method allows for an expansion of the upper limit of the heart rate without causing any noise. On the other hand, if a single characteristic minute oscillation in a single heartbeat is smaller than a predetermined noise threshold, it is not selected as a candidate for the maximum amplitude peak, thus expanding the lower limit of the heart rate. Furthermore, by weighting the time interval of the heart rate amplitude peak according to the magnitude of the amplitude value of the heart rate amplitude peak, the heart rate can be calculated with high accuracy.

[0076] In the first to third heartbeat detection processes, as the average heartbeat interval, τ ave = Στ√(p n p n+1 ) / Σ√(p n p n+1 ), even when calculating, it is not necessary to align all hierarchical values of τ. And after calculating the weight of the time interval of the amplitude peak, a weighted average of the two-dimensional data consisting of the time interval of the amplitude peak and the weight of the time interval is calculated. Therefore, the detection period of the heart rate can be shortened. Also, as the average heartbeat interval, τ n ←(1 - λ)τ n + λτ n-1 When calculating, the forgetting coefficient λ can be set according to √(p n-1 p n ), and compared with when calculating the weighted average of the two-dimensional data in the heart rate detection period, the calculation burden of τ n can be reduced. In addition to setting √(p n-1 p n ) as the weighting coefficient, max(p n-1 , p n ) can also be set, or min(p n-1 , p n ) can be set.

[0077] Here, when calculating the average heartbeat interval τ n ←(1 - λ)τ n + λτ n-1 , when the forgetting coefficient λ is made small (large), noise is easily removed (difficult to remove), but it is difficult (easy) to shorten the convergence time. And when applying the first-order IIR filter, the filter can be simplified compared to when applying the moving average filter and the FIR filter.

[0078] In this embodiment, the heart rate of 60 / τ (bpm) is calculated with high accuracy based on the time interval τ (seconds) of the amplitude peaks. As a variation, the heart rate of 60k / To (bpm) may be calculated more simply based on the number of amplitude peaks k (peaks) extracted within a predetermined time Tо (seconds). For example, if only 10 amplitude peaks are extracted during a 10-second measurement, a heart rate of 60 bpm can be calculated. However, if one extra amplitude peak is mistakenly extracted during a 10-second measurement, a heart rate of 66 bpm will be calculated, resulting in a +10% error. Therefore, using the time interval of the amplitude peaks is more accurate, while using the number of extracted amplitude peaks requires less computation.

[0079] (Procedure for respiratory detection processing as described herein) The procedure for the respiration detection process of this disclosure is shown in Figure 14. The heart rate component extraction unit 1 extracts frequency components (ri[n], rq[n]) and (li[n], lq[n]), and the heart rate component multiplication unit 2 calculates frequency components (mi[n], mq[n]) (step S4). The respiration phase extraction unit 5 extracts the respiration phase change due to minute vibrations of respiration from the frequency components (mi[n], mq[n]), with the heart rate phase change due to minute vibrations of the heartbeat removed (step S5). The respiration rate calculation unit 6 calculates the respiration rate based on the frequency components of the respiration phase change (e.g., Fourier transform components) (step S6). Specifically, the processes shown in Figures 15 to 18 are executed.

[0080] The principle of the respiration detection process of this disclosure is shown in Figures 15 to 17. In Figure 15, the radar signal or ultrasound signal (carrier band) reflected from the body surface of patient P is subjected to heart rate phase modulation by minute vibrations of the heartbeat, and also to respiratory phase modulation by minute vibrations of respiration. Here, the radar signal or ultrasound signal (carrier band) reflected from the body surface of patient P is also subjected to heart rate amplitude modulation, but this is less certain than heart rate phase modulation and is therefore unfavorable for high-precision respiration extraction.

