JP2012179209A - Pulsation detector - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a pulsation detector that can effectively eliminate noise while reducing processing load of a detector.SOLUTION: The pulsation detector 100 includes: a pulse wave sensor 10 for outputting a pulse wave signal d in which a pulsation signal and a noise signal including a body movement noise signal resulting from a body movement of a subject are mixed; a pulse wave signal filtering part 200 including an adaptive filter 202 which adaptively filters the pulse wave signal; an adaptive filter reconstruction part 17 for changing a state of the adaptive filter 202 at a first time point during a period in which the pulse wave signal filtering part 200 continuously operates, to the state at a second time point which is a time point prior to the first time point and includes the time point of starting of the operation of the adaptive filter 202; and a pulse wave frequency analyzing part 400 which performs a frequency analyzing processing for each predetermined time period to specify a pulsation presenting spectrum based on a signal output from the pulse wave signal filtering part 200 after filtering.

Description

  The present invention relates to a pulsation detection device and the like.

  The pulsation detecting device is a device for detecting a pulsation derived from a heartbeat of a human body. For example, the pulsation detecting device is based on a signal (pulse wave signal) from a pulse wave sensor attached to an arm, a palm, a finger or the like. This is a device for removing a signal component (body motion influence signal) generated by the influence of body motion as noise and detecting a signal (beat signal) derived from the heartbeat.

  A pulse wave sensor of the type worn on a human finger or wrist is described in, for example, Patent Document 1. A technique for removing a noise component included in a pulse wave signal output from a pulse wave sensor by filtering is described in Patent Document 2, for example. In the technique described in Patent Document 2, a bandpass filter that passes a signal having a frequency close to the frequency indicating the pulse at that time is selected from a plurality of bandpass filters.

JP 2005-198829 A JP 2007-54471 A

  In the technique described in Patent Document 1, the pulse wave sensor is mounted on a person's finger, palm, wrist, or the like. Fingers, palms, wrists, and the like are parts of the body that are particularly moving. Therefore, when a pulse wave sensor is attached to those parts, a lot of noise is mixed in the pulse wave signal output from the pulse wave sensor. For example, when the wearer moves the hand or a part around the hand, a change in blood flow is generated independently of the blood flow due to the heartbeat, and noise is mixed in the signal caught by the pulse wave sensor.

  Further, for example, when the wearer bumps his hand against an object or another body part of the wearer himself, another blood flow change occurs independently of blood flow fluctuation due to the heartbeat. Since the pulse wave sensor catches this blood flow change, noise is mixed in the pulse wave sensor output signal.

  In particular, a wrist-type pulsation detection device (a type of pulsation detection device that is worn on a wrist such as a person) is compared to a finger-type pulsation detection device (a type of pulsation detection device that is worn on a person's finger). Thus, since the obtained pulse wave signal is weak, sufficient noise countermeasures are required.

  For example, bones, tendons, and muscles such as ulna and retaining bones are gathered on the wrist, and the shape of tendons and muscles changes greatly by moving fingers, hands, wrists, and the like. At this time, a change in blood flow occurs. When observing blood flow in arteries and veins, arterial blood flow changes more clearly than in veins, and thus the heartbeat rhythm appears more clearly in the pulse signal. With venous blood flow, it is difficult to see the movement of the heartbeat. However, there are few arterial blood vessels in the subcutaneous tissue (shallow region) outside the wrist. That is, the wrist has few arteries, and the arteries are often deeper. Therefore, when a change in blood flow is captured by the pulse wave sensor, the blood flow change due to an external factor is more dominant than the blood flow change due to the heartbeat. For this reason, when a hand movement or an impact on the periphery of the hand enters, there is a high possibility that a change in blood flow due to a heartbeat becomes difficult to see.

  In the technique described in Patent Document 2, a bandpass filter that passes a signal having a frequency close to the frequency indicating the pulse at that time is selected from a plurality of bandpass filters. However, in order to realize such a process, for example, it is necessary to perform a large number of determination processes on software, which increases the processing burden and increases the time required for the process. I cannot deny. This leads to an increase in the size of the pulsation detecting device and an increase in power consumption. For example, in order to realize a pulsation detecting device as large as a wristwatch, it is important to implement effective noise countermeasures while reducing the processing load and power consumption of the device.

  According to at least one aspect of the present invention, for example, it is possible to realize a pulsation detection device capable of executing effective noise countermeasures while reducing the processing load on the device. In addition, for example, in a pulsation detection device of a type in which a pulse wave sensor is attached to a portion where it is difficult to acquire a pulse wave signal, such as the outside of the wrist (a portion that contacts the back surface of the wristwatch), a pulsation presentation that represents the pulsation signal The specific performance of the spectrum can be improved.

  (1) One aspect of the pulsation detection device of the present invention is a pulsation detection device that detects a pulsation signal derived from the pulsation of a subject, and the pulsation signal and the body motion of the subject are detected. A pulse wave sensor for outputting a pulse wave signal mixed with a noise signal including a derived body motion noise signal, a filter for filtering the pulse wave signal, an adaptive filter for self-adapting frequency response characteristics, and the adaptation The state of the adaptive filter at the first time point during the continuous operation of the filter is the state of the second time point that is a time point before the first time point and includes the time point when the adaptive filter starts operating. A pulse wave signal filtering unit including an adaptive filter reconfiguring unit for reconfiguring, and a frequency analysis process at predetermined intervals based on a signal output from the pulse wave signal filtering unit Go and including a frequency analysis section for detecting a pulsation presentation spectrum representing the beat signal.

  The adaptive filter is a digital filter that can change the transfer function in a self-adaptive manner based on an input signal. That is, the adaptive filter is a filter that self-adapts the frequency response characteristic. The adaptive filter is generally implemented as a digital filter that performs digital signal processing. A predetermined algorithm (eg, an optimization algorithm) is used to maintain the desired performance of the filter (eg, the performance of minimizing a noise component contained in the input signal). A signal obtained based on the filtered signal is fed back to the predetermined algorithm, and the filter coefficient is adaptively changed by the adaptive processing by the algorithm. As a result, the frequency response characteristic of the adaptive filter is changed. During the period in which the adaptive filter is continuously operating, the adaptive processing by the algorithm is performed continuously (intermittently), and the desired performance of the adaptive filter is maintained.

  However, if the adaptive filter operates continuously for a certain period of time, the performance of the adaptive filter may be reduced below a desired level. This is because the pulse wave signal output from the pulse wave sensor contains a pulsation signal (stationary component) and a noise signal (non-stationary component) including a body movement noise signal derived from the body movement of the subject. It is caused by doing. That is, the performance of the adaptive filter changes following the pulsation signal (stationary component), but may also follow the noise signal. The level of the pulsation signal as well as the noise component (signal amplitude) varies with time, and there may be sudden and non-stationary fluctuations, for example. Therefore, for example, when the level of the pulsation component suddenly decreases and the level of the noise component increases at the same time, the performance of the adaptive filter changes following the noise component. Since the adaptive filter selects and outputs a signal with high autocorrelation from the input signal, the follow-up to the noise gradually progresses with the passage of time, and the noise signal gradually passes. In this case, after a certain amount of time has passed, the performance of the adaptive filter is considerably degraded, and it becomes difficult to return to a normal state.

  For example, a wristwatch-type pulsation detecting device is used by being worn on the wrist of a subject, and is used continuously for a long time, for example, to record changes in the operating state of the subject over time. There is. Therefore, it is preferable that the performance of the adaptive filter does not deteriorate even when continuously used for a long time.

  Thus, in this aspect, an adaptive filter reconstruction unit is provided to reconfigure the adaptive filter at an appropriate timing. That is, the adaptive filter reconstruction unit sets the state of the adaptive filter at a first time point in the period in which the pulse wave signal filtering unit is continuously operating, the time point before the first time point, and the adaptive filter It is reconfigured to the state of the second time point including the time point when the operation is started.

  Here, the filter can be reconfigured by, for example, initializing the filter coefficient (returning the filter coefficient value back to the filter coefficient value at the start of operation) or changing the filter coefficient value in the past during the filter operation period. Setting the filter coefficient value at the time point (for example, when the preferred filter coefficient has been obtained). That is, the second time point is a time point before the first time point including the operation start time point. Further, for example, a case where a pair of adaptive filters is prepared and one filter is switched to the other filter is also included in the filter reconstruction.

