JP3924636B2 - Pulse wave signal processing method - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
  The present inventionPulse waveThe present invention relates to a signal processing method, and more particularly to pulse wave noise removal that can be used in oxygen saturation measurement or the like by pulse photometry.
[0002]
[Prior art]
  Conventionally, a pulse oximeter is used to continuously measure the oxygen saturation of arterial blood in a noninvasive manner. In this pulse oximeter, the probe is attached to the fingertip or earlobe of the subject, and the living body is irradiated with light of different wavelengths of red and infrared from the probe in a time-sharing manner, and obtained from transmitted light or reflected light of two different wavelengths. The oxygen saturation S is measured from the ratio of the pulsation component of absorbance. A reference wavelength of, for example, 660 nm is used for red light, and a wavelength of, for example, 940 nm is used for infrared light, and two light emitting diodes emitting these wavelengths and one photodiode for receiving light are incorporated in the probe. Has been. Now, assuming that the pulsation component of the absorbance at the wavelength of red light is ΔA1, and the pulsation component of the absorbance at the wavelength of infrared light is ΔA2, the ratio of the absorbance at two different wavelengths is given by the following equation.
  Ф = ΔA1 / ΔA2
  The oxygen saturation S can be calculated as a function f of this absorbance ratio.
  S = f (Ф)
[0003]
  In such a pulse oximeter, if body motion or the like occurs in the patient during measurement, a noise component is mixed into the pulse wave detected by the probe, and the oxygen saturation S cannot be measured accurately. Thus, various attempts have been made to remove the influence of such noise.
[0004]
  Japanese Patent Application Laid-Open No. 2-172443 describes a technique for detecting a pulse in an oximeter and performing frequency filtering on the photoelectric pulse wave signal based on the pulse frequency. Here, as a specific pulse wave number detection method, the output of the light receiving unit is logarithmically converted, the pulse wave signal obtained from the logarithmically converted output is binarized, and the predetermined number of binarized pulses are further converted. It is described that the period of the wave signal is obtained and the pulse rate is obtained from the reciprocal of the period. Furthermore, frequency filtering is performed by using a bandpass filter with a low selection frequency when the pulse rate is small, a bandpass filter with a medium selection frequency when the pulse rate is medium, and a selection frequency when the pulse rate is high. It is described that a high bandpass filter is used for each.
[0005]
  However, such an invention has the following problems.
  <Problem of binarization>
  When body motion occurs, the pulse wave signal is mixed with noise caused by body motion and buried. If the pulse wave signal mixed with this body motion noise is logarithmically calculated, the ratio of the logarithmically calculated body motion component to the logarithmically calculated pulse wave component is the pulse wave component before the logarithmic calculation of the body motion component before the logarithmic calculation. Smaller than the ratio to. That is, in the above-described conventional technique, a pulse wave signal mixed with logarithmically calculated body motion noise is distributed to 1 or 0 with a certain threshold value by binary calculation. However, since the noise component is not necessarily smaller than the pulse wave component, there is a strong possibility that the noise component is distributed to 1 by binarization while being a noise component. In this case, the pulse rate cannot be detected with high accuracy.
  <Problems of bandpass filter>
  Only by using a band pass filter to remove noise, when the noise component is in the pass band, the noise is mixed into the pulse wave. That is, the pulse wave cannot be extracted well.
[0006]
[Problems to be solved by the invention]
  The present invention has been made in order to solve the above-mentioned problems, and its purpose is (1) calculating the fundamental frequency of a signal with high accuracy, and (2) using the fundamental frequency and the frequency of the high frequency as a filtering frequency. The signal processing is used to extract a characteristic frequency component of the signal to form a signal. In particular, in the application to the measurement of arterial oxygen saturation, even if the pulse wave is superimposed with noise due to body motion or the pulse wave is fluctuating, the basic pulse rate is accurate. By measuring the frequency and processing the pulse wave signal using the fundamental frequency and the harmonic frequency as the filtering frequency, a characteristic frequency component is extracted from the pulse wave on which the noise is superimposed, and the pulse wave signal is extracted. It can form, remove noise, and reproduce the original pulse wave. Another object of the present invention is to provide an apparatus for accurately measuring oxygen saturation by pulse photometry from a pulse wave from which noise has been removed.
[0007]
[Means for Solving the Problems]
  According to claim 1 of the present inventionPulse waveThe signal processing method processes a continuous first signal IR and second signal RD having the same fundamental frequency.Pulse waveIn the signal processing method,
  For each of the first signal IR and the second signal RD,Constant based on the number of samples of the pulse wave signalIn a period, a frequency spectrum or a frequency power spectrum is calculated, and a first spectrum Spc.IR of the first signal IR and a second spectrum Spc.RD of the second signal RD are calculated.And eachSpectrum calculation step to be calculated;
  A spectrum that takes a difference between the first spectrum Spc.IR and the second spectrum Spc.RD calculated in the spectrum calculation step on the frequency axis.Is calculatedOr for calculating the spectrum of differences between each otherin additionFurther normalization spectrumAnd thatCalculating a fundamental frequency of the first signal IR and the second signal RD based on a result; and
  In the fundamental frequency calculation step,
(Spc.IR−Spc.RD) /Spc.RD,
(Spc.IR−Spc.RD) /Spc.IR,
(Spc.IR−Spc.RD) / (Spc.IR + Spc.RD),
{1- (Spc.RD / Spc.IR)} / (Spc.RD / Spc.IR),
{1- (Spc.RD / Spc.IR)} / (Spc.IR/Spc.IR), or
{1- (Spc.RD / Spc.IR)} / {1+ (Spc.RD / Spc.IR)}
The fundamental frequency is calculated based on the calculation result of one of the spectra.
[0008]
  Of the present inventionClaim 2Pertaining toPulse waveThe signal processing method processes a continuous first signal IR and second signal RD having the same fundamental frequency.Pulse waveIn the signal processing method,
  For each of the first signal IR and the second signal RD,Constant based on the number of samples of the pulse wave signalIn a period, a frequency spectrum or a frequency power spectrum is calculated, a first spectrum Spc.IR is calculated from the first signal IR, and a second spectrum Spc.RD is calculated from the second signal RD.,RespectivelySpectrum calculation step to be calculated;
  The first spectrum Spc.IR and the second spectrum Spc.RD by the spectrum calculation stepMultiple timesSpectral addition average calculation step for averaging,
  Using the first spectrum Av.Spc.RD added and averaged in the spectrum addition averaging step and the second spectrum Av.Spc.RD added and averaged, spectra that take a difference on the frequency axisIs calculatedOr for calculating the spectrum of differences between each otherin additionFurther normalization spectrumAnd thatAnd a fundamental frequency calculating step of calculating a fundamental frequency of the first signal IR and the second signal RD based on the result.
[0009]
  Of the present inventionClaim 3Pertaining toPulse waveThe signal processing method isClaim 2Described inPulse waveIn the signal processing method,
  In the fundamental frequency calculation step,
(Av.Spc.IR−Av.Spc.RD) /Av.Spc.RD,
(Av.Spc.IR−Av.Spc.RD) /Av.Spc.IR,
(Av.Spc.IR−Av.Spc.RD) / (Av.Spc.IR + Av.Spc.RD),
{1- (Av.Spc.RD / Av.Spc.IR)} / (Av.Spc.RD / Av.Spc.IR),
{1- (Av.Spc.RD / Av.Spc.IR)} / (Av.Spc.IR/Av.Spc.IR), or
{1- (Av.Spc.RD / Av.Spc.IR)} / {1+ (Av.Spc.RD / Av.Spc.IR)}
The fundamental frequency is calculated based on the calculation result of one of the spectra.
[0010]
  Of the present inventionClaim 4Pertaining toPulse waveA signal processing method comprises:One of 3Described inPulse waveIn the signal processing method,
  Furthermore, a filter forming step of forming a filter using the fundamental frequency calculated in the fundamental frequency calculating step and its harmonics, and
  A filtering step of filtering at least the first signal IR or the second signal RD by a filter formed by the filter forming step.
[0011]
  Of the present inventionClaim 5Pertaining toPulse waveSignal processing method is claimed4Described inPulse waveIn the signal processing method, when there are a plurality of the fundamental frequencies calculated in the fundamental frequency calculating step, the filter forming step forms a filter using each fundamental frequency and each harmonic frequency. Features.
[0012]
  Of the present inventionClaim 6Pertaining toPulse waveSignal processing method is claimed4 or 5Described inPulse waveIn the signal processing method, in the filter forming step, the characteristic of the filter is a rectangular shape having a width centered on the fundamental frequency and the harmonic frequency.