[0081] Then, the heart rate component extraction unit 1 extracts Ae as a positive frequency component due to minute vibrations of the heartbeat. j{θr+(θp+φ)} (A is amplitude, θ)r θ is a change in respiratory phase. p (where is the heart rate phase change and φ is the initial phase of the heart rate phase change) is extracted, and Ae is used as the negative frequency component due to minute oscillating heartbeats. j{θr-(θp+φ)-π} (π is the phase difference between the upper and lower sidebands of the phase modulation) is extracted. Then, the heart rate component multiplication unit 2 takes the complex multiplied positive and negative frequency components as Ae j{θr+(θp+φ)} *Ae j{θr-(θp+φ)-π} =|A| 2 e j(2θr-π) The respiratory phase extraction unit 5 calculates the complex multiplication of the positive and negative frequency components |A|. 2 e j(2θr-π) Therefore, the heart rate phase change θ due to minute oscillating heartbeats. p +φ removed, respiratory phase change θ due to minute vibrations of respiration r It is possible to extract it.

[0082] In Figure 16, the heart rate component multiplication unit 2 multiplies the positive and negative frequency components by complex multiplication, and |A| 2 e j(2θr-π) The simulation results are shown before the calculation of θ. In the left column of Figure 16, the respiratory phase change θ is shown. r However, as it changes from 0 to π / 2, the heart rate phase change θ p +φ superimposes minute changes. In the upper right column of Figure 16, the respiratory phase change θ r The changes in θ are the main feature observed, and in the lower right section of Figure 16 (a magnified view of a part of the time domain), the respiratory phase change θ is observed. r Changes in this can be observed, as well as changes in heart rate phase θ. p The changes in +φ are also clearly visible.

[0083] In Figure 17, the heart rate component multiplication unit 2 multiplies the positive and negative frequency components by complex multiplication, and |A| 2 e j(2θr-π) The simulation results are shown after the calculation of the following. In the left column of Figure 17, the respiratory phase change 2θ is shown. r As -π changes from -π to 0, the heart rate phase change θ p +φ does not superimpose minute changes (although changes in amplitude are observed, the phase change is monotonous). In the upper right column of Figure 17, the respiratory phase change 2θr The changes in -π are the main feature observed, and in the lower right section of Figure 17 (a magnified view of a part of the time domain), the respiratory phase change 2θ is observed. r Although a change in -π can be observed, the heart rate phase change θ p There is hardly any noticeable change in +φ.

[0084] In this way, by extracting frequency components from minute vibrations of the heartbeat, robustness can be improved and the influence of external disturbances can be reduced. Furthermore, by calculating the respiratory rate based on the frequency components of respiratory phase changes from which heartbeat phase changes have been removed, it is possible to prevent multiple counts and expand the detection range while considering the reflected signal from the chest without considering the reflected signal from the abdomen. In addition, since the data used to calculate the respiratory rate is different from the data used to calculate the heart rate (as described above), respiratory heartbeat can be separated.

[0085] A specific example of the respiration detection process of this disclosure is shown in Figure 18. The respiration phase extraction unit 5 complex multiplies the positive and negative frequency components |A| 2 e j(2θr-π) From this, amplitude peaks due to minute heart rate oscillations are extracted, and the respiratory phase change 2θ is observed at the amplitude peaks. r -π is extracted, while the respiratory phase change 2θ occurs between amplitude peaks. r -π is not extracted and zero padding is applied (step S5). The respiratory rate calculation unit 6 calculates the respiratory phase change 2θ r As the maximum peak amplitude of the -π frequency component increases, the weight given to the respiratory rate is increased, and the average respiratory rate is calculated (step S6).

[0086] In Figure 18, at peak time n0, an amplitude peak exists, and the respiratory phase change 2θ r -π=0 is extracted. At peak time n1, an amplitude peak exists, and the respiratory phase change 2θ r -π=π / 2 is extracted. At peak time n2, an amplitude peak exists, and the respiratory phase change 2θ r -π=π is extracted. At peak time n3, an amplitude peak exists, and the respiratory phase change 2θ r-π=π is extracted. At peak time n4, an amplitude peak exists, and the respiratory phase change 2θ r -π=π / 2 is extracted. At peak time n5, an amplitude peak exists, and the respiratory phase change 2θ r -π=π / 2 is extracted. Between peak times n0, n1, n2, n3, n4, and n5, there are no amplitude peaks, and zero padding I=Q=0 is applied.