  Even if the performance of the filter deteriorates due to continuous use of the adaptive filter (for example, even if the follow-up to the noise component becomes dominant and the beat signal does not follow), By performing the reconstruction of the adaptive filter, the adaptive filter can start a new adaptation process after the reconstruction, resulting in self-adaptation to capture the beat signal. Therefore, the adaptive filter can return to a normal state following the pulsation signal.

  In this aspect, since an adaptive filter is used, it is not necessary to use a plurality of bandpass filters. Further, the reconfiguration of the filter can be realized by initializing the filter coefficient, etc., and the processing load on the apparatus is small, and the increase in power consumption of the apparatus is not particularly problematic.

  Therefore, according to this aspect, for example, it is possible to realize a pulsation detection device (for example, a pulsation detection device that can withstand long-time use) that can perform effective noise countermeasures while reducing the processing burden on the device. it can. In addition, for example, in a pulsation detection device of a type in which a pulse wave sensor is attached to a portion where it is difficult to acquire a pulse wave signal, such as the outside of the wrist (a portion that contacts the back surface of the wristwatch), a pulsation presentation that represents the pulsation signal The specific performance of the spectrum can be improved.

  (2) In another aspect of the pulsation detecting device of the present invention, from the time when the adaptive filter starts to operate or the time immediately before the adaptive filter is reconfigured to the third time when the first time has elapsed. The second period when the first period in which the reconfiguration of the adaptive filter is unnecessary and the second period in which the second time elapses from the third time point to the fourth time point when the second time has elapsed is set as the second period. The time point is set in the first period, and the first time point is set in the second period.

  In this aspect, the first period in which the reconfiguration of the adaptive filter is unnecessary and the second period in which the adaptive filter can be reconfigured are set, and the adaptive filter is reconfigured in the second period. In addition, the state of the adaptive filter changes to the state of the past first time point due to the reconstruction of the adaptive filter, and this first time point is set in the first period. After reconfiguration, the adaptive filter can start a new adaptation process from the preferred state, which results in self-adaptation to capture the beat signal. Therefore, the adaptive filter can maintain appropriate performance continuously.

  In the following aspects (4) to (7), whether or not the adaptive filter can be reconfigured is determined based on a predetermined condition. In this case, the predetermined condition may not be satisfied in the second period (a period during which the adaptive filter can be reconfigured). In this case, the adaptive filter can be forcibly reconfigured when the fourth time point, which is the end point of the second period, is reached, regardless of whether or not the predetermined condition is satisfied. In this way, an unfavorable situation in which the period during which the adaptive filter cannot be reconstructed continues for a long time does not occur.

  (3) In another aspect of the pulsation detecting device of the present invention, the first time point is set at a predetermined cycle with reference to a time point when the adaptive filter starts operation.

  In this aspect, the reconfiguration of the adaptive filter is executed periodically (periodically) at a predetermined cycle with the operation start time of the adaptive filter (pulse wave signal filtering unit) as a reference. In this aspect, the timing of reconfiguring the adaptive filter can be easily managed by a timer, for example, and is easy to implement.

  (4) In another aspect of the pulsation detecting device of the present invention, the adaptive filter reconstruction unit detects the pulsation presenting spectrum by the frequency analysis processing at the nearest time point before the first time point. The adaptive filter is reconfigured on the condition that

  The reconstruction of the adaptive filter is performed in the second period in which the adaptive filter can be reconstructed. However, in this aspect, the filter reconstruction cannot be performed at any time in the second period, and a predetermined condition is satisfied. Can be done. That is, in this aspect, the condition for adaptive filter reconstruction is that the pulsation presentation spectrum is detected by the previous frequency analysis by the pulse wave frequency analysis unit.

  That is, the pulse wave frequency analysis unit performs a frequency analysis process at predetermined time intervals (for example, every 4 seconds) based on the filtered signal output from the pulse wave signal filtering unit, and specifies a pulsation presentation spectrum. If the adaptive filter reconstruction unit is to execute the filter reconstruction process at the first time point, the frequency analysis at the most recent time point before the first time point (that is, the previous frequency analysis process based on the first time point) ), It is confirmed whether the pulsation presentation spectrum (frequency spectrum indicating the pulsation period and its signal intensity) has been successfully detected. When the condition is satisfied, the adaptive filter is reconfigured at the first time point. To do.

  When the adaptive filter is reconfigured, the adaptive filter returns to the state at the second time point in the past, and newly starts an adaptive operation. At the start of this new operation, for example, if the noise mixed in the pulse wave signal is small and the cleanliness of the waveform of the pulse wave signal is within an acceptable range, the adaptive filter performs appropriate self-adaptation. Is possible. On the other hand, when there is a lot of noise mixed in the pulse wave signal and the degree of cleanliness of the waveform of the pulse wave signal is outside the allowable range, the adaptive filter is likely to follow the noise. Since the reconstruction of the adaptive filter is also a process of breaking the tracking of the signal that is captured at that time and considered to have a high autocorrelation, if the capture of the pulsation presentation spectrum immediately after the reconstruction fails, for example, The result is that the pulsation signal component that has been detected until just before the first time point is cut, which is rather bad.

  Here, when the pulsation signal component has been successfully detected in the previous frequency analysis, it can be estimated that the cleanliness of the pulse wave signal at the first time point is within the allowable range. . Therefore, in this aspect, the condition for adaptive filter reconstruction is that the pulsation presentation spectrum is detected by the previous frequency analysis.

  (5) In another aspect of the pulsation detection device of the present invention, the pulsation detection device further includes a signal evaluation unit that determines a degree of noise amount of the noise signal based on a frequency analysis result of the pulse wave signal, In the component, the pulsation presentation spectrum is detected by the frequency analysis processing at the most recent time before the first time, and the noise amount is equal to or less than a predetermined reference by the signal evaluation unit. The adaptive filter is reconfigured on the condition that it is determined as follows.

  In this aspect, in addition to the condition in the above aspect (4), the signal evaluation unit determines that the amount of noise mixed in the pulse wave signal (the amount of disturbance noise component) is equal to or less than a predetermined reference. Was used as a filter reconstruction condition. By determining the amount of noise by the signal evaluation unit, the degree of cleanliness of the pulse wave signal can be estimated more accurately.

  That is, by reconfiguring the filter when there is little disturbance noise, appropriate self-adaptation is executed by the reconstructed adaptive filter, and the pulsation component can be distinguished from the noise component. Therefore, it is possible to more accurately determine whether or not the adaptive filter can be reconfigured.

  (6) In another aspect of the pulsation detection device of the present invention, the adaptive filter reconstruction unit includes a signal evaluation unit that determines a motion state of the subject based on a frequency analysis result of the pulse wave signal. Is that the pulsation presenting spectrum is detected by the frequency analysis processing at the most recent time before the first time point, and the motion state of the subject is stable by the signal evaluation unit. The adaptive filter is reconfigured on the condition that it has been determined.

  The state of motion of the subject (human or animal) can change constantly. In a steady motion state, the pulsation signal is stable, but when the motion state suddenly changes (transitional phase of the motion state change), the amplitude and frequency of the pulsation signal become unstable, for example There may be a sudden change. In such a case, if the reconstruction of the adaptive filter is executed, there is a high possibility that the adaptive process will follow not only the pulsation signal component but also the noise component.

  Therefore, in this aspect, the condition that the movement state of the subject is confirmed to be stable is added as a result of the frequency analysis by the signal evaluation unit to the condition of the above aspect (4) (weighting). ) Therefore, it is possible to more accurately determine whether or not the adaptive filter can be reconfigured.

  For example, when the movement state of the person who is the subject is in a state that does not correspond to either a steady movement such as walking or jogging, or a resting state (for example, a ball game, In the case of a state of gymnastics, the adaptive filter is not reconfigured. Therefore, the failure of pulsation detection and the occurrence of false detection can be reduced.

  (7) In another aspect of the pulsation detection device of the present invention, a signal evaluation unit that determines the degree of noise amount of the noise signal and the motion state of the subject based on the frequency analysis result of the pulse wave signal. The adaptive filter reconstruction unit detects the pulsation presentation spectrum by the frequency analysis process at the most recent time before the first time point, and the signal evaluation unit sets the noise amount to a predetermined value. The adaptive filter is reconfigured on the condition that it is determined that the motion condition of the subject is in a stable state.