[0013]
  Of the present inventionClaim 7Pertaining toPulse waveSignal processing method is claimed4 or 5Described inPulse waveIn the signal processing method, in the filter forming step, the filter is a Gaussian characteristic filter centered at the fundamental frequency and the harmonic frequency thereof.
[0014]
  Of the present inventionClaim 8Pertaining toPulse waveA signal processing method comprises:Or 7 EitherDescribed inPulse waveIn the signal processing method, the first signal IR and the second signal RD are both pulse wave signals.
[0015]
  Of the present inventionClaim 9Pertaining toPulse waveSignal processing method is claimed4 or 7 EitherDescribed inPulse waveIn the signal processing method, each of the first signal IR and the second signal RD is a pulse wave signal, and further, one of the first signal IR and the second signal RD that is filtered in the filtering step. The step of displaying a signal of:
[0016]
  Of the present inventionClaim 10Pertaining toPulse waveSignal processing method is claimed4 or 7 EitherDescribed inPulse waveIn the signal processing method,
  The first signal IR is a pulse wave signal obtained by infrared light transmitted or reflected through a living artery,
  The second signal RD is a pulse wave signal obtained by red light transmitted or reflected through a living artery,
  Filtering both the first signal IR and the second signal RD in the filtering step;
  Further, the method further includes an oxygen saturation calculation step of calculating an oxygen saturation using the first signal IR or the second signal RD filtered in the filtering step.
[0017]
  Of the present inventionClaim 11The pulse wave signal processing method according to the above is a pulse wave signal processing method for processing a first pulse wave signal and a second pulse wave signal of a living body having the same fundamental frequency.
  A first pulsation component detection step of detecting a pulsation component ΔA1 of the first pulse wave signal;
  A second pulsation component detection step of detecting a pulsation component ΔA2 of the second pulse wave signal;
  A pulsation component calculation step of calculating a ratio Φ of the pulsation component of the first pulse wave signal and the pulsation component of the second pulse wave signal;
  Constant based on the number of samples of the pulse wave signalTo calculate the frequency spectrum or frequency power spectrum in the periodMore than oncePerforming spectral computation steps;
  SaidMore than onceSpectral addition average calculation step of adding and averaging the spectrum calculated in minutes;
  And a fundamental frequency calculating step of calculating a fundamental frequency of the pulse wave signal from the summed and averaged spectrum.
[0018]
  Of the present inventionClaim 12The pulse wave signal processing method according to claim 11, wherein the pulse wave signal processing method according to claim 11,
  A filter forming step of forming a filter using the fundamental frequency calculated by the fundamental frequency calculating step and a frequency of its harmonics;
  A filtering step of filtering at least the first pulse wave signal or the second pulse wave signal with the filter formed by the filter forming step.
[0019]
  Of the present inventionClaim 13The pulse wave signal processing method according to claim12In the pulse wave signal processing method according to claim 1, when there are a plurality of the fundamental frequencies calculated by the fundamental frequency calculating step, the filter forming step uses a filter using each fundamental frequency and each harmonic frequency thereof. It is characterized by forming.
[0020]
  Of the present inventionClaim 14The pulse wave signal processing method according to claim12 or 13In the pulse wave signal processing method described in (1), in the filter forming step, the characteristic of the filter is a rectangular shape having a width centered on the fundamental frequency and the harmonic frequency.
[0021]
  Of the present inventionClaim 15The pulse wave signal processing method according to claim12 or 13In the pulse wave signal processing method described in (1), in the filter forming step, the filter is a Gaussian characteristic filter centered on the fundamental frequency and the harmonic frequency.
[0022]
  Of the present inventionClaim 16Pertaining toPulse waveSignal processing method is claimedAny of 12-15Described inPulse waveIn the signal processing method, further, in the filtering step, the firstPulse waveSignal or said secondPulse waveOf the filtered signalPulse waveAnd displaying the signal.
[0023]
  Of the present inventionClaim 17Pertaining toPulse waveSignal processing method is claimedAny of 12-15Described inPulse waveIn the signal processing method,
  The first pulse wave signal is a pulse wave signal obtained by red light transmitted or reflected through a living artery,
  The second pulse wave signal is a pulse wave signal obtained by infrared light transmitted or reflected through a living artery,
  Filtering both the first pulse wave signal and the second pulse wave signal in the filtering step;
  The method further includes an oxygen saturation calculation step of calculating an oxygen saturation using the first pulse wave signal or the second pulse wave signal filtered in the filtering step.
[0024]
DETAILED DESCRIPTION OF THE INVENTION
  Embodiments of a pulse wave signal processing and a pulse oximeter using the pulse wave signal processing according to the present invention will be described below in detail with reference to the drawings.
[0025]
  FIG. 1 shows a device configuration of a pulse oximeter as an embodiment of the present invention. The probe 1 includes a light emitting unit 2 and a light receiving unit 3, and is configured to sandwich a fingertip (living tissue) 4 by these. The light emitting unit 2 includes two light emitting diodes that respectively emit red light (wavelength λ1: 660 nm) that is first wavelength light and infrared light (wavelength λ2: 940 nm) that is second wavelength light. The light emitting unit 2 is driven by the light emitting unit driving circuit 5, and red light and infrared light are emitted alternately. The light receiving unit 3 includes a photodiode, receives light transmitted through the fingertip, and outputs an electrical signal corresponding to the transmitted light intensity. The output signal of the light receiving unit 3 is amplified by the received light signal amplification circuit 6 and demodulated by the demodulation circuit 7. The demodulating circuit 7 outputs the signals corresponding to the red light and the infrared light separately. These signals are amplified by amplifiers 9a and 9b, converted to digital signals by A / D converters 10a and 10b, and input to a CPU (Central Processing Unit) 8. The CPU 8 controls the demodulation circuit 7 and the light emitting unit drive circuit 5, processes signals given from the A / D converters 10 a and 10 b, and outputs the result to the display unit 11. The present invention is characterized by the pulse wave noise removal processing in this signal processing. The result output to the display unit is a pulse wave waveform, pulse rate, and SpO2 value (oxygen saturation) from which noise has been removed.
[0025]
  FIG. 2 shows a rough processing flow of the apparatus. When the measurement is started (S10), red light and infrared light emitted from the light emitting unit 2 of the probe 1 are alternately emitted, and red light and infrared light transmitted through the living tissue 4 are alternately received by the light receiving unit 3. Is detected. Then, it is amplified by the received light signal amplifier circuit 6, separated into red light and infrared light by the demodulator circuit 7, and A / D converted by the A / D converters 10a and 10b. Thereby, the pulse wave is detected (S11). Next, for each pulse wave of infrared light and red light, Fourier transform for a predetermined period is performed to obtain the absolute value of each spectrum and the value of the power spectrum. Although a method for obtaining a power spectrum is not described in detail in this specification, it can be obtained by Fourier transform after obtaining autocorrelation of infrared light and red light for a predetermined period. Then, the fundamental frequency of the pulse wave is calculated based on the spectrum calculation for infrared light and red light (S12). This corresponds to the pulse frequency. And the matched filter which permeate | transmits the fundamental frequency of a pulse wave and the frequency of the harmonic is formed (S13). The pulse wave is filtered using this matched filter (S14). The filtered pulse wave is displayed on the display unit 11, and the pulse and the SpO 2 value (oxygen saturation) obtained from the filtered pulse wave are also displayed on the display unit 11.
[0026]
  This series of processing is continuously repeated. Hereinafter, it will be described in detail from “Calculation of fundamental frequency of pulse wave” S12 shown in FIG.
[0027]
  First embodiment
  A first embodiment will be described with reference to FIGS. The pulse waves for infrared light (IR) and red light (D) each have a sampling interval of 16 msec. In the uppermost stage shown in FIG. 3, infrared light (IR) and red light (RD) including noise are shown.
  The period [i] is a period corresponding to the number of 1024 samples and is 16.384 seconds (= 1024 * 16 msec). First, Fourier transform is performed using 1024 data samples of infrared light (IR) and red light (RD) in the period [i], and the absolute value of the spectrum is obtained. This spectrum is shown in FIG. 4 (a) and FIG. 4 (b). FIG. 4A shows an infrared light (IR) spectrum (Spc.IR), and FIG. 4B shows a red light (RD) spectrum (Spc.RD).
