JP2013208312A - Signal analyzer, electronic equipment and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a signal analyzer capable of performing frequency resolution processing with improved real-time properties while maintaining resolution equivalent to resolution for a conventional method, and to provide electronic equipment, a program and the like.SOLUTION: A signal analyzer 100 includes a signal acquisition unit 110 which acquires a detection signal from a detection unit 200 having a sensor, and a signal processing unit 130 which performs window function processing for sampling data of the detection signal and Fourier transform processing for data after the window function processing as frequency resolution processing. The jth sampling data Dof sampling data D-Dis data sampled in timing after timing when the ith sampling data Dis sampled (0≤i<N/2<j≤N). The signal processing unit 130 performs the window function processing by an asymmetric window function in which a window function value h (j) corresponding to Dis a maximum value and in which the window function value h (i) corresponding to Dis a value smaller than the window function value h (j).

Description

    The present invention relates to a signal analysis device, an electronic device, a program, and the like.

  2. Description of the Related Art Conventionally, electronic devices including a pulse detection device (signal analysis device) such as a pulse meter have been widely used. The pulsation detection device is a device for detecting pulsation derived from the heartbeat of the human body, for example, based on a signal from a pulse wave sensor attached to an arm, palm, finger, etc. It is an apparatus for detecting a signal to be detected. Actually, when presenting information (pulsation information) related to a signal derived from a heartbeat to a user, for example, a pulse rate is calculated as the pulsation information and presented.

  In calculating the pulse information such as the pulse rate, firstly, calculating the pulse information corresponding to the latest pulse wave detection signal as much as possible, second, calculating more accurate pulse information, These two points are the main issues.

  To solve such a problem, in the technique disclosed in Patent Document 1, FFT (Fast Fourier Transform) processing is performed on a pulse wave detection signal for a predetermined period, and the maximum frequency component is determined as a pulse spectrum. The frequency is converted into a pulse rate.

JP 2007-054471 A

  However, there are the following problems when measuring the pulse rate by such an approach. First, in order to obtain the pulse rate more accurately, the resolution of the frequency resolution process may be increased. However, in order to increase the resolution of the frequency resolution process, a single pulse rate calculation process (beat information calculation process) is performed. It is necessary to lengthen the detection period of the pulse wave detection signal to be used. On the other hand, as the detection period of the pulse wave detection signal is lengthened, the frequency component of the past pulse wave detection signal is included in the frequency decomposition processing result. That is, the real-time property of the pulse rate decreases as the resolution of the frequency resolution process is increased.

  That is, with the conventional method, it is difficult to achieve both real-time characteristics of the pulse rate and frequency resolution. The cause of this problem is not the lack of hardware processing capability, but the above-described frequency resolution processing algorithm (method).

  According to some aspects of the present invention, it is possible to provide a signal analysis device, an electronic device, a program, and the like that can perform frequency decomposition processing with improved real-time characteristics while maintaining the same resolution as that of the conventional method. .

One aspect of the present invention is a signal acquisition unit that acquires a detection signal from a detection unit having a sensor, and window function processing for sampling data D _{0 to} D _N (N is a positive integer of 2 or more) of the detection signal; A signal processing unit that performs Fourier transform processing on the data after the window function processing as frequency decomposition processing, and the j-th sampling data D _j among the sampling data D _{0 to DN} is the i-th sampling The data D _i is data sampled at a timing later than the timing at which the data D _i was sampled (integers i and j are integers satisfying 0 ≦ i <N / 2 <j ≦ N), and the signal processing unit window function value corresponding to the sampling data D _j of the j h (j) is the maximum value, the second window function value corresponding to the sampling data D _i of i h (i) is the window function Asymmetric window function becomes a value smaller than the value h (j), relating to the signal analysis apparatus performs the window function processing.