  In this aspect, the above condition (5) is further added (weighted) to the condition of the above aspect (6) as a condition for adaptive filter reconstruction. This makes it possible to more accurately determine whether the adaptive filter can be reconfigured.

  (8) In another aspect of the pulsation detecting device of the present invention, the pulse wave signal filtering unit includes a delay processing unit that delays the pulse wave signal for a predetermined time, and a filter coefficient update that updates a coefficient of the adaptive filter. The adaptive filter outputs a first signal having a high autocorrelation among the output signals of the delay processing unit, and the subtraction unit extracts the pulse wave signal from the pulse wave signal. Subtracting the first signal to generate a second signal having a lower autocorrelation than the first signal, supplying the second signal to the filter coefficient update unit, and the filter coefficient update unit includes the second signal The coefficient of the adaptive filter is updated so that the signal is suppressed, and the adaptive filter reconstruction unit converts the coefficient of the adaptive filter into the adaptive filter when the adaptive filter reconstruction unit reconfigures the adaptive filter at the first time point. Setting the coefficient at the time the coater has started operation.

  In this aspect, the pulse wave signal filtering unit includes a delay processing unit, an adaptive filter, a filter coefficient update unit, and a subtraction unit. The filter coefficient updating unit suppresses (for example, minimizes) the filter coefficient so that the second signal output from the subtraction unit (a signal with low correlation and may be referred to as an error signal) is suppressed. Update. Further, in this aspect, the adaptive filter reconstruction unit initializes the filter coefficient when performing the reconstruction of the adaptive filter. That is, the filter coefficient is set to the filter coefficient value at the start of the operation of the adaptive filter (for example, the filter coefficient value is set to all zeros).

  According to this aspect, the adaptive filter included in the pulse wave signal filtering unit can be reinitialized at an appropriate timing. Therefore, even when continuous measurement is performed for a long time, the performance of the adaptive filter can be maintained at an appropriate level. Thereby, the failure of pulsation detection and the occurrence of false detection can be reduced.

  Thus, according to at least one of the above-described aspects of the present invention, for example, by reconfiguring the adaptive filter that constitutes the pulse wave signal filtering unit at an appropriate timing, the performance of the adaptive filter is brought to an appropriate level. Can be maintained. In addition, for example, the effective use of the adaptive filter can simplify the filter configuration, and it is relatively easy to reconfigure the adaptive filter. Therefore, it is effective while reducing the processing load on the apparatus. It is possible to realize a pulsation detecting device that can execute various noise countermeasures.

The figure which shows the structure of an example of the pulsation detection apparatus of this invention. 2A to 2C are diagrams illustrating an example of the configuration of a pulse wave sensor and a pulsation detection device. FIGS. 3A to 3C are diagrams showing an example of the waveform and frequency spectrum of the signal before FFT for explaining the determination operation of the degree of cleanness of the pulse wave signal (degree of noise). The flowchart which shows an example of the determination process of the exercise state of the subject The figure which shows an example of the timing when the reconstruction of an adaptive filter is performed Flow chart showing an example of the procedure of adaptive filter reconfiguration processing The flowchart which shows an example of the procedure of adaptive filtering processing (pre-processing) and frequency analysis processing (post-processing) FIGS. 8A to 8C are diagrams showing an example of a pulse wave signal waveform in actual processing and a frequency spectrum obtained by FFT. FIGS. 9A and 9B correspond to the time lapse of the S / N index (SN3) in each of the example in which the adaptive filter reconstruction processing is executed and the case in which the adaptive filter reconstruction processing is not performed (comparative example). Diagram showing the results of examining fluctuations

  Hereinafter, this embodiment will be described. In addition, this embodiment demonstrated below does not unduly limit the content of this invention described in the claim. In addition, all the configurations described in the present embodiment are not necessarily essential configuration requirements of the present invention.

(First embodiment)
(Overall configuration example)
FIG. 1 is a diagram showing a configuration of an example of a pulsation detecting device of the present invention. A pulsation detection device 100 shown in FIG. 1 is a kind of sensor device that detects pulsation signals derived from the pulsation of a subject (including humans and animals) and biological information such as heartbeats corresponding to the pulsation signals. It is.

  Here, the pulsation means a movement that occurs when the periodical contraction and relaxation of not only the heart but also the internal organs are repeated medically. Here, the movement as a pump in which the heart periodically sends blood is called pulsation. The heart rate refers to the number of heart beats per minute. The pulse rate refers to the number of pulsations in the peripheral blood vessels. When the heart pumps out blood, pulsation occurs in the artery, and this number is called the pulse rate or simply the pulse. As long as the pulse is measured with an arm, it is usually referred to as a pulse rate instead of a heart rate in medical terms. Moreover, in the following description, the term body movement is used. Body movement, in a broad sense, means everything that moves the body. Further, steady and periodic body movements of the subject may be referred to as narrow body movements. The narrowly defined body movement is, for example, a steady and periodic movement of the arm (near the site where the pulse meter is worn) accompanying walking, jogging, or the like.

  The pulsation detecting device 100 shown in FIG. 1 includes a pulse wave sensor 10, a pulse wave signal accumulation unit (a first buffer memory 13 for accumulating data of a pulse wave signal d for 4 seconds, and a pulse wave signal for 16 seconds. a second buffer memory 15 for storing data d), a pulse wave signal filtering unit 200 including an adaptive filter reconstruction unit 17, a body motion sensor (acceleration sensor, gyro sensor, etc.) 11, and body motion noise It includes a removal processing unit 300, a pulse wave frequency analysis unit 400, and a display unit (including a liquid crystal panel) 600 that displays detection results and the like.

  The pulse wave sensor 10 is, for example, a photoelectric pulse wave sensor and a pulse wave sensor based on the principle thereof. The pulse wave sensor 10 outputs a pulse wave signal d in which a pulsation signal and a noise signal including a body motion noise signal derived from the body motion of the subject (human or animal) are mixed.

  Here, the pulse wave signal d includes, for example, a pulsation component signal (stationary component or periodic component), a body motion noise component (stationary or periodic component), and a disturbance noise component (impact noise or the like). Stationary or aperiodic component).

  A signal for 4 seconds of the pulse wave signal d output from the pulse wave sensor 10 is accumulated in the first buffer memory 13. The pulse wave signal d for 4 seconds is transferred to the second buffer memory 15 at a cycle of 4 seconds. The second buffer memory 15 is a FIFO (first-in first-out) memory, and the pulse wave signal for 16 seconds is updated every 4 seconds. The pulse wave signal for 16 seconds is accumulated because it is necessary to observe the transition of the signal in a certain time width and carefully examine the presence or absence of correlation when identifying the pulsation component by frequency analysis. is there.

  The pulse wave signal filtering unit 200 is a kind of adaptive filter capable of separating and outputting a stationary frequency component and other non-stationary components when the input signal includes a stationary frequency component.

  The pulse wave signal filtering unit 200 includes a delay processing unit 201 that delays the pulse wave signal for a predetermined time (here, one sample time), and a first signal y having high autocorrelation among output signals of the delay processing unit 201. Filter 202 for selecting and outputting the filter, filter coefficient updating unit (nLMS filter coefficient updating unit) 210 for updating the coefficient of adaptive filter 202, and subtracting first signal y from pulse wave signal d to obtain the first signal a subtractor 204 that generates a second signal e (= d−y) having lower autocorrelation than y and supplies (feeds back) the second signal e to a filter coefficient update unit (nLMS filter coefficient update unit) 210; have.

  A functional block including the delay processing unit 201, the adaptive filter 202, and the filter coefficient update unit (nLMS filter coefficient update unit) 210 may be referred to as an adaptive line spectrum enhancer.

  The nLMS filter coefficient updating unit 210 adjusts the coefficient of the adaptive filter (in this case, the normalized minimum mean square coefficient (nLMS coefficient)) so that the value of the second signal e is suppressed (for example, minimized). Are updated adaptively.