[0028]
  As can be seen, until the pulse wave is Fourier-transformed to obtain the spectrum, the spectrum of the noise component is large and it is difficult to detect the fundamental frequency of the pulse wave. Therefore, it has been found that the fundamental frequency of the pulse wave can be easily detected by performing any of the following calculations (1) and (3). The calculation here is the sum / quotient calculation of each spectrum for the same frequency. What is common in the operations of (1) to (3) is to use the difference between the spectrum of infrared light (IR) (Spc.IR) and the spectrum of red light (RD) (Spc.RD) as a molecule. It is.
[0029]
(1) (Spc.IR-Spc.RD) /Spc.RD
(2) (Spc.IR−Spc.RD) /Spc.IR
(3) (Spc.IR−Spc.RD) / (Spc.IR + Spc.RD)
[0030]
  Note that the following (1) 'to (3)' may be used as these equivalent arithmetic expressions.
(1) '{1- (Spc.RD / Spc.IR)} / (Spc.RD / Spc.IR)
(2) '{1- (Spc.RD / Spc.IR)} / (Spc.IR/Spc.IR)
(3) '{1- (Spc.RD / Spc.IR)} / {1+ (Spc.RD / Spc.IR)}
[0031]
  Here, what is common in the operations of (1) ′ to (3) ′ is that {1- (Spc.RD / Spc.IR)} is used as a numerator. In order to avoid the denominator becoming zero, a predetermined value may be added in advance to the denominator, and the calculation result is substantially the same as the calculation result without adding the predetermined value in advance. The same applies to the spectrum calculation shown below.
[0032]
  An example of the result is shown in FIGS. 5 (a) and 5 (b). FIG. 5A is a graph showing a spectrum calculation result when the pulse hardly fluctuates in the period [i]. In the graph shown in FIG. 5A, (1), (2), and (3) indicate the calculation results of (1), (2), and (3), respectively. As this graph shows, since the pulse hardly fluctuated in the period [i], the spectrum at the fundamental frequency of the pulse wave is conspicuous, and the frequency is determined almost at one point. Since the second high frequency is also emphasized for the normalization calculation, it is easy to determine the fundamental frequency. Further, it is shown that the fundamental frequency of the pulse wave can be obtained from any of the calculation results of (1), (2), and (3). FIG. 5B is a graph showing a spectrum calculation result when the pulse fluctuates greatly in the period [i]. (1), (2), and (3) in the graph shown in FIG. 5 (b) indicate the calculation results of the above (1), (2), and (3), respectively. As shown in this graph, since the pulse greatly fluctuated in the period [i], it can be seen that although the spectrum at the fundamental frequency of the pulse wave is conspicuous, the frequency has a width.
[0033]
  Here, it will be described that the graph of the calculation result of this spectrum is as shown in FIG. There is a difference between the amplitudes of the pulse waves of the infrared light (IR) and the red light (RD) that do not contain noise, and the amplitude of the pulse wave of the infrared light (IR) is usually larger. On the other hand, the amplitude of noise due to body movement or the like may be larger than the amplitude of the pulse wave of infrared light (IR) or red light (RD), but in infrared light (IR) and red light (RD). In many cases, the amplitude of the noise is the same. Therefore, when the spectrum is obtained by Fourier transforming each pulse wave of infrared light (IR) and red light (RD) including noise, the fundamental frequency of the pulse wave in the pulse wave spectrum of infrared light (IR) The spectrum at is larger than that in the red light (RD) pulse wave spectrum.
[0034]
  On the other hand, the noise spectrum has a relatively small difference. Therefore, if the difference between the pulse wave spectrum of the infrared light (IR) and the pulse wave spectrum of the red light (RD) is taken for each frequency, the infrared light (IR) and red that are originally present in the spectrum of the fundamental frequency of the pulse wave are obtained. A spectral difference caused by the difference between the amplitudes of each pulse wave of light (RD) appears. Further, the difference in the frequency of the noise component is suppressed because the amplitude of noise in the infrared light (IR) and the red light (RD) is often the same.
[0035]
  Therefore, in the calculations (1) to (3) above, the numerator is (Spc.IR-Spc.RD). Furthermore, by taking Spc.RD, Spc.IR, or (Spc.IR + Spc.RD) as the denominator, the difference in the noise component spectrum is divided by the noise component spectrum. Is suppressed. On the other hand, even if the spectral difference at the fundamental frequency of the pulse wave is divided by Spc.RD, Spc.IR, or (Spc.IR + Spc.RD) as the denominator, it stands out without being suppressed as much as the noise. . However, in the period of the data subjected to Fourier transform, when the pulse greatly fluctuates, the fundamental frequency of the pulse wave fluctuates, so as shown in FIG. The fundamental frequency of the pulse wave has a width. In addition, since the obtained spectrum is normalized, it can be expressed between about 0.4 and 1, and has a role as a discriminant function.
[0036]
  Next, the step of “filter formation” S13 shown in FIG. 2 will be described. As shown in FIG. 5A, when the pulse hardly fluctuates, a filter is formed based on the fundamental frequency (referred to as fs) of the pulse wave. As shown in FIG. 6A, the filtering frequency is assumed to be 2 fs, 3 fs, 4 fs,..., N · fs, which are the fundamental frequency fs of the pulse wave and the harmonics thereof. Alternatively, as shown in FIG. 6 (b), the filtering frequency is set to a predetermined width (band) centered on the fundamental frequency fs of the pulse wave and the harmonic frequencies 2fs, 3fs, 4fs,. ) May be a rectangular shape. Alternatively, as shown in FIG. 6C, even a Gaussian characteristic filter centered on the fundamental frequency fs of the pulse wave and the harmonic frequencies thereof, 2fs, 3fs, 4fs,..., N · fs. Good.
[0037]
  Although various detection methods of fs are conceivable, in the calculation result shown in FIG. 5A, for example, (1) the peak from the frequency range corresponding to the maximum possible range of the pulse rate in the living body. (2) Provide a sufficient threshold that will appear as a spectral value if it is the fundamental frequency of the pulse wave from a statistical point of view, and the frequency at which the spectral value exceeds this threshold (3) Provide a sufficient threshold that will appear if it is the fundamental frequency of the pulse wave from a statistical point of view. Among the frequencies exceeding the value, there is a method of finding the lowest frequency among the frequencies that appear almost in integral multiples and can be recognized as harmonics.
[0038]
  As shown in FIG. 5B, when the pulse fluctuates greatly, the fundamental frequency of the pulse wave has a width. In the frequency band having this width, the maximum frequency is fhigh and the minimum frequency is low. The center frequency is defined as fc, which is defined as fc = √ (flow) × √ (fhigh). Although various detection methods of fhigh and flow are conceivable, in the calculation result shown in FIG. 5B, for example, (1) From the range of frequencies corresponding to the maximum possible range of pulse rate variation in the living body, Statistically, if there is a fundamental frequency of the pulse wave, a sufficient threshold that will appear as a spectrum value is provided, and the frequency band that has the lowest frequency band out of the frequency band that exceeds this threshold exceeds the threshold. Among them, the minimum frequency value is set to “low” and the maximum frequency value is set to “fhigh”. (2) A sufficient threshold that will appear if it is the fundamental frequency of the pulse wave is provided statistically. Of the frequency bands that exceed the spectrum value, the lowest frequency value among the lowest frequency bands that can be recognized as harmonics, with the frequency band appearing almost integer times. flow, and fhigh the maximum frequency value, there is a method such as. The calculation for obtaining fc may be fc = (flow + fhigh) / 2 instead of the above.
[0039]
  Next, there are various methods for forming the filtering frequency. Here, seven methods shown in FIG. 7 will be described.
[0040]
  The filtering frequency range of the method shown in FIG. 7A is set to fhigh to low, and 2fhigh to 2flow, 3fhigh to 3flow,..., N · fhigh to n · flow as this harmonic range. It is a rectangular shape.
  The range of the filtering frequency of the method shown in FIG. 7B is the same as the range of fhigh to low including the center frequency fc and the range of fhigh to flow centering on 2fc as the range of this harmonic. A range in which a range having the same width as fhigh to flow around 3fc is set, and a range having the same width as fhigh to flow around n · fc is set. is there.
  In the method shown in FIG. 7C, fhigh to flow are divided into a predetermined number, and the center frequencies of the divided frequency widths are fs-1, fs-2, fs-3, ..., fs-n, respectively. And And the range of the filtering frequency is the frequency divided above with the center frequency of 2fs-1, 2fs-2, 2fs-3, ... 2fs-m as the range of fhigh to low and the range of this harmonic. Range having the same width as the width, 3fs-1, 3fs-2, 3fs-3,..., 3fs-m as the center frequency, the range having the same width as the divided frequency width,. .., n.fs-1, n.fs-2, n.fs-3,..., N.fs-m as the center frequency, a range having the same width as the divided frequency width was set. It is rectangular.