In one aspect of the present invention, by making the window function value h (i) corresponding to the _i th sampling data D _i smaller than the window function value h (j) corresponding to the _{j th} sampling data D _j , The signal value of the sample of 0 ≦ i <N / 2 is compressed larger than the sample of N / 2 <j ≦ N.

  Thereby, in a window function process, the past sample can be compressed compared with the recent sample. Therefore, the real-time property of the frequency decomposition processing result (pulse rate) can be improved as compared with the case where the window function processing is not performed or the case where the window function processing is performed using the symmetric window function. This effect can also be rephrased as follows. When the window function processing is not performed, or when the specification is sufficient with real-time characteristics equivalent to the case where the window function processing is performed using a symmetric window function, the sampling period can be made longer, so the frequency resolution A high frequency distribution can be obtained. Hereinafter, description of the latter effect is omitted.

In one embodiment of the present invention, the signal processing unit, the asymmetry the sampling data D ₀ to D window function corresponding to the N sampled data D _N of the _N values h (N) is the minimum value The window function processing may be performed using a window function.

  As a result, it is possible to improve the side lobe characteristics compared to the case where the window function processing is not performed. In addition, the real-time property of the frequency resolution processing result (pulse rate) can be improved as compared with the case where the window function processing is performed using a symmetric window function.

In one embodiment of the present invention, the signal processing unit, the asymmetry the sampling data D ₀ to D window function corresponding to the N sampled data D _N of the _N values h (N) is the maximum value The window function processing may be performed using a window function.

  According to this, compared with the case where window function processing is not performed or the case where window function processing is performed using a symmetric window function, it is possible to improve the real-time property of the frequency decomposition processing result (pulse rate). Become.

In one embodiment of the present invention, the signal processing unit, the sampling data _{D N} (integer k sampling data _{D k} ~ N-th of the first k of the sampling data _D 0 to D _N, the N / 2 < The window function processing may be performed by the asymmetric window function in which all of the window function values h (k) to h (N) corresponding to an integer satisfying k <N are maximized.

  As a result, not only the Nth sampling data but also sampling data that is traced a predetermined number of times from the Nth sampling data can be maximized to reduce the compression rate. Therefore, it is possible to improve the real-time property of the frequency decomposition processing result (pulse rate) compared to the case where the window function processing is not performed or the case where the window function processing is performed using a symmetric window function.

In the aspect of the invention, the signal processing unit may be configured to obtain p-th sampling data D _p (the integers p, M, and N are 0 ≦ p ≦ 3M−1) among the sampling data D _{0 to DN.} , N = an integer satisfying 4M−1), the window function processing is performed by the following equation (1), and the q-th sampling data D _q (integers q, M and N are 3M ≦ q ≦ 4M). −1 and an integer satisfying N = 4M−1), the window function processing may be performed by the following equation (2).

  As a result, it is possible to improve the side lobe characteristics compared to the case where the window function processing is not performed. In addition, the real-time property of the frequency resolution processing result (pulse rate) can be improved as compared with the case where the window function processing is performed using a symmetric window function. Therefore, for example, even when the level of the pulse wave detection signal is low, it is possible to easily specify the pulse frequency.

In the aspect of the invention, the signal processing unit may be configured to obtain p-th sampling data D _p (the integers p, M, and N are 0 ≦ p ≦ 3M−1) among the sampling data D _{0 to DN.} , N = integer satisfying 4M−1), the window function processing is performed by the following equation (3), and the q-th sampling data D _q (integers q, M and N are 3M ≦ q ≦ 4M). −1 and an integer satisfying N = 4M−1), the window function processing may be performed by the following equation (4).

  That is, in the 0th and Nth sampling data, it is possible to set not only the signal value but also the slope to 0. As a result, it is possible to improve the side lobe characteristics compared to the case where the window function processing is not performed. In addition, the real-time property of the frequency resolution processing result (pulse rate) can be improved as compared with the case where the window function processing is performed using a symmetric window function. Therefore, for example, even when the level of the pulse wave detection signal is low, it is possible to easily specify the pulse frequency.