  Each of the first signal y and the second signal e (= d−y) is multiplied by each of the gain coefficients h1 and h2 (this processing is performed by coefficient multipliers 207a and 207b included in the gain adjustment unit 206). Then both signals are summed by summer 208. For example, h1 ≧ 1.0 is set and h2 <1.0 is set. According to this, while being able to reduce the influence by an impact, the followability to the sudden change of a pulsation component or a body motion component can be improved. For example, when a person who has a resting pulse of 60 starts jogging a fast pace and the pulse suddenly rises to 150, the follow-up performance of the adaptive filter 202 is delayed from the rise of the pulse. The adaptive filter 202 may block the signal of the rapidly increasing heart rate component. This can be eliminated.

  The adaptive filter 202 can change the transfer function in a self-adaptive manner based on the input signal. The adaptive filter 202 is generally implemented as a digital filter that performs digital signal processing. A predetermined algorithm (eg, an optimization algorithm) is used to maintain the desired performance of the filter (eg, the performance of minimizing a noise component contained in the input signal). A signal obtained based on the filtered signal is fed back to the predetermined algorithm, and the filter coefficient is adaptively changed by the adaptive processing by the algorithm. As a result, the frequency response characteristic of the adaptive filter is changed. During the period in which the adaptive filter is continuously operating, the adaptive processing by the algorithm is performed continuously (intermittently), and the desired performance of the adaptive filter 202 is maintained.

  However, if the adaptive filter 202 is continuously operated for a certain period of time, the performance of the adaptive filter 202 may be reduced below a desired level. This is because the pulse wave signal d output from the pulse wave sensor 10 includes a pulsation signal (stationary component) and a noise signal (non-stationary component) including a body movement noise signal derived from the body movement of the subject. This is due to the mixture. That is, the performance of the adaptive filter 202 changes following the pulsation signal (stationary component), but may also follow the noise signal. The level of the pulsation signal as well as the noise component (signal amplitude) varies with time, and there may be sudden and non-stationary fluctuations, for example. Therefore, for example, when the level of the pulsating component suddenly decreases and at the same time the level of the noise component increases, the performance of the adaptive filter 202 changes following the noise component. Since the adaptive filter 202 basically selects and outputs a signal having high autocorrelation from the input signal, the tracking of the noise gradually progresses over time, and the noise signal gradually increases. Let it pass. In this case, after a certain amount of time has elapsed, the performance of the adaptive filter 202 may be significantly degraded and it may be difficult to return to a normal state.

  Therefore, in the present embodiment, the adaptive filter reconstruction unit 17 performs the reconstruction of the adaptive filter 202 at an appropriate timing. As shown in FIG. 1, the adaptive filter reconstruction unit 17 includes a fast Fourier transform unit 212 that performs FFT (Fast Fourier Transform), fast Fourier transform units 213 and 215 that can be provided as necessary, and a signal. An evaluation unit 214, a timing determination unit (timer) 216, and a filter reconstruction processing unit 218 are included.

  The fast Fourier transform unit 212 performs fast Fourier transform on the pulse wave signal d. When the fast Fourier transform units 213 and 215 are provided, the fast Fourier transform unit 213 performs fast Fourier transform on the acceleration signal (X-direction component) output from the body motion sensor 11, and the fast Fourier transform unit 215 The acceleration signal (Y direction component) output from the motion sensor 11 is fast Fourier transformed.

  The signal evaluation unit 214 determines the degree of cleanliness of the pulse wave signal d (that is, the amount of noise) as basic information for determining whether or not the adaptive filter 202 can be reconfigured. (Noise amount degree determination unit) 217 and an exercise state determination unit 219 for determining the exercise state of the subject. The operations of the signal cleanness determination unit (noise amount degree determination unit) 217 and the motion state determination unit 219 will be described later with reference to FIGS. 2 and 3.

  The timing determination unit (timer) 216 outputs timing information necessary for determining the reconfiguration timing of the adaptive filter 202. The filter reconstruction processing unit 218 reconfigures the adaptive filter 202 when it satisfies a temporal condition and a predetermined substantive condition related to the degree of cleanness of the pulse wave signal.

  In the filter reconstruction processing unit 218, information on whether or not the pulsation presentation spectrum (pulsation component) has been detected from the pulsation component specifying unit 408 included in the pulse wave frequency analysis unit 400 includes the pulsation component. Supplied every time detection processing is performed. The filter reconstruction processing unit 218 is provided with a memory (not shown), and information on the success / failure of detection of the pulsation presentation spectrum is accumulated in the memory as history information. For example, the filter reconstruction processing unit 218 reconstructs the adaptive filter 202 based on timing information from the timing determination unit (timer) 216, information on the success / failure of the detection of the pulsation presentation spectrum in the previous detection, and the like. The adaptive filter 202 is reconfigured when it is determined that it is possible (for example, the filter coefficient is initialized as all zeros).

  For example, the filter reconstruction processing unit 218 sets the state of the adaptive filter 202 to a time point before the first time point at a first time point in a period in which the pulse wave signal filtering unit 200 is continuously operating, and The state is changed to the state at the second time point including the time point when the adaptive filter 202 starts operation. Here, the reconstruction of the adaptive filter 202 includes, for example, initialization of the filter coefficient (returning the filter coefficient value to the filter coefficient value at the start of the operation) and the filter coefficient value during the operation period of the filter. Is set to the filter coefficient value at a past time point (for example, when a preferable filter coefficient has been obtained). That is, the second time point is a time point before the first time point including the operation start time point. Further, for example, a case where a pair of adaptive filters is prepared and one filter is switched to the other filter is also included in the filter reconstruction.

  Even if the performance of the filter is deteriorated by continuously using the adaptive filter by performing the reconstruction of the adaptive filter 202 (for example, the pulsation signal is obtained by following the noise component dominantly). The adaptive filter can start a new adaptation process after reconfiguration, resulting in self-adaptation to capture the beat signal. Therefore, the adaptive filter 202 can return to a normal state following the pulsation signal. Details will be described later.

  Further, in FIG. 1, a body motion sensor 11 is a sensor that detects a body motion of a subject, for example, a steady, periodic arm (near the site where the pulse meter is attached) accompanying walking or jogging, For example, an acceleration sensor or a gyro sensor can be included.

  The body motion noise removal processing unit 300 includes a first adaptive filter 302 and a second adaptive filter 304 for body motion noise removal. Body motion noise (body motion noise component) is caused by steady motion or motion (body motion) such as a human being included in the pulse wave signal (more precisely, the output signal of the summing unit 208). This is a noise component indicating a change in the volume of the blood vessel. In the case of a pulse meter worn on an arm or finger, the volume of the blood vessel changes in accordance with the rhythm of the arm swing due to the influence of the arm swing while walking or jogging. When a human performs a steady operation, the pulse wave signal becomes a component signal having the frequency of the operation. It has been found that the body motion noise component has a high correlation with the waveform of the signal output from the acceleration sensor mounted near the pulse wave sensor mounting site.

  The pulse wave frequency analysis unit 400 includes a fast Fourier transform unit 402 to which a pulse wave signal after removal of body motion noise is input, and a fast Fourier to which an acceleration signal (X-axis direction component) from the body motion sensor 11 is input. A conversion unit 404, a fast Fourier transform unit 406 to which an acceleration signal (Y-axis direction component) from the body motion sensor 11 is input, a pulsation component identification unit (a pulsation presentation spectrum identification unit) 408, and a pulse rate calculation unit 410, and may further include an exercise state determination unit 412 as needed.

  The pulsation component identification unit (pulsation presentation spectrum identification unit) 408 performs frequency analysis on the pulse wave signal for 16 seconds after FFT every 4 seconds, and based on the spectrum value, spectrum distribution, etc. in the past The correlation with the obtained pulsation component is examined, and the pulsation presentation spectrum is specified. The pulsation presentation spectrum is a frequency spectrum indicating a pulsation period and its signal intensity among frequency spectra obtained as an FFT result of a pulsation component signal for a certain period. The body motion presentation spectrum is a frequency spectrum indicating the cycle of body motion (for example, arm swing during walking) and the signal intensity among the spectrum obtained as the FFT result of the body motion noise component signal for a certain period.