  FIGS. 7D to 7F show cases where a Gaussian characteristic filter is set instead of a rectangular shape in FIGS. 7A to 7C. That is, FIG. 7D shows an area of fhigh to low, which is the band of the filtering frequency shown in FIG. 7A, 2fhigh to 2flow, 3fhigh to 3flow,..., N · fhigh to n · flow. Similarly, the setting of the Gaussian characteristic filter corresponding to each band is shown.
  FIG. 7E shows a range from fhigh to low including the center frequency fc, which is the band of the filtering frequency shown in FIG. 7B, and a range having the same width as the width from fhigh to flow around 2fc. In the range having the same width as the width of fhigh to flow around the center,..., In the range having the same width as the width of fhigh to flow around n · fc, the respective areas are the same, The setting of the Gaussian characteristic filter corresponding to each band is shown.
  FIG. 7 (f) shows a range of fhigh to flow divided into a predetermined number which is the band of the filtering frequency shown in FIG. 7 (c), 2fs-1, 1fs-2, 2fs-3,..., 2fs−. Range having the same width as the frequency width divided with m as the center frequency, 3fs-1, 3fs-2, 3fs-3,..., the same width as the frequency width divided with 3fs-m as the center frequency In a range having the same width as the divided frequency width with n · fs-m as the center frequency, and corresponding to each band with the same area. It shows the setting of a Gaussian characteristic filter.
  FIG. 7G shows the setting of the filtering frequency at the Fourier transform resolution level. Here, since the data length used for the Fourier transform is 16.384 [seconds] (= 1024 × 16 msec), the resolution is 0.061 [Hz] = 1 / 16.384 [seconds].
[0041]
  Therefore, the frequency of 0.061 [Hz], which is the resolution between flow and fhigh, that is, flow, flow + 0.061 [Hz], flow + 0.061 [Hz] 2,..., Flow + 0.061 [Hz] Xn,..., Hhigh are the fundamental frequencies of the pulse wave. Next, a frequency that is an integral multiple of each fundamental frequency of these pulse waves is set as a harmonic frequency. That is, a frequency of 0.061 [Hz] × 2 frequency interval from 2 flow to 2 fhigh, a frequency of 0.061 [Hz] × 3 frequency interval from 3 flow to 3 fhigh, 0 from n · flow to n · fhigh A frequency with a frequency interval of .061 [Hz] × n is a harmonic frequency. The fundamental frequency and the harmonic frequency are employed as the filtering frequency. The basic concept of forming the filtering frequency described with reference to FIGS. 7A to 7G is that when a plurality of basic frequencies are generated due to fluctuations in the pulse wave, the filtering frequency includes each basic frequency and its harmonics. This is a frequency obtained by ORing the frequencies.
[0042]
  Here, the reason why the fundamental frequency of the pulse wave and the harmonic frequency thereof are used as the filtering frequency forming method will be described. That is, the FFT has a low resolution at a low frequency and a high resolution at a high frequency. Therefore, when the wave is fluctuating, even if the fluctuation cannot be extracted by filtering with a low frequency, it can be extracted by filtering with a harmonic frequency. Therefore, if it is assumed that the fundamental frequency of the pulse wave fluctuates, the fluctuation of the pulse wave can also be extracted by filtering based on the harmonic frequency.
[0043]
  For this reason, in order to extract the pulse wave waveform as faithfully as possible, it is necessary to perform filtering to a considerable harmonic frequency. On the other hand, in order to measure the SpO2 value, it is sufficient if the amplitude of the pulse wave can be detected as an operation. Therefore, filtering up to the harmonic frequency is not necessarily required, and filtering of about 2nd to 4th order is sufficient. What order of high frequency is used for filtering may be appropriately determined. Note that if the filtering frequency is too wide, the filtered pulse wave will probably contain noise, so the filtering frequency is set appropriately.
[0044]
  The matched filter obtained from 1024 data in the period [i] shown in FIG. 3 has a period [I] before the period [i] and the period [i + 1] forming the next matched filter are started. Used to filter the pulse wave. Then, the matched filter obtained from the data of the period [i + 1] for forming the next matched filter has the period [I + 1] before the period [i] and the period for forming the next matched filter are started. Used to filter the pulse wave. The period for forming the matched filter may be set every predetermined period, or the biological information parameters such as the heart rate obtained by the electrocardiogram measurement and the pulse rate obtained by the pulse wave measurement are not less than a predetermined value. You may provide when it fluctuates. In this embodiment, the sampling interval is set to 16 msec and the period for forming the matched filter is set to a period corresponding to the number of 1024 samples. However, the present invention is not limited to this.
[0045]
  Second embodiment
  Next, a second embodiment will be described with reference to FIG. As in the first embodiment, the pulse wave sampling interval is 16 msec and the number of data used to form the matched filter is 1024. Then, Fourier transform is performed from the data in the period [i], the absolute value of the spectrum is obtained, the fundamental frequency of the pulse wave is obtained by performing any one of the following (1) to (3), and the matched filter The formation of is the same as in the first embodiment.
[0046]
(1) (Spc.IR-Spc.RD) /Spc.RD
(2) (Spc.IR−Spc.RD) /Spc.IR
(3) (Spc.IR−Spc.RD) / (Spc.IR + Spc.RD)
[0047]
  Note that the following (1) 'to (3)' may be used as these equivalent arithmetic expressions.
(1) '{1- (Spc.RD / Spc.IR)} / (Spc.RD / Spc.IR)
(2) '{1- (Spc.RD / Spc.IR)} / (Spc.IR/Spc.IR)
(3) '{1- (Spc.RD / Spc.IR)} / {1+ (Spc.RD / Spc.IR)}
[0048]
  The feature of the second embodiment is that the period for forming the matched filter is made continuous. That is, in FIG. 8, the matched filter formed using the data in the period [i] is used for filtering the pulse wave in the period [i]. In the continuous period [i + 1], a new matched filter is formed using the data in that period and is used to filter the pulse wave in that period [i + 1]. The same applies to the next consecutive period [i + 2]. In this way, a matched filter is formed continuously for each predetermined period, and pulse waves are filtered using the matched filter.
[0049]
  Third embodiment
  Next, a third embodiment will be described with reference to FIG. As in the first embodiment, the pulse wave sampling interval is 16 msec and the number of data used to form the matched filter is 1024. Then, Fourier transform is performed from the data in the period [i], the absolute value of the spectrum is obtained, the fundamental frequency of the pulse wave is obtained by performing any one of the following (1) to (3), and the matched filter The formation of is the same as in the first embodiment.
[0050]
(1) (Spc.IR-Spc.RD) /Spc.RD
(2) (Spc.IR−Spc.RD) /Spc.IR
(3) (Spc.IR−Spc.RD) / (Spc.IR + Spc.RD)
[0051]
  Note that the following (1) 'to (3)' may be used as these equivalent arithmetic expressions.
(1) '{1- (Spc.RD / Spc.IR)} / (Spc.RD / Spc.IR)
(2) '{1- (Spc.RD / Spc.IR)} / (Spc.IR/Spc.IR)
(3) '{1- (Spc.RD / Spc.IR)} / {1+ (Spc.RD / Spc.IR)}
[0052]
  The feature of the third embodiment is that processing is performed while partially overlapping the period for forming the matched filter. That is, in FIG. 9, the matched filter formed using the data in the period [i] is used for filtering the pulse wave in the period [i]. The start point of the period [i + 1] for forming the next new matched filter is, for example, after a period of 64 samples of data from the start point of the period [i]. The matched filter formed using the data of period [i + 1] filters the pulse wave of 64 samples in period [I + 1] from the end of period [i] to the end of period [i + 1] Used to do. Furthermore, the start point of the period [i + 2] for forming the next new matched filter is after a period of 64 samples of the same data as the start point of the period [i + 1]. The matched filter formed using the data of period [i + 2] has a pulse wave of 64 samples in period [I + 2] from the end of period [i + 1] to the end of period [i + 2]. Is used for filtering. Similarly, the matched filter used for filtering the pulse wave for 64 samples of data in the period [I + 3] has a period [i + starting after a delay of 64 samples from the start point of the period [i + 2]. 3]. In this way, while the period for forming the matched filter is shifted by a predetermined number of samples, the termination period for the shifted period is filtered by the formed matched filter. In the present embodiment, the sampling interval is set to 16 msec, the period for forming the matched filter is set to a period for 1024 samples, and the period for shifting the period for forming the matched filter is set to a period for 64 samples. It does not need to be limited to this.