[0087] Figure 18 shows the respiratory phase change 2θ. r Based on the maximum peak frequency of the -π frequency component, the respiratory frequency and thus the respiratory rate are calculated. Then, the respiratory phase change 2θ r A weighting coefficient for respiratory rate is calculated based on the maximum peak amplitude of the -π frequency component. Here, as with calculating heart rate, when calculating respiratory rate, the weight of respiratory rate may be calculated according to the maximum peak amplitude, and the forgetting coefficient λ may be set according to the weighting coefficient.

[0088] Thus, by taking into account heart rate variability, and utilizing information on respiratory phase changes at the heart rate amplitude peak time, while not utilizing respiratory phase change information between heart rate amplitude peaks, the respiratory rate can be calculated with high accuracy. Here, the respiratory phase change θ r Instead of performing frequency conversion, it performs respiratory phase change 2θ r To perform a frequency conversion of -π (2θ r is θ r Compared to [another method], the maximum peak frequency of the frequency component can be calculated with high accuracy (it is doubled). Furthermore, because the respiratory rate is weighted according to the magnitude of the maximum peak amplitude of the frequency component of the respiratory phase change, the respiratory rate can be calculated with high accuracy.

[0089] In this embodiment, the respiratory phase change 2θ due to minute vibrations of respiration is obtained from the frequency components (mi[n], mq[n]). r -π is extracted. Simultaneously, from the frequency components (mi[n], mq[n]), the respiratory amplitude change due to minute vibrations of respiration |A| is calculated. 2It also extracts the positive and negative frequency components |A| obtained by complex multiplication. 2 e j(2θr-π) It is calculating this.

[0090] As a variation, one of the respiratory phase changes and respiratory amplitude changes due to minute vibrations of respiration may be extracted from the frequency components (mi[n], mq[n]). Alternatively, one or both of the respiratory phase changes and respiratory amplitude changes due to minute vibrations of respiration may be extracted from the frequency components (ri[n], rq[n]), (li[n], lq[n]), and b[n].

[0091] (Results of the heart rate and respiration detection process in this disclosure) The results of the heart rate detection process of this disclosure are shown in Figures 19 and 20. In the upper part of Figure 19, the amplitude P[n] = √(mi[n]) due to minute vibrations of the heart rate is shown. 2 +mq[n] 2 The time evolution of ) is shown. In the lower panel of Figure 19, the amplitude P[n] = √(mi[n] due to minute oscillating heartbeats is shown. 2 +mq[n] 2 This shows the peak time. Here, the processes shown in Figures 8 and 9 are executed.

[0092] In the left column of Figure 20, the heart rate interval τ a The weights are shown. In the middle column of Figure 20, the heart rate interval τ b The weights are shown. In the right column of Figure 20, the heart rate interval τ c The weights are shown. Here, the heart rate interval 0 < τ a <50 and heart rate interval 150<τ c At <200, the upper and lower limits of the heart rate interval are exceeded and are therefore ignored. And the heart rate interval τ a ~90 and heart rate interval τ b At a value of ~90, clusters based on heart rate interval weights are extracted. Therefore, a heart rate of 61 bpm can be calculated. Note that when the weight is low, the detection result may be ignored, or a warning may be issued. Furthermore, k-means algorithm or DBSCAN may be applied when extracting clusters.

[0093] In the lower section of Figure 20, unlike the left, middle, and right sections of Figure 20, the heart rate interval τ is shown. a , τ b , τ c Without showing the weights separately, the heart rate interval τ a , τ b , τ c The weights are combined and shown. Here, clusters are extracted from this 2D data, and outliers are removed. Then, a weighted average of heart rate intervals is calculated for these clusters, resulting in higher accuracy of heart rate.

[0094] The results of the respiratory detection process of this disclosure are shown in Figures 21 and 22. In the upper part of Figure 21, the amplitude P[n] = √(mi[n]) due to minute oscillating heartbeats is shown. 2 +mq[n] 2 The time evolution of ) is shown. The lower panel of Figure 21 shows the respiratory phase change 2θ of the amplitude peak. r This shows that [n]-π=360·(4f / c)r[n] (where r[n] is the distance). Here, the processes shown in Figures 15 and 18 are performed.