  The pulse rate calculation unit 410 calculates the pulse rate. If the position (frequency) of the pulsation presentation spectrum on the frequency axis is determined, the pulse rate is uniquely determined corresponding to the position of the spectrum. The motion state determination unit 412 determines the motion state (steady motion state, unsteady motion state, rest state, etc.) of the subject based on the pulse wave signal or the body motion signal. The display unit 600 can display the detected pulse rate and the waveform indicating the pulsation, the detected exercise state, the calorie consumption of the subject, the current time, and the like.

  In the pulsation detection device 100 shown in FIG. 1, the reconstruction of the adaptive filter 202 is executed at an appropriate timing. Therefore, even if the performance of the filter deteriorates due to continuous use of the adaptive filter 202 (for example, a case where the pulsation signal is not followed occurs due to the dominant follow-up to the noise component). Even so, the adaptive filter 202 can start a new adaptation process after reconfiguration, resulting in self-adaptation to capture the beat signal. Therefore, the adaptive filter 202 can return to a normal state following the pulsation signal.

  In the pulsation detection device 100 of FIG. 1, since the adaptive filter 202 is used, it is not necessary to use a plurality of bandpass filters. Further, the reconfiguration of the filter can be realized by initializing the filter coefficient, etc., and the processing load on the apparatus is small, and the increase in power consumption of the apparatus is not particularly problematic.

  Therefore, according to the present embodiment, for example, the pulsation detection device 100 (for example, the pulsation detection device 100 that can withstand long-time use) that can execute effective noise countermeasures while reducing the processing burden on the device is realized. can do. In addition, for example, in a pulsation detection device of a type in which the pulse wave sensor 10 is attached to a portion where it is difficult to obtain a pulse wave signal, such as the outside of the wrist (a portion that contacts the back surface of the wristwatch), the specification of the pulsation presentation spectrum is performed. The performance can be improved.

(Example of wrist watch type pulse meter)
FIG. 2A to FIG. 2C are diagrams illustrating an example of a configuration of a pulse wave sensor and a pulsation detection device. FIG. 2A shows an example of a cross-sectional structure of the pulse wave sensor 10. A pulse wave sensor 10 shown in FIG. 2A includes a light emitting element 1 provided on the back surface (lower surface) of a substrate 3, a transparent cover (a contact member made of a light transmissive material) 2, a light It has a reflective dome (light reflecting portion) 4 and a light receiving portion 5 provided on the surface (upper surface) of the substrate 3. The light R1 emitted from the light emitting element 1 reaches the blood vessel (biological information source) O at the detection site SA (here, the wrist) and is reflected. Since the volume of the blood vessel O periodically varies with the pulsation, the intensity of the reflected light R1 ′ varies periodically according to the pulsation. The reflected light R 1 ′ is reflected by the reflecting dome 4 and then enters the light receiving unit 5. The light receiving unit 5 converts incident light into an electrical signal.

  FIG. 2B shows an example of a circuit configuration in the pulse wave sensor 10. Light emission of the light emitting unit 1 is controlled by the control circuit 161. The light reception signal output from the light receiving unit 5 is amplified by the amplifier circuit 162 and then converted into a digital signal by the A / D conversion circuit 163. In this way, the pulse wave signal d is obtained.

  FIG. 2C shows an example of use of a wristwatch-type pulsometer (incorporating the pulsation detecting device of the present embodiment). The pulsometer has a wristband 500, a display unit 600, and a pulsation detecting device 100 (built-in) of the present embodiment. In the example of FIG. 2C, the pulsometer (that is, the pulsation detecting device 100) is attached to the left wrist of the person (user) who is the subject.

(Determining the cleanliness of the pulse wave signal (degree of noise) and determining the motion state of the subject)
When determining whether or not the adaptive filter 202 can be reconfigured, it may be necessary to determine the degree of cleanliness of the pulse wave signal (degree of noise), the state of motion of the subject, and the like. Hereinafter, these determination operations will be described with reference to FIGS.

  First, determination of the degree of cleanliness of the pulse wave signal (degree of noise amount) will be described. FIGS. 3A to 3C are diagrams showing an example of the waveform and frequency spectrum of the signal before FFT for explaining the determination operation of the degree of cleanness of the pulse wave signal (degree of noise). It is.

  3A to 3C, the signal waveform of the pulse wave signal d before the FFT for 16 seconds is shown on the upper side. The horizontal axis represents time, and the vertical axis represents signal amplitude. On the lower side, a frequency spectrum in a frequency band of 0 to 4 Hz is shown. The horizontal axis indicates the frequency, and the vertical axis indicates the spectrum value.

  Here, FIG. 3 (A) shows the waveform and frequency spectrum of the pulse wave signal d when the noise is low (clean), and FIG. 3 (B) shows the pulse wave when the noise is moderate (so-so). The waveform and frequency spectrum of the signal are shown. FIG. 3C shows the waveform and frequency spectrum of the pulse wave signal d when the pulse wave signal d contains a lot of noise (noisy). As is clear from the comparison between FIGS. 3A to 3C, the waveform of the pulse wave signal d and the frequency spectrum are closely related, and the frequency of the pulse wave signal corresponds to the waveform of the pulse wave signal. The distribution state and spectrum value of the spectrum change. Therefore, it is possible to estimate the state of noise to be superimposed on the pulse wave signal d (degree of noise amount) based on the frequency spectrum obtained by FFT.

  In the present embodiment, a ratio of spectrum values of main frequency spectra (that is, a ratio of the height of the baseline) is used as an index for estimating the degree of noise. Specifically, the indices r5 and r10 are used (however, they are only examples, and other statistical indices such as standard deviation may be used). Here, r5 is the frequency spectrum of the pulse wave signal for 16 seconds, when five spectra are arranged in order of the magnitude of the peak value (that is, when sorting), This is an index obtained by using the spectrum value (power) as the denominator and the spectrum value (power) of the fifth spectrum as the numerator.

  R10 is the spectrum value of the first spectrum when 10 spectra are arranged in order of the peak value from the frequency spectrum of the pulse wave signal for 16 seconds (that is, when sorting). This is an index obtained by using (power) as the denominator and the spectrum value (power) of the tenth spectrum as the numerator.

  Here, as an example, when r5 <0.5 and r10 <0.2, the noise is small (clean), and when r5> 0.7 and r10 <0.5, the noise is noisy. If it is not, the noise is moderate.

  In the example of FIG. 3A, since r5 = 0.14 and r10 = 0.08, it is determined that the noise is small (clean). In the example of FIG. 3B, since r5 = 0.56 and r10 = 0.35, it is determined that the noise is moderate. In the example of FIG. 3C, since r5 = 0.82 and r10 = 0.62, it is determined that there is a lot of noise (noisy).

  Such determination processing is executed by the signal cleanliness determination unit (noise amount degree determination unit) 217 included in the signal evaluation unit 214 shown in FIG.

  In addition, the motion state of the subject can also be estimated using the above-described indices r5 and r10. This is because when the waveform of the pulse wave signal d changes depending on the motion state of the subject, the change appears as a change in the frequency spectrum, and the change in the frequency spectrum is reflected in the indices r5 and r10.

  The exercise state of the subject is executed by the exercise state determination unit 219 included in the signal evaluation unit 214 shown in FIG. The pulse wave frequency analysis unit 400 that performs post-processing also includes an exercise state determination unit 412. The exercise state determination unit 412 also performs similar processing to determine the exercise state of the subject. .

  FIG. 4 is a flowchart illustrating an example of the determination process of the exercise state of the subject. First, 256 samples of pulse wave signal data corresponding to 16 seconds are acquired (step ST1). Next, the FFT processing is performed on the pulse wave signal data (step ST2).

  Subsequently, for example, the signal cleanliness determination unit (noise amount degree determination unit) 217 included in the signal evaluation unit 214 executes a frequency spectrum sorting process (step ST3). Here, the sorting process is, for example, a process of arranging a plurality of frequency spectra in descending order of spectrum values. Next, indices r5 and r10 are calculated (step ST4).

  Next, the signal cleanness determination unit (noise amount level determination unit) 217 determines whether r5> 0.7 and r10 <0.5, that is, whether the noise is in a noisy state. (Step ST5). If so (in the case of Y), since there is a lot of unsteady noise, it is determined that the unsteady movement is being performed (step ST6).