[0053]
  Fourth embodiment
  Next, a fourth embodiment will be described with reference to FIG. Similar to the first embodiment, the pulse wave sampling interval is 16 msec, and the number of data units for obtaining the spectrum is 1024. A feature of the fourth embodiment is that a period for obtaining an absolute value of a pulse wave spectrum is set by shifting a plurality of periods, an absolute value of a plurality of obtained spectra is averaged, and then a matched filter is formed. It is in. Period [i], period [i + 1], period [i + 2],..., Period [i + n],... Shown in FIG. It is a period. In each period [i], period [i + 1], period [i + 2],..., Period [i + n],..., Fourier transform is performed using 1024 data samples. The absolute value of the spectrum is obtained. An example showing the spectrum (absolute value; the same applies hereinafter) is the same as in FIGS. 4A and 4B shown in the first embodiment. FIG. 4A shows a spectrum (Spc.IR) of infrared light (IR), and FIG. 4B shows a spectrum (Spc.RD) of red light (RD). The n spectra obtained by Fourier transform in the period [i] to the period [i + n] in the period [I] are averaged for each frequency component. That is, the spectrum of infrared light (IR) is averaged (Av.Spc.IR), and the spectrum of red light (RD) is averaged (Av.Spc.RD).
[0054]
  Then, by using the average infrared light spectrum (Av.Spc.IR) and the average red light spectrum (Av.Spc.RD), performing any one of the following (4) to (6), Detect the fundamental frequency of the pulse wave. The calculation here is the sum / quotient calculation of each spectrum for the same frequency. What is common in the calculations of (4) to (6) is that the difference between the average infrared light spectrum (Av.Spc.IR) and the average red light spectrum (Av.Spc.RD) is used as a molecule. .
[0055]
(4) (Av.Spc.IR-Av.Spc.RD) /Av.Spc.RD
(5) (Av.Spc.IR−Av.Spc.RD) /Av.Spc.IR
(6) (Av.Spc.IR−Av.Spc.RD) / (Av.Spc.IR + Av.Spc.RD)
[0056]
  Note that the following (4) 'to (6)' may be used as these equivalent arithmetic expressions.
(4) '{1- (Av.Spc.RD / Av.Spc.IR)} / (Av.Spc.RD / Av.Spc.IR)
(5) '{1- (Av.Spc.RD / Av.Spc.IR)} / (Av.Spc.IR/Av.Spc.IR)
(6) '{1- (Av.Spc.RD / Av.Spc.IR)} / {1+ (Av.Spc.RD / Av.Spc.IR)}
[0057]
  In the period [I], when the pulse hardly fluctuates, the spectrum at the fundamental frequency of the pulse wave stands out, so it is decided at almost one point. When the pulse fluctuates greatly, the spectrum at the fundamental frequency of the pulse wave stands out. However, the fact that the frequency has a width is the same as in FIGS. 5A and 5B shown in the first embodiment. The matched filter formation method is the same as the method described in the first embodiment with reference to FIGS. 6A to 6C and FIGS. 7A to 7G. The obtained matched filter is used for filtering the pulse wave in the period [I].
[0058]
  Next, each of the n spectra obtained by Fourier transform in the period [i + 1] to the period [I + n + 1] in the period [I + 1] includes infrared light (IR) and red light ( For each RD), the frequency components are added and averaged, and an average infrared light spectrum (Av.Spc.IR) and an average red light spectrum (Av.Spc.RD) are calculated. Then, by using the average infrared light spectrum (Av.Spc.IR) and the average red light spectrum (Av.Spc.RD), performing any one of the above (4) to (6), Detect the fundamental frequency of the pulse wave. About the formation method of the matched filter using the fundamental frequency of the pulse wave, as above-mentioned, using FIG. 6 (a)-(c) and FIG. 7 (a)-(g), it is 1st implementation. This is the same as the method described in the embodiment. The matched filter formed using the data of the period [I + 1] is used for filtering 64 samples of data in the period [N + 1] from the end of the period [I] to the end of the period [I + 1]. It is done.
[0059]
  In this embodiment, the sampling interval is 16 msec, the unit for obtaining the spectrum is the period for the number of 1024 samples, and the period for shifting the period for forming the matched filter is the period for the number of 64 samples. It does not have to be. In addition, the period for shifting the period for forming the matched filter may be a period for forming the matched filter (a period corresponding to the number of 1024 samples). At this time, the period for filtering by the matched filter is 1024 data. Here, the advantage of obtaining the fundamental frequency of the pulse wave using the average of the spectrum is that the noise due to body movements or the like unrelated to the pulse wave is canceled and eliminated by averaging the n times. This is because the N ratio is improved by √ (n) times.
[0060]
  Fifth embodiment
  Next, a fifth embodiment will be described with reference to FIG. 3, FIG. 11, and FIG. The feature of this embodiment is that, when obtaining the fundamental frequency of the pulse wave, the spectrum is obtained by Fourier-transforming the difference between the pulse wave of infrared light (IR) and red light (RD). The pulse wave sampling interval for infrared light (IR) and red light (RD) is 16 msec, and the number of data used to form a matched filter is 1024. First Embodiment It is the same. First, the difference between pulse waves of infrared light (IR) and red light (RD) in period [i] is taken, and this is taken as aI−aR (= infrared light (IR) pulse wave−red light (RD) pulse wave. Next, the differential pulse wave is subjected to Fourier transform to obtain the absolute value of the spectrum, as shown in Fig. 11. As can be seen, the difference between the pulse waves is Fourier transformed to obtain the spectrum. Therefore, it is difficult to detect the fundamental frequency of the pulse wave because the spectrum of the noise component is large, so it is easy to detect the fundamental frequency of the pulse wave by performing any of the following calculations (7) to (9). The calculation here is the sum / quotient calculation of each spectrum at the same frequency, and what is common in the calculations of (7) to (9) is infrared light (IR) as a molecule. ) And the red light (RD) pulse wave difference obtained by Fourier transform. It is to use Kuttle (Spc. (AI-aR)).
[0061]
(7) (Spc. (AI-aR)) / Spc.aI
(8) (Spc. (AI-aR)) / Spc.aR
(9) (Spc. (AI-aR)) / (Spc.aI + Spc.aR)
  Where
Spc.aI: Infrared (IR) pulse wave spectrum
Spc.aR: Red light (RD) spectrum
[0062]
  Among these, an example of the calculation result of (9) is shown in FIG. The example shown in FIG. 12 is a graph of a spectrum calculation result when the pulse hardly fluctuates in the period [i]. Therefore, the spectrum at the fundamental frequency of the pulse wave is conspicuous, and the frequency is determined almost at one point. Note that when the pulse fluctuates greatly in the period [i], the spectrum at the fundamental frequency of the pulse wave is conspicuous, but the frequency has a width.
[0063]
  If the fundamental frequency of the pulse wave can be determined by the above operations (7) to (9), FIGS. 6 (a) to (c) and FIGS. 7 (a) to (g) are performed in the first embodiment. As described above, the filtering frequency is determined to form a matched filter. That is, when the pulse hardly fluctuates, the fundamental frequency and its harmonic frequency are set as the filtering frequency as shown in FIG. 6A, or as shown in FIGS. 6B and 6C. In addition, the filtering frequency is set such that the fundamental frequency and its harmonic frequency have a rectangular shape with a width or a Gaussian filter. When the pulse fluctuates greatly, there is a range in the fundamental frequency of the pulse wave, so as shown in FIGS. 7 (a) to 7 (g), depending on the width of the fundamental frequency of the pulse wave caused by the fluctuation of the pulse. Thus, the filtering frequency is set by adding a width to the filtering frequency or by using a Gaussian filter together with the harmonic frequency.
[0064]
  The matched filter obtained from 1024 data in the period [i] in FIG. 3 is the period [I] until the period [i] and the period [i + 1] forming the next matched filter are started. Used to filter the pulse wave. Then, the matched filter obtained from the data of the period [i + 1] for forming the next matched filter has the period [I + 1] before the period [i] and the period for forming the next matched filter are started. Used to filter the pulse wave. Note that the period for forming the matched filter may be set for each required period, or may be provided when a biological information parameter such as a heart rate or a pulse rate fluctuates more than a predetermined value. In this embodiment, the sampling interval is set to 16 msec and the period for forming the matched filter is set to a period corresponding to the number of 1024 samples. However, the present invention is not limited to this.