[0095] The left column of Figure 22 shows the constellation of amplitude peaks. The right column of Figure 22 shows the respiratory phase change 2θ. r The frequency transformation of [n]-π is shown. Here, the respiratory phase change 2θ is observed from n=38 through n=131 to n=219. r A monotonic change of [n]-π is detected. Furthermore, the maximum peak frequency of the frequency transformation is calculated at a frequency component of 0.15Hz. Therefore, a respiratory rate of 9 bpm can be calculated.

[0096] The results of the respiratory detection process described herein are also shown in Figure 23. In Figure 23, the amplitude P[n] = √(mi[n]) due to minute oscillating heartbeats is shown. 2 +mq[n] 2 This shows the time evolution of ). Here, the amplitude P[n] due to minute oscillating heartbeats forms a very regular peak. In this case, the respiratory amplitude change due to minute oscillating respiration |A| 2Even simply extracting this information allows for a reasonably accurate calculation of the respiratory rate. However, the amplitude P[n] due to minute fluctuations in the heartbeat does not necessarily form regular peaks. In this case as well, the respiratory phase change 2θ due to minute fluctuations in respiration. r By extracting -π, the respiratory rate can be calculated with high accuracy regardless of the circumstances. [Industrial applicability]

[0097] The heart rate and respiratory rate detection device and heart rate and respiratory rate detection program of this disclosure can calculate heart rate and respiratory rate based on radar signals (which may include ultrasonic signals) reflected from the body surface, while reducing the burden on nurses and lowering the risk of infection. [Explanation of symbols]

[0098] S: Spectrogram B: Bandpass filter result P:Patient R: Radar transceiver, ultrasonic transceiver M: Heart rate and respiration detection device Wa, Wb, Wc, Wd: Heart rate observation time window W0, W1, W2, W3: Peak detection time window W: Peak detection time window 1: Heart rate component extraction unit 2: Heart rate component multiplication section 3: Heart rate peak extraction section 4: Heart rate calculation unit 5:Respiration phase extraction part 6:Respiration rate calculation part

Claims

1. A heart rate component extraction unit extracts frequency components due to minute vibrations of the heartbeat from radar signals or ultrasonic signals reflected from the body surface over a predetermined heart rate observation time window, A heart rate peak extraction unit extracts amplitude peaks from the frequency components due to the minute vibrations of the heartbeat, and extracts one characteristic minute vibration from among several characteristic minute vibrations in a single heartbeat. A heart rate calculation unit calculates the heart rate based on the time interval of the amplitude peaks or the number of amplitude peaks extracted within a predetermined time period, A heart rate detection device characterized by comprising the following features.

2. The heart rate component extraction unit extracts frequency components due to the minute oscillations of the heart rate over a predetermined heart rate observation time window that includes multiple characteristic minute oscillations in a single heartbeat. The heart rate detection device according to claim 1, characterized in that...

3. The heart rate peak extraction unit extracts the maximum amplitude peak from the amplitude peaks among the frequency components due to minute vibrations of the heart rate within a predetermined peak detection time window. A heart rate detection device according to claim 1 or 2, characterized in that...

4. The heart rate peak extraction unit moves the predetermined peak detection time window so as to extract the maximum amplitude peak in a time domain excluding the vicinity of both ends of the predetermined peak detection time window. The heart rate detection device according to claim 3, characterized in that...

5. The heart rate calculation unit calculates the heart rate based on the time interval of the maximum amplitude peak extracted in adjacent or non-adjacent predetermined peak detection time windows. A heart rate detection device according to claim 3 or 4, characterized in that it is a heart rate detection device according to claim 3 or 4.

6. The heart rate calculation unit calculates the average heart rate by increasing the weight of the time interval of the amplitude peak as the amplitude value of the amplitude peak increases. A heart rate detection device according to any one of claims 1 to 5, characterized in that

7. The heart rate calculation unit increases the weight of the time interval of the amplitude peak as the amplitude value of the amplitude peak increases, and then calculates the heart rate based on clustering of two-dimensional data consisting of the time interval of the amplitude peak and the weight of the time interval. A heart rate detection device according to any one of claims 1 to 5, characterized in that

8. A heart rate detection program for causing a computer to execute each processing step corresponding to each processing unit of a heart rate detection device according to any one of claims 1 to 7.

Images