  In step ST5, in the case of N, the signal cleanliness determination unit (noise amount degree determination unit) 217 satisfies r5 <0.5 and r10 <0.2, that is, noise is small (clean. ) Is determined (step ST7). If so (in the case of Y), next, the body movement (for example, body movement with periodicity such as arm swing) is performed within the top 10 spectrum sorting results in which the frequency spectrum is sorted in step ST3. It is determined whether there is a corresponding frequency spectrum (step ST8). In this determination, the accuracy of the determination is improved by specifying the body motion spectrum with reference to the body motion signal obtained from the body motion sensor 11. In this case, the configuration indicated by the broken line in FIG. 1 (the configuration in which the acceleration signals in the X and Y directions output from the body motion sensor 11 are supplied to the fast Fourier transform units 213 and 215) is used. be able to.

  In step ST8, in the case of Y, since it is in a state of low noise (clean) and a body motion spectrum can be specified, it is determined that it is during steady exercise (for example, walking at the same speed) (step ST9). ). In step ST8, in the case of N, since it is a state of low noise (clean) and no body motion spectrum is found, it is determined to be a resting state (step ST10).

  In Step ST7, in the case of N, it is next determined whether there is a frequency spectrum corresponding to body movement (for example, periodic body movement such as arm swing) within the top 10 spectrum sorting results. (Step ST11). In this determination, the accuracy of the determination is improved by specifying the body motion spectrum with reference to the body motion signal obtained from the body motion sensor 11. In this case, the configuration indicated by the broken line in FIG. 1 (the configuration in which the acceleration signals in the X and Y directions output from the body motion sensor 11 are supplied to the fast Fourier transform units 213 and 215) is used. be able to.

  In step ST11, if Y, the amount of noise is moderate (so-so) and the body motion spectrum can be specified. Therefore, during steady motion, but there are unsteady elements (for example, the same It is determined that the vehicle is walking at a speed but is in an unsteady state such as watching a wristwatch (step ST12).

  In Step ST11, when N, the amount of noise is moderate (so-so) and the body motion spectrum is not found. Therefore, while resting, there is an unsteady element (for example, It is determined that the person is lying on the bed but is performing an unsteady operation such as turning over (step ST13).

  In this way, the motion state of the subject can be estimated. However, the above-mentioned example is an example and is not limited to this.

(Regarding the timing conditions for adaptive filter reconfiguration timing)
FIG. 5 is a diagram illustrating an example of timing at which adaptive filter reconstruction is executed. For example, a wristwatch-type pulsation detecting device as shown in FIG. 2 is used by being worn on the wrist of a subject. For example, in order to record changes in the operating state of the subject over time, May be used continuously. Therefore, it is preferable that the performance of the adaptive filter 202 is not deteriorated even when used continuously for a long time. Therefore, the adaptive filter 202 is reconfigured at an appropriate timing during the period in which the adaptive filter 202 is continuously operating.

  In FIG. 5, the first reconfiguration of the adaptive filter 202 is unnecessary from the time point when the adaptive filter starts operating (time t1) to the third time point (time t2) when the first time (for example, 10 minutes) has elapsed. The period TA is defined as the second period TC in which the adaptive filter can be reconfigured from the third time point (time t2) to the fourth time point (time t4) when the second time (for example, 5 minutes) has elapsed. Note that a period TB from time t1 to time t4 is the maximum period in which the adaptive filter 202 can be used without reconfiguration.

  The timing (first time point) at which the adaptive filter 202 is reconfigured is set in the second period TC. The state of the adaptive filter 202 returns to the state of the past second time point as a result of the reconstruction, and this second time point is set in the first period TA.

  The adaptive filter reconfiguration unit 17 determines whether or not the filter can be reconfigured from time t2 in consideration of other conditions (described later). If it is determined that the reconfiguration is possible, the adaptive filter reconfiguration unit 17 executes the reconfiguration process of the adaptive filter 202. In the example of FIG. 5, the first reconfiguration of the adaptive filter 202 is executed at time t3. Time t3 is the first timing that satisfies a predetermined condition (determination requirement) for determining whether filter reconstruction is possible. After reconfiguration, the adaptive filter 202 can start a new adaptation process from a favorable state, and as a result, self-adaptation is performed so as to capture a pulsation signal. Therefore, the adaptive filter 202 can maintain appropriate performance continuously.

  Note that, as a result of determining whether or not the adaptive filter can be reconfigured according to a predetermined condition (conditions other than the temporal condition: described later with reference to FIG. 6), the predetermined condition is not satisfied, and the state is in the second period. (Adaptive filter reconfigurable period) It may continue over TC. In this case, the fourth time point (time t4) that is the end point of the second period TC is set as the limit time point, and when the fourth time point (time t4) is reached, regardless of whether the predetermined condition is satisfied or not, The adaptive filter 202 can be forced to reconfigure. In this way, an unfavorable situation in which the period during which the adaptive filter 202 cannot be reconstructed continues for a long time does not occur. The above operation is then repeated.

  Next, the first period TA is started from the most recent time (here, time t3) when the adaptive filter is reconfigured. The period from time t5 to time t7 is the second period TC. Time t6 is the second filter reconstruction timing. Such an operation is repeatedly executed in the continuous operation period (TD) of the adaptive filter 202 after time t7. The operation of the adaptive filter 202 ends at time t8.

  Further, the first time point (time t3 in the example of FIG. 5) that is the reconfiguration timing of the adaptive filter 202 is set periodically (periodically) with the operation start time (time t0) of the adaptive filter 202 as a reference. You can also. In this case, the reconfiguration timing of the adaptive filter can be easily managed by, for example, the timing determination unit (timer) 216 (see FIG. 1), and is easy to implement.

(Example of specific processing procedure for filter reconfiguration)
FIG. 6 is a flowchart illustrating an exemplary procedure of adaptive filter reconstruction processing. First, pulse wave signal data corresponding to 256 samples for 16 seconds and body motion signal data are acquired (step ST101). Next, it is determined whether or not the time condition (condition described with reference to FIG. 5) is satisfied (step ST102).

  If the time condition is satisfied in step ST102 (step ST102: Y), next, whether or not the present time has reached the limit time (time t4 or time t7 in FIG. 5), that is, the adaptive filter It is determined whether the accumulated use time of 202 is equal to or longer than the time corresponding to the second period TC (step ST103). If the limit time has been reached (step ST103: Y), the adaptive filter reconstruction unit 17 reconfigures the adaptive filter 202 (step ST105).

  In step ST103, in the case of No, the process proceeds to step ST104, and it is determined whether or not a predetermined condition other than the time condition is satisfied. Various conditions can be considered as the predetermined condition. For example, when the condition I is satisfied, the conditions I and II are satisfied, the conditions I and III are satisfied, and the conditions I, II, and III are all satisfied Etc. can be considered. If the predetermined condition is satisfied in step ST104 (step ST104: Y), the adaptive filter reconstruction unit 17 reconfigures the adaptive filter 202 (step ST105). Hereinafter, the contents of the predetermined conditions I to III and a plurality of examples for determining by combining each condition will be described.

(Judgment example A in step ST104: judgment based on condition I)
The adaptive filter reconstruction unit 17 is conditioned on the condition that the pulsation presenting spectrum is detected by the frequency analysis by the pulse wave frequency analysis unit 400 (see FIG. 1) at the most recent time before the first time point (this is the condition I), the reconstruction of the adaptive filter 202 is performed.

  The reconstruction of the adaptive filter 202 is performed in the second period TC in which the adaptive filter 202 can be reconstructed. However, the reconstruction of the adaptive filter 202 is not always performed in the second period TC, and a predetermined condition is satisfied. Can be done if you meet. Here, the condition (condition I) for adaptive filter reconstruction is that the pulsation presentation spectrum is detected by the previous frequency analysis by the pulse wave frequency analysis unit 400.

  That is, the pulse wave frequency analysis unit 400 performs frequency analysis processing every predetermined time (for example, every 4 seconds) based on the filtered signal output from the pulse wave signal filtering unit, and specifies the pulsation presentation spectrum. . If the adaptive filter reconstruction unit 17 is to execute the filter reconstruction process at the first time point, the frequency analysis at the most recent time point before the first time point (that is, the previous frequency analysis based on the first time point). In the processing), it is confirmed whether or not the pulsation presentation spectrum (the frequency spectrum of the pulsation baseline corresponding to the pulsation) has been successfully detected, and when the condition (condition I) is satisfied, Perform a reconfiguration.