[0065]
  Sixth embodiment
  Next, a sixth embodiment will be described with reference to FIG. The pulse wave sampling interval is 16 msec, the number of data used to form the matched filter is 1024, and the difference between the pulse wave of infrared light (IR) and red light (RD) is Fourier transformed. The method for obtaining the fundamental frequency from the calculation of any of (7) to (9) using the spectrum obtained in the above and the method for forming the matched filter from the fundamental frequency of the pulse wave are the same as in the fifth embodiment. is there. A feature of the sixth embodiment is that a period for forming a matched filter is made continuous. That is, in FIG. 8, the matched filter formed using the data in the period [i] is used for filtering the pulse wave in the period [i]. In the continuous period [i + 1], a new matched filter is formed using the data in that period and is used to filter the pulse wave in that period [i + 1]. The same applies to the next consecutive period [i + 2]. In this way, a matched filter is formed continuously for each predetermined period, and pulse waves are filtered using the matched filter.
[0066]
  Seventh embodiment
  Next, a seventh embodiment will be described with reference to FIG. The pulse wave sampling interval is 16 msec, the number of data used to form the matched filter is 1024, and the difference between the pulse wave of infrared light (IR) and red light (RD) is Fourier transformed. The method for obtaining the fundamental frequency from the calculation of any of (7) to (9) using the spectrum obtained in the above and the method for forming the matched filter from the fundamental frequency of the pulse wave are the same as in the fifth embodiment. is there. The seventh embodiment is that processing is performed while partially overlapping the period for forming the matched filter. That is, in FIG. 9, the matched filter formed using the data in the period [i] is used for filtering the pulse wave in the period [i]. The start point of the period [i + 1] for forming the next new matched filter is, for example, after a period of 64 samples of data from the start point of the period [i]. The matched filter formed using the data of period [i + 1] filters the pulse wave of 64 samples in period [I + 1] from the end of period [i] to the end of period [i + 1] Used to do. Furthermore, the start point of the period [i + 2] for forming the next new matched filter is after a period of 64 samples of the same data as the start point of the period [i + 1]. The matched filter formed using the data of period [i + 2] has a pulse wave of 64 samples in period [I + 2] from the end of period [i + 1] to the end of period [i + 2]. Is used for filtering. Similarly, the matched filter used for filtering the pulse wave for 64 samples of data in the period [I + 3] has a period [i + starting after a delay of 64 samples from the start point of the period [i + 2]. 3]. In this way, while the period for forming the matched filter is shifted by a predetermined number of samples, the termination period for the shifted period is filtered by the formed matched filter. In the present embodiment, the sampling interval is set to 16 msec, the period for forming the matched filter is set to a period for 1024 samples, and the period for shifting the period for forming the matched filter is set to a period for 64 samples. It does not need to be limited to this.
[0067]
  Eighth embodiment
  Next, an eighth embodiment will be described with reference to FIG. As in the fifth embodiment, the pulse wave sampling interval is 16 msec, and the number of data units for obtaining the spectrum is 1024. A feature of the eighth embodiment is that a period for obtaining an absolute value of a pulse wave spectrum is set by shifting a plurality of periods, an absolute value of a plurality of obtained spectra is averaged, and then a matched filter is formed. It is in. Period [i], period [i + 1], period [i + 2],..., Period [i + n],... Shown in FIG. It is a period. In each period [i], period [i + 1], period [I + 2],..., Period [i + n], calculation is next performed for each 1024 data samples.
[0068]
(1) The spectrum Spc (aI-aR) is obtained by Fourier transforming the difference (aI-aR) between the pulse wave of infrared light (IR) and the pulse wave of red light (RD).
(2) The spectrum Spc.IR and / or Spc.RD is obtained by Fourier transforming the pulse wave of infrared light (IR) and / or the pulse wave of red light (RD).
(3) An average of spectrum Spc (aI−aR) for n periods [i], period [i + 1], period [i + 2],..., Period [i + n], Calculate Av.Spc (aI-aR).
(4) Add the spectrum Spc.IR and / or Spc.RD for each period [i], period [i + 1], period [i + 2], ..., period [i + n] Average to determine Av.Spc.IR and / or Av.Spc.RD.
(5) The fundamental frequency of the pulse wave is obtained by calculating one of the following (10) to (12).
(10) (Av.Spc. (AI-aR)) / Av.Spc.RD
(11) (Av.Spc. (AI-aR)) / Av.Spc.IR
(12) (Av.Spc. (AI-aR)) / (Av.Spc.IR + Av.Spc.RD)
[0069]
  What is common in the calculations of (10) to (12) is that, as a molecule, the spectrum of the difference between the pulse of infrared light (IR) and the pulse wave of red light (RD) is averaged (Av. Spc. (aI-aR)). The calculation result by the above (12) is almost as shown in FIG. 12, but the noise component is canceled and eliminated by averaging, so that the fundamental frequency of the pulse wave becomes easier to detect.
[0070]
  In the period [I], when the pulse hardly fluctuates, the spectrum at the fundamental frequency of the pulse wave stands out, so it is decided at almost one point. When the pulse fluctuates greatly, the spectrum at the fundamental frequency of the pulse wave stands out. Although there is a width in the frequency range, it is the same as in the first embodiment described with reference to FIGS. 5 (a) and 5 (b). The matched filter formation method is the same as the method described in the first embodiment with reference to FIGS. 6A to 6C and FIGS. 7A to 7G. The obtained matched filter is used for filtering the pulse wave in the period [I].
[0071]
  Next, the matched filter forming method in the period [I + 1] is the same as the matched filter forming method in the period [I]. That is, using the data of each period [i + 1] to period [i + n + 1], the above processes (1) to (5) are performed to obtain the fundamental frequency of the pulse wave. And using it, a matched filter is formed by the method demonstrated in 1st Embodiment using Fig.6 (a)-FIG.7 (g). The matched filter formed using the data of the period [I + 1] is used for filtering 64 samples of data in the period [N + 1] from the end of the period [I] to the end of the period [I + 1]. It is done. Further, similarly, the matched filter in the period [I + 2] is used to filter the data in the period [N + 2], and the matched filter in the period [I + 3] is used in the data in the period [N + 3]. Used to filter This process is repeated.
[0072]
  In the present embodiment, the sampling interval is set to 16 msec, the unit for obtaining the spectrum is set to a period corresponding to the number of 1024 samples, and the period for shifting the period for forming the matched filter is set to a period corresponding to the number of 64 samples. May be. In addition, the period for shifting the period for forming the matched filter may be a period for forming the matched filter (a period corresponding to the number of 1024 samples). At this time, the period for filtering by the matched filter is 1024 data.
[0073]
  Here, the points to be considered when designing the algorithm will be described. The resolution of the Fourier transform is determined by the reciprocal of the data length used for the Fourier transform. In this embodiment, since the data length used for Fourier transform is 16.384 [seconds] (= 1024 × 16 msec), the resolution is 0.061 [Hz] = 1 / 16.384 [seconds]. . In this way, the necessary Fourier transform resolution is taken into consideration in the design. Moreover, since it is necessary to display the filtered pulse wave continuously on the display device, (1) the time length of the pulse wave to be displayed on the display device, and (2) real-time display of the pulse wave is permitted. The time difference between the appearance of the pulse wave appearing on the living body and the display timing on the display, (3) the time during which the pulse wave detected sequentially can be stored and processed by the above-described various embodiments, etc. will also need to be considered.
[0074]
  Ninth embodiment
  Next, a ninth embodiment will be described. After the pulse wave of infrared light (IR) and the pulse wave of red light (RD) are detected, at least one of them is used to determine the pulse rate conventionally used. Then, using the pulse frequency as the fundamental frequency of the pulse wave, a matched filter forming method is formed by the matched filter forming method described in the first embodiment with reference to FIGS. The formed matched filter is used to filter the pulse wave of infrared light (IR) and the pulse wave of red light (RD). Since the pulse rate is detected sequentially, a matched filter can be formed sequentially accordingly, and the filtering frequency of the matched filter can be shifted following the fluctuation of the pulse rate. Since the pulse rate obtained by the pulse wave measurement and the heart rate obtained by the electrocardiogram measurement are substantially the same, in this embodiment, the electrocardiogram measurement is performed in order to obtain the fundamental frequency of the pulse wave, and the heart rate frequency is calculated. You may use as a fundamental frequency of a pulse wave.
[0075]
  In the above-described embodiment, nine embodiments are exemplified, but these may be continuously switched according to the situation. For example, in the initial measurement stage, there is not enough data to average the spectrum, so the first to third embodiments or the fifth to seventh and ninth embodiments that do not perform the averaging are adopted. Also good. Alternatively, in the fourth or eighth embodiment, the average number of times n may be reduced. When sufficient data has been accumulated, the average number of additions n may be gradually increased in the fourth or eighth embodiment. In any case, it can be said that the matched filter is adaptive in that the matched filter is formed according to the fundamental frequency of the pulse wave that changes over time.