  When the adaptive filter 202 is reconfigured, the adaptive filter 202 returns to the state at the second time point in the past, and newly starts an adaptive operation. At the start of this new operation, for example, if there is little noise mixed in the pulse wave signal d and the degree of cleanliness of the waveform of the pulse wave signal is within an allowable range, the adaptive filter will Adaptation is possible.

(Judgment example B in step ST104: judgment based on condition I and condition II)
On the other hand, when there is a lot of noise mixed in the pulse wave signal d and the degree of cleanliness of the waveform of the pulse wave signal is outside the allowable range, the adaptive filter 202 is likely to follow the noise. Since the reconstruction of the adaptive filter 202 is also a process of cutting off the tracking of the signal that is captured at that time and considered to have a high autocorrelation, if the capture of the pulsation presentation spectrum immediately after the reconstruction fails, for example, As a result, the pulsation signal component that has been detected until just before the first time point is cut off, which may be rather bad.

  The adaptive filter reconstruction unit 17 detects a pulsation presentation spectrum (satisfies condition I) by the frequency analysis by the pulse wave frequency analysis unit 400 at the most recent time before the first time point, and the signal evaluation unit If it is determined by 214 that the amount of noise is less than or equal to a predetermined criterion (condition II), the adaptive filter 202 is reconfigured. In the determination of the noise amount, the determination method described with reference to FIGS. 3A to 3C can be used.

  That is, in this determination example B, in addition to the condition I, the signal evaluation unit 214 determines that the amount of noise mixed in the pulse wave signal d (the amount of disturbance noise component) is not more than a predetermined reference. (Condition II) is also required to be satisfied. By determining the amount of noise by the signal evaluation unit 214, the degree of cleanness of the pulse wave signal d can be estimated more accurately.

  In this example, by reconfiguring the filter when the disturbance noise is low, appropriate self-adaptation is executed by the reconstructed adaptive filter, and the pulsation component can be distinguished from the noise component. Therefore, it is possible to more accurately determine whether or not the adaptive filter can be reconfigured.

(Judgment example C in step ST104: judgment based on condition I and condition III)
In this example, the adaptive filter reconstruction unit 17 has detected a pulsation presenting spectrum by the frequency analysis by the pulse wave frequency analysis unit 400 at the most recent time before the first time (condition I is satisfied). In addition, the adaptive filter 202 is reconfigured by the motion state determination unit 412 (see FIG. 1) when the motion state of the subject is stable (that is, when the condition III is also satisfied).

  The state of motion of the subject (human or animal) can change constantly. In a steady motion state, the pulsation signal is stable, but when the motion state suddenly changes (transitional phase of the motion state change), the amplitude and frequency of the pulsation signal become unstable, for example There may be a sudden change. In such a case, if the reconstruction of the adaptive filter is executed, the possibility of following the noise component instead of the pulsation signal component increases.

  Therefore, in the determination example C, a condition (condition III) that confirms that the motion state of the subject is a stable state is further added (weighted) to the above condition I. Therefore, it is possible to more accurately determine whether the adaptive filter 202 can be reconfigured. In the determination of the movement state of the subject, the determination method described with reference to FIG. 4 can be used.

  For example, when the movement state of the person who is the subject is in a state that does not correspond to either a steady movement such as walking or jogging, or a resting state (for example, a ball game, In the case of a state of physical exercise or the like, the adaptive filter 202 is not reconfigured. Therefore, the failure of pulsation detection and the occurrence of false detection can be reduced.

(Determination example D in step ST104: Determination based on condition I, condition II and condition III)
In this example, the adaptive filter reconstruction unit 17 has detected a pulsation presentation spectrum by frequency analysis by the pulse wave frequency analysis unit at the most recent time before the first time point (satisfies condition I), and the signal When the evaluation unit 214 determines that the amount of noise is equal to or less than a predetermined criterion (satisfies Condition II) and the subject's motion state is a stable state including a steady motion state and a rest state (that is, When the condition III is satisfied), the adaptive filter 202 is reconfigured.

  In the determination example D, it is determined whether or not the conditions I, II, and III are all satisfied as the conditions for reconfiguring the adaptive filter 202. Therefore, it is possible to more accurately determine whether the adaptive filter 202 can be reconfigured.

  Next, procedures of adaptive filtering processing (pre-processing executed by the pulse wave signal filtering unit 200 before frequency analysis) and frequency analysis processing (post-processing) performed by the pulse wave frequency analysis unit 400 will be described.

  FIG. 7 is a flowchart illustrating an example of procedures of adaptive filtering processing (preprocessing) and frequency analysis processing (postprocessing). In the preprocessing, first, the delay wave processing unit 201 delays the pulse wave signal d (step ST106). Next, it is determined whether the number of processed data has reached the total number of samples (step ST107). If the number has not reached (step ST107: Y), processing by the adaptive filter 202 is performed and autocorrelation is performed. The first signal (stationary component) y, which is a highly reliable signal, is output (step ST108). Subsequently, a signal e with low autocorrelation is output by subtracting the first signal y from the pulse wave signal d (step ST109).

  Next, it is determined whether or not an excessive amplitude is detected in the pulse wave signal within the sampling interval (step ST110). When an excessive amplitude is detected (step ST110: Y), it is highly likely that an impact noise has been added, so the gain coefficient (for example, h1 = 1.2, h2 = 0) in the impact detection mode is selected. The output value is calculated based on the selected gain coefficient (step ST111).

  In step ST110, when an excessive amplitude is not detected (step ST110: N), it is determined whether or not an excessive amplitude has been detected for a certain period after the excessive amplitude is detected last (step ST112). If not (step ST112: Y), adaptive filter coefficient update processing is executed (step ST113), and then the normal mode gain coefficient (for example, h1 = 1.0, h2 = 0.5) is selected. The output value is calculated based on the selected gain coefficient (step ST114).

  Next, a post-processing procedure will be described. First, a filtered signal and an acceleration signal (body motion signal) are acquired (step ST115). Next, body motion removal adaptive filtering processing is executed by the body motion noise removal processing unit 300 (see FIG. 1) (step ST116). Next, the pulse wave frequency analysis unit 400 executes frequency analysis processing (step ST117). Next, the pulse rate calculation unit 410 calculates the pulse rate (step ST118).

(Actual processing example)
8A to 8C are diagrams illustrating an example of a pulse wave signal waveform in actual processing and a frequency spectrum obtained by FFT. In FIG. 8, the waveform of the pulse wave signal for 16 seconds (upper stage) and its FFT result (lower stage) are shown. Note that the pulse wave signal handled in this example is a pulse wave signal obtained when a periodic motion such as walking or jogging is not performed.

  In the upper waveform of FIG. 8A, a signal that seems to be a disturbance noise component is seen in the vicinity of the terminal portion of the signal for 16 seconds (portion A indicated by a dotted line). In the lower FFT results, the pulse wave signal contains a pulsation signal component MP1a that is a detection target component and a disturbance noise signal component SP1a.

  Next, reference is made to FIG. FIG. 8B shows a pulse wave signal obtained by filtering the pulse wave signal having the waveform shown in the upper part of FIG. 8A with an adaptive filter that is not subjected to filter reconstruction processing, and the FFT result thereof. Show. Here, the adaptive filter is in a state of being continuously operated for about 40 minutes from the start of measurement (a state in which the filter coefficient has been used while being adaptively updated) (however, the filter reconstruction process is performed). It has not been).

  As is clear from FIG. 8B, the disturbance noise component (noise signal component) SP1b remains without being cut. In the example of FIG. 8B, since the spectrum value of the pulsation presentation spectrum MP1b is the largest, the signal MP1b is determined to be the pulsation presentation spectrum by sorting processing based on the peak value of each spectrum. Is possible. However, when the pulsation presentation spectrum MP1b is small, it is difficult to distinguish from the disturbance noise component (noise signal component) SP1b, and the possibility that the detection of the pulsation component will fail increases.

  Next, reference is made to FIG. FIG. 8C is obtained by filtering the pulse wave signal having the waveform shown in the upper part of FIG. 8A with an adaptive filter 202 (when filter reconstruction processing is performed at intervals of 10 minutes, for example). The pulse wave signal and its FFT result are shown. Here, the adaptive filter 202 is in a state of continuously operating for about 40 minutes from the start of measurement (a state in which the filter coefficient has been used while being adaptively updated). It was carried out at 10 minute intervals).