[0076]
  In the above embodiment, the absolute value of the spectrum of the pulse wave of each wavelength is used to obtain the basic period of the pulse wave, but a power spectrum may be used instead. The spectrum according to the present application is a concept including a power spectrum in addition to an absolute value of a spectrum.
[0077]
  In the first to third embodiments, the following calculations (1) to (3) are used to specify the fundamental frequency of the pulse wave.
(1) (Spc.IR-Spc.RD) /Spc.RD
(2) (Spc.IR−Spc.RD) /Spc.IR
(3) (Spc.IR−Spc.RD) / (Spc.IR + Spc.RD)
[0078]
  Instead, when the pulsation component of the absorbance at the wavelength of red light is ΔA1, and the pulsation component of the absorbance at the wavelength of infrared light is ΔA2, the difference between ΔA1 and ΔA2 on the time axis is obtained, and the difference signal is obtained. The fundamental frequency may be detected by performing Fourier transform using data samples in the same predetermined period as described above to obtain the absolute value of the spectrum.
[0079]
  The pulse wave of infrared light (IR) and the pulse wave of red light (RD) filtered according to each embodiment can be displayed by a display. Either one may be displayed on the display. Further, the SpO2 value (oxygen saturation) can be obtained by an existing method or the like using the pulse wave of the filtered infrared light (IR) and the pulse wave of the red light (RD). That is, the pulsation component obtained from the filtered pulse wave of red light is ΔA1, the pulsation component obtained from the filtered pulse wave of infrared light is ΔA2, and the absorbance ratio Ф = ΔA1 / ΔA2 is obtained, The oxygen saturation is obtained from the oxygen saturation S = f (Ф).
[0080]
  In the above embodiment, the basic frequency is obtained from the pulse wave signal, a matched filter is formed, and noise is removed by this filter. However, the calculation for obtaining the fundamental frequency and the processing for removing the noise in the present invention are not limited to the pulse wave signal, and can be applied to other biological signals or more generally to industrial signals in general. It is. In particular, it is suitable for application to continuous signals having periodicity and two signals having the same fundamental frequency. Therefore, in the above embodiment, the pulse wave of infrared light (IR) and the pulse wave of red light (RD) can be replaced with two industrial general signals having the same fundamental frequency. Accordingly, Spc.IR, Spc.RD, Av.Spc.IR, and Av.Spc.RD can be applied by replacing the spectrum obtained by Fourier transforming an industrial general signal.
[0081]
【The invention's effect】
  As detailed above, according to claim 1 of the present inventionPulse waveAccording to the signal processing method, since the first signal IR and the second signal RD have the same fundamental frequency, each spectrum (( Spc.IR−Spc.RD) or calculation based on {1- (Spc.RD / Spc.IR)}, the basic frequency of the signal can be calculated accurately even if the signal contains noise. it can.Moreover, the component resulting from noise can be suppressed and the fundamental frequency can be more easily extracted..
[0082]
  Of the present inventionClaim 2Pertaining toPulse waveAccording to the signal processing method, it is possible to calculate the fundamental frequency of the signal while suppressing the noise appearing in the spectrum by using the spectrum averaging calculation.
[0083]
  Of the present inventionClaim 3Pertaining toPulse waveAccording to the signal processing method, the component resulting from noise can be suppressed and the fundamental frequency can be extracted more easily by using the averaged spectrum for calculation.
[0084]
  Of the present inventionClaim 4Pertaining toPulse waveSignal processing methodAccording toFrom the pulse wave signal from which noise has been removed, a fundamental frequency component that is a characteristic frequency component of the pulse wave signal and a frequency component of its harmonics can be extracted to form a pulse wave signal.
[0085]
  Of the present inventionClaim 5Pertaining toPulse waveSignal processing methodAccording toEven when the fundamental frequency fluctuates, each fundamental frequency component and the frequency component of each harmonic can be extracted to form a pulse wave signal.
[0086]
  Of the present inventionClaim 6Pertaining toPulse waveSignal processing methodAccording toThe pulse wave signal can be formed by taking in the frequency components around the fundamental frequency and the harmonic frequency.
[0087]
  Of the present inventionClaim 7Pertaining toPulse waveSignal processing methodAccording toA pulse wave signal can be formed by incorporating frequency components around the fundamental frequency and the harmonic frequency thereof in a Gaussian distribution.
[0088]
  Of the present inventionClaim 8Pertaining toPulse waveSignal processing methodAccording toA fundamental frequency can be obtained from a pulse wave signal of a living body, a characteristic frequency component can be extracted, and a pulse wave signal can be formed.
[0089]
  Of the present inventionClaim 9Pertaining toPulse waveSignal processing methodAccording toThe characteristic frequency component of the pulse wave can be extracted from the pulse wave signal of the living body, and the pulse wave signal can be formed and displayed.
[0090]
  Of the present inventionClaim 10Pertaining toPulse waveSignal processing methodAccording toThe oxygen saturation can be accurately calculated using the filtered pulse wave.
[0091]
  Of the present inventionClaim 11Pulse wave signal processing method according toAccording toWhen calculating the difference between the pulsating components to obtain the fundamental frequency of the pulse wave, noise can be suppressed by calculating the addition average of the spectrum, and the fundamental frequency of the pulse wave can be extracted more easily.
[0092]
  Of the present inventionClaim 12Pertaining toPulse waveAccording to the signal processing method, a filter for forming a pulse wave can be formed by extracting the frequency components of the fundamental frequency at which the characteristic of the pulse wave signal appears and the harmonic frequency thereof.
[0093]
  Of the present inventionClaim 13Pertaining toPulse waveAccording to the signal processing method, even if the fundamental frequency of the pulse wave signal fluctuates, a filter for extracting the frequency component of each fundamental frequency and each of its harmonics and forming the pulse wave signal Can be formed.
[0094]
  Of the present inventionClaim 14Pertaining toPulse waveAccording to the signal processing method, a pulse wave signal can be formed by taking in frequency components around the fundamental frequency and the harmonic frequency thereof.
[0095]
  Of the present inventionClaim 15Pertaining toPulse waveAccording to the signal processing method, a pulse wave signal can be formed by taking in frequency components around the fundamental frequency and the harmonic frequency thereof in a Gaussian distribution.
[0096]
  Of the present inventionClaim 16Pertaining toPulse waveAccording to the signal processing method, the filtered pulse wave signal can be displayed.
[0097]
  Of the present inventionClaim 17Pertaining toPulse waveAccording to the signal processing method, the oxygen saturation can be accurately calculated using the filtered pulse wave.
[Brief description of the drawings]
[Figure 1] BookInvention1 shows a configuration of an oxygen saturation measuring apparatus that performs pulse wave signal processing according to an embodiment.Block connectionFIG.
[Figure 2] BookInventionPulse wave signal processing according to the embodimentFlowchart showing the programIt is.
[Figure 3] BookInventionThe time chart of the pulse wave signal processing concerning an embodiment is shown.IllustrationIt is.
FIG. 4 (a) shows the spectrum of an infrared light pulse wave.Characteristic diagram, (B) shows the spectrum of red light pulse waveCharacteristic diagramIt is.
FIG. 5A shows a spectrum calculation result for obtaining a fundamental frequency of a pulse wave.Characteristic diagram, (B) shows the calculation result of the spectrum for obtaining the fundamental frequency of the pulse wave.Characteristic diagramIt is.
FIG. 6A is a filtering frequency for filtering a pulse wave.Characteristic diagram of, (B) is a rectangular frequency characteristic filter for filtering pulse wavesCharacteristic diagram of, (C) is a Gaussian characteristic filter for filtering the pulse waveCharacteristic diagram ofIt is.
7A is a rectangular frequency characteristic filter for filtering pulse waves. FIG.Characteristic diagram of, (B) is a rectangular frequency characteristic filter for filtering pulse wavesCharacteristic diagram of, (C) is a rectangular frequency characteristic filter for filtering pulse wavesCharacteristic diagram of, (D) is a Gaussian characteristic filter for filtering the pulse waveCharacteristic diagram of, (E) is a Gaussian characteristic filter for filtering pulse wavesCharacteristic diagram of, (F) is a Gaussian characteristic filter for filtering the pulse waveCharacteristic diagram of, (G) is a filtering frequency for filtering the pulse waveCharacteristic diagram ofIt is.
[Figure 8] BookInventionThe time chart of the pulse wave signal processing concerning an embodiment is shown.IllustrationIt is.
[Figure 9] BookInventionThe time chart of the pulse wave signal processing concerning an embodiment is shown.IllustrationIt is.
FIG. 10 BookInventionThe time chart of the pulse wave signal processing concerning an embodiment is shown.IllustrationIt is.
FIG. 11 shows the spectrum of the difference between the infrared light pulse wave and the red light pulse wave.Characteristic diagramIt is.
FIG. 12 shows the calculation result of the spectrum for obtaining the fundamental frequency of the pulse wave.Characteristic diagramIt is.
[Explanation of symbols]
1 Probe
2 Light emitting part
3 Light receiver
4 Living tissue (finger)
5 Light emitting part drive circuit
6 Light receiving signal amplifier circuit
7 Demodulation circuit
8 CPU
9a amplifier
9b amplifier
10a A / D converter
10b A / D converter
11 Display

Claims (17)

In a pulse wave signal processing method for processing a first signal IR and a second signal RD which have the same fundamental frequency and are continuous,
For each of the first signal IR and the second signal RD, a frequency spectrum or a frequency power spectrum is calculated in a certain period based on the number of samples of the pulse wave signal , and the first signal IR first A spectrum calculation step for calculating a spectrum Spc.IR and a second spectrum Spc.RD of the second signal RD , respectively .
Using the first spectrum Spc.IR and the second spectrum Spc.RD calculated in the spectrum calculation step, calculate a spectrum that takes a difference between them on the frequency axis, or calculate a spectrum that takes a difference between them. A fundamental frequency calculating step of calculating a spectrum to be further normalized, and calculating a fundamental frequency of the first signal IR and the second signal RD based on the result,
In the fundamental frequency calculating step,
(Spc.IR−Spc.RD) /Spc.RD,
(Spc.IR−Spc.RD) /Spc.IR,
(Spc.IR−Spc.RD) / (Spc.IR + Spc.RD),
{1- (Spc.RD / Spc.IR)} / (Spc.RD / Spc.IR),
{1- (Spc.RD / Spc.IR)} / (Spc.IR/Spc.IR), or
{1- (Spc.RD / Spc.IR)} / {1+ (Spc.RD / Spc.IR)}
A pulse wave signal processing method, wherein the fundamental frequency is calculated based on a calculation result of any one of the spectra. In a pulse wave signal processing method for processing a first signal IR and a second signal RD which have the same fundamental frequency and are continuous,
For each of the first signal IR and the second signal RD, a frequency spectrum or a frequency power spectrum is calculated in a certain period based on the number of samples of the pulse wave signal , and the first signal IR is calculated based on the first signal IR. A spectrum calculation step of calculating a spectrum Spc.IR and a second spectrum Spc.RD from the second signal RD , respectively .
A spectral addition average calculation step of adding and averaging a plurality of times of the first spectrum Spc.IR and the second spectrum Spc.RD by the spectrum calculation step;
Using the first spectrum Av.Spc.RD added and averaged by the spectrum addition average calculating step and the second average Av.Spc.RD added and averaged, a spectrum that takes a difference between them on the frequency axis is calculated. fundamental frequency, and or further calculates the spectrum normalized in addition to the calculation of the spectrum taking the difference between each other, based on the result, calculates the fundamental frequency of said first signal IR and the second signal RD A pulse wave signal processing method comprising: an arithmetic step. In the pulse wave signal processing method according to claim 2 ,
In the fundamental frequency calculating step,
(Av.Spc.IR−Av.Spc.RD) /Av.Spc.RD,
(Av.Spc.IR−Av.Spc.RD) /Av.Spc.IR,
(Av.Spc.IR−Av.Spc.RD) / (Av.Spc.IR + Av.Spc.RD),
{1- (Av.Spc.RD / Av.Spc.IR)} / (Av.Spc.RD / Av.Spc.IR),
{1- (Av.Spc.RD / Av.Spc.IR)} / (Av.Spc.IR/Av.Spc.IR), or
{1- (Av.Spc.RD / Av.Spc.IR)} / {1+ (Av.Spc.RD / Av.Spc.IR)}
A pulse wave signal processing method, wherein the fundamental frequency is calculated based on a calculation result of any one of the spectra. In the pulse wave signal processing method according to any one of claims 1 to 3 ,
Furthermore, a filter forming step of forming a filter using the fundamental frequency calculated in the fundamental frequency calculating step and its harmonics, and
Wherein the filter formed by the filter formation step, pulse wave signal processing method characterized by comprising a filtering step of filtering at least the first signal IR or the second signal RD, a. In the pulse wave signal processing method according to claim 4 ,
The filter forming step, when the fundamental frequency calculated by the fundamental frequency calculating step there is a plurality, and forming a filter by using the frequency of the harmonics of each and each fundamental frequency pulse Signal processing method. The pulse wave signal processing method according to claim 4 or 5 ,
In the filter forming step, characteristics of the filter, the fundamental frequency and the pulse wave signal processing method characterized in that the frequency of the harmonics is rectangular having a width about each. The pulse wave signal processing method according to claim 4 or 5 ,
In the filter forming step, wherein the filter, the fundamental frequency and the pulse wave signal processing method characterized in that the frequency of the harmonics is a Gaussian characteristic filter centered respectively. In the pulse wave signal processing method according to any one of claims 1 to 7 ,
The pulse signal processing method, wherein both the first signal IR and the second signal RD are pulse signals. In the pulse wave signal processing method according to any one of claims 4 to 7 ,
The first signal IR and the second signal RD are both pulse wave signals,
The pulse wave signal processing method further includes the step of displaying the filtered signal of the first signal IR or the second signal RD in the filtering step. In the pulse wave signal processing method according to any one of claims 4 to 7 ,
The first signal IR is a pulse wave signal obtained by infrared light transmitted or reflected through a living artery,
The second signal RD is a pulse wave signal obtained by red light transmitted or reflected through a living artery,
Filtering both the first signal IR and the second signal RD in the filtering step;
The pulse wave signal processing method further includes an oxygen saturation calculation step of calculating an oxygen saturation using the first signal IR or the second signal RD filtered in the filtering step. In a pulse wave signal processing method for processing a first pulse wave signal and a second pulse wave signal of a living body having the same fundamental frequency,
A first pulsation component detection step of detecting a pulsation component ΔA1 of the first pulse wave signal;
A second pulsation component detection step of detecting a pulsation component ΔA2 of the second pulse wave signal;
A pulsation component calculation step of calculating a difference between the pulsation component of the first pulse wave signal and the pulsation component of the second pulse wave signal;
A spectrum calculation step of performing a frequency spectrum or a frequency power spectrum a plurality of times in a certain period based on the number of samples of the pulse wave signal ;
And spectral averaging calculation step of averaging the multiple partial computed spectrum,
A fundamental frequency calculating step of calculating a fundamental frequency of the pulse wave signal from the averaged spectrum. In the pulse wave signal processing method according to claim 11,
A filter forming step of forming a filter using the fundamental frequency calculated by the fundamental frequency calculating step and a frequency of its harmonics;
A pulse wave signal processing method comprising: a filtering step of filtering at least the first pulse wave signal or the second pulse wave signal with a filter formed by the filter forming step. The pulse wave signal processing method according to claim 12 ,
In the filter forming step, when there are a plurality of the fundamental frequencies calculated in the fundamental frequency calculating step, a filter is formed using each fundamental frequency and each harmonic frequency thereof. Signal processing method. The pulse wave signal processing method according to claim 12 or 13 ,
In the filter forming step, the characteristic of the filter is a rectangular shape having a width centered on the fundamental frequency and the harmonic frequency thereof, respectively. The pulse wave signal processing method according to claim 12 or 13 ,
In the filter forming step, the filter is a Gaussian characteristic filter centered on the fundamental frequency and the harmonic frequency thereof, respectively. The pulse wave signal processing method according to any one of claims 12 to 15 ,
Further, the pulse wave signal processing method characterized by the step of displaying pulse wave signal towards a filtered out of the first pulse wave signal or the second pulse wave signal in the filtering step comprises. The pulse wave signal processing method according to any one of claims 12 to 15 ,
The first pulse wave signal is a pulse wave signal obtained by red light transmitted or reflected through a living artery,
The second pulse wave signal is a pulse wave signal obtained by infrared light transmitted or reflected through a living artery,
Filtering both the first pulse wave signal and the second pulse wave signal in the filtering step;
The pulse wave signal processing further includes an oxygen saturation calculation step of calculating an oxygen saturation using the first pulse wave signal or the second pulse wave signal filtered in the filtering step. Method.

Images