  As is clear from FIG. 8C, the disturbance noise component SP1c is sufficiently suppressed by the adaptive filtering process, while the pulsation presentation spectrum MP1b is emphasized, so that the pulsation component is reliably identified. can do.

The performance of the filter can be objectively evaluated using the S / N index. Hereinafter, the effect of this embodiment is demonstrated by evaluation based on the S / N index. In the following description, an evaluation value called SN3 is used as the S / N index. Here, SN3 can be expressed by the following calculation formula.
SN3 = (total of the pulsation presentation spectrum and one of its left and right spectrum values) / (total of the spectrum values at all frequencies 0 to 4 Hz) (unit:%)
The S / N index (SN3) in the pulse wave signal (original signal) shown in FIG. 8 (A) was 13.8%. The S / N index (SN3) in the pulse wave signal (without reconfiguring the adaptive filter) shown in FIG. 8B is 13.3%, which is slightly lower than the original signal. On the other hand, the S / N index (SN3) in the pulse wave signal (with adaptive filter reconstruction) shown in FIG. 8C is 18.7%, and the S / N is greatly improved. Yes. The effect of reconfiguring the adaptive filter 202 can be numerically confirmed by the fact that the S / N index is improved.

(Second Embodiment)
FIGS. 9A and 9B correspond to the time lapse of the S / N index (SN3) in each of the example in which the adaptive filter reconstruction processing is executed and the case in which the adaptive filter reconstruction processing is not performed (comparative example). It is a figure which shows the result of having investigated the fluctuation | variation.

  Here, a person who is a subject wears a wristwatch-type pulsometer (that is, a pulsation detecting device 100), the subject walks for 20 minutes, the pulse wave signal is filtered by an adaptive filter, and the filtered The S / N index (SN3) was calculated for the signal, and the calculation result was plotted on the time axis. 9A and 9B, the horizontal axis is the time axis, and the vertical axis is the axis indicating the value (%) of SN3 that is the S / N index. FIG. 9A shows the change over time of SN3 from the start of walking by the user as the subject until 10 minutes have passed. FIG. 9 (B) shows the change over time of SN3 from the time when 15 minutes have elapsed until 20 minutes have passed. The data corresponding to the present embodiment is indicated by a black square, and the data of the comparative example is indicated by a white diamond.

  From the start of measurement until about 10 minutes (FIG. 9A), there is almost no difference in the value of SN3 between this embodiment and the comparative example. However, as is clear from FIG. 9B, the difference appears clearly by about 15 minutes.

  9A and 9B, the adaptive filter 202 constituting the pulse wave signal filtering unit (adaptive line spectrum enhancer) is reconfigured at an appropriate timing, so that the performance of the adaptive filter 202 is improved. It was proved that it can be maintained at a certain level. Therefore, even when long-time continuous measurement is performed, the performance of the adaptive filter 202 can be maintained at an appropriate level, and the occurrence of pulsation detection failure and false detection can be reduced.

  As described above, according to at least one embodiment of the present invention, the performance of the adaptive filter is maintained at an appropriate level by reconfiguring the adaptive filter constituting the pulse wave signal filtering unit at an appropriate timing. be able to. Therefore, for example, even when continuous measurement is performed for a long time, the performance of the adaptive filter can be maintained at an appropriate level. Thereby, the failure of pulsation detection and the occurrence of false detection can be reduced.

  In addition, for example, the effective use of the adaptive filter can simplify the filter configuration, and it is easy to reconfigure the adaptive filter. Therefore, effective noise can be reduced while reducing the processing burden on the apparatus. It is possible to realize a pulsation detecting device capable of executing countermeasures.

  Although the present embodiment has been described in detail as described above, it will be easily understood by those skilled in the art that many modifications can be made without departing from the novel matters and effects of the present invention. Accordingly, all such modifications are intended to be included in the scope of the present invention. For example, a term described with a different term having a broader meaning or the same meaning at least once in the specification or the drawings can be replaced with the different term in any part of the specification or the drawings.

10 pulse wave sensor, 11 body motion sensor (acceleration sensor, gyro sensor, etc.),
12 pulse wave signal storage unit, 17 filter reconstruction unit,
200 Pulse wave signal filtering unit,
300 body motion noise removal processing unit, 400 pulse wave frequency analysis unit, 600 display unit

Claims (8)

A pulsation detection device for detecting a pulsation signal derived from the pulsation of a subject,
A pulse wave sensor that outputs a pulse wave signal in which the pulse signal and a noise signal including a body motion noise signal derived from the body motion of the subject are mixed;
A filter that filters the pulse wave signal, and an adaptive filter that self-adapts a frequency response characteristic; and a state of the adaptive filter at a first time point during a period in which the adaptive filter is continuously operating, A pulse wave signal filtering unit including: an adaptive filter reconfiguring unit configured to reconfigure a second time point including a time point before the adaptive filter starts operating,
A frequency analysis unit that performs a frequency analysis process at predetermined intervals based on a signal output from the pulse wave signal filtering unit and detects a pulsation presentation spectrum representing the pulsation signal, and To detect pulsation. The pulsation detecting device according to claim 1,
The time from when the adaptive filter starts to operate or from the most recent time when the adaptive filter is reconfigured to the third time when the first time has elapsed is defined as a first period in which the reconfiguration of the adaptive filter is unnecessary, When the second period in which the adaptive filter can be reconfigured from the third time point to the fourth time point when the second time has elapsed,
The pulsation detecting device, wherein the second time point is set in the first period, and the first time point is set in the second period. The pulsation detecting device according to claim 2,
The pulsation detecting device according to claim 1, wherein the first time point is set at a predetermined period with reference to a time point when the adaptive filter starts to operate. The pulsation detecting device according to claim 2 or 3,
The adaptive filter reconstruction unit reconfigures the adaptive filter on the condition that the pulsation presenting spectrum is detected by the frequency analysis processing at the most recent time point before the first time point. A characteristic pulsation detection device. The pulsation detecting device according to claim 2 or 3,
Based on the frequency analysis result of the pulse wave signal, a signal evaluation unit that determines the degree of noise amount of the noise signal,
In the adaptive filter reconstruction unit, the pulsation presentation spectrum is detected by the frequency analysis processing at the most recent time before the first time point, and the noise amount is determined by the signal evaluation unit. A pulsation detecting device, characterized in that the adaptive filter is reconfigured on the condition that it is determined to be below a reference. The pulsation detecting device according to claim 2 or 3,
Based on the frequency analysis result of the pulse wave signal, the signal evaluation unit for determining the motion state of the subject,
In the adaptive filter reconstruction unit, the pulsation presentation spectrum is detected by the frequency analysis processing at the most recent time before the first time point, and the motion state of the subject is detected by the signal evaluation unit. The pulsation detecting device is characterized in that the adaptive filter is reconfigured on the condition that is determined to be in a stable state. The pulsation detecting device according to claim 2 or 3,
Based on the frequency analysis result of the pulse wave signal, the signal evaluation unit for determining the degree of noise amount of the noise signal and the motion state of the subject,
In the adaptive filter reconstruction unit, the pulsation presentation spectrum is detected by the frequency analysis processing at the most recent time before the first time point, and the noise amount is equal to or less than a predetermined reference by the signal evaluation unit. And the adaptive filter is reconfigured on the condition that the motion state of the subject is determined to be a stable state. The pulsation detecting device according to any one of claims 1 to 7,
The pulse wave signal filtering unit
A delay processor for delaying the pulse wave signal for a predetermined time;
A filter coefficient update unit for updating the coefficient of the adaptive filter;
A subtracting unit,
The adaptive filter outputs a first signal having high autocorrelation among output signals of the delay processing unit,
The subtracting unit subtracts the first signal from the pulse wave signal to generate a second signal having lower autocorrelation than the first signal, and supplies the second signal to the filter coefficient updating unit. ,
The filter coefficient update unit updates the coefficient of the adaptive filter so that the second signal is suppressed,
The adaptive filter reconstruction unit sets a coefficient of the adaptive filter to a coefficient at the time when the adaptive filter starts operation when reconfiguring the adaptive filter at the first time point. Motion detection device.

Images