In the aspect of the invention, the signal processing unit may be configured to obtain p-th sampling data D _p (the integers p, M, and N are 0 ≦ p ≦ 3M−1) among the sampling data D _{0 to DN.} , N = integer satisfying 4M−1), the window function processing is performed by the following equation (5), and the q-th sampling data D _q (integers q, M and N are 3M ≦ q ≦ 4M). −1 and an integer satisfying N = 4M−1), the window function processing may be performed by the equation h (q) = 1.

  In other words, it is possible to realize not only the Nth sampling data but also maximize the window function value for the sampling data that is a predetermined number back from the Nth sampling data. This makes it possible to improve the real-time property of the frequency decomposition processing result (pulse rate) compared to the case where the window function processing is not performed or the case where the window function processing is performed using a symmetric window function.

  Moreover, in one mode of the present invention, the signal processing unit includes a signal state determination unit that performs a signal state determination process on the detection signal acquired by the signal acquisition unit, and the signal processing unit has the first signal state as the signal state. If it is determined that the signal state is the second signal state, the first asymmetric window function is performed. The window function processing may be performed using a second asymmetric window function different from the above.

  This makes it possible to change the asymmetric window function used for window function processing according to the signal state.

  Moreover, in one mode of the present invention, the signal processing unit includes a signal state determination unit that performs a signal state determination process on the detection signal acquired by the signal acquisition unit, and the signal processing unit has the first signal state as the signal state. If the signal function is determined to be, the window function processing by the asymmetric window function and the Fourier transform processing on the data after the window function processing are performed as the frequency decomposition processing, and the signal state is the second signal. When it is determined that the state is the state, a power spectrum density estimation process for the detection signal may be performed as the frequency resolution process.

  That is, for example, in the case of the first signal state in which the SN ratio is lower than the second signal state, the frequency decomposition process having a high sidelobe suppression effect is performed, and the second SN ratio is higher than that in the first signal state. In the case of the signal state, it is possible to perform frequency decomposition processing that emphasizes real-time characteristics.

  This makes it possible to improve the real-time property of the frequency decomposition processing result (pulse rate) as compared with the case where the window function processing is not performed or the case where the window function processing is performed using the symmetric window function.

  In one aspect of the present invention, the signal acquisition unit includes a signal state determination unit that performs a signal state determination process on the detection signal acquired by the signal acquisition unit, wherein the detection unit is the first sensor that is the sensor. The signal acquisition unit further includes a first detection signal based on the first sensor signal acquired by the first sensor and a second sensor acquired by the second sensor. A second detection signal based on a sensor signal of the second detection signal is acquired from the detection unit, and the signal state determination unit performs a process of determining the signal state of the second detection signal as the determination process, and the signal When it is determined that the signal state of the second detection signal is the first signal state, the processing unit performs the window function processing using the first asymmetric window function as the asymmetric window function. For the first detection signal When the signal state of the second detection signal is determined to be the second signal state, the window function is calculated using a second asymmetric window function different from the first asymmetric window function. Processing may be performed on the first detection signal.

  Accordingly, it is possible to change the asymmetric window function used for the window function processing according to the signal state of the second detection signal different from the first detection signal, and perform the frequency decomposition processing.

  In one aspect of the present invention, the signal acquisition unit includes a signal state determination unit that performs a signal state determination process on the detection signal acquired by the signal acquisition unit, wherein the detection unit is the first sensor that is the sensor. The signal acquisition unit further includes a first detection signal based on the first sensor signal acquired by the first sensor and a second sensor acquired by the second sensor. A second detection signal based on a sensor signal of the second detection signal is acquired from the detection unit, and the signal state determination unit performs a process of determining the signal state of the second detection signal as the determination process, and the signal When it is determined that the signal state of the second detection signal is the first signal state, the processing unit uses the first detection signal that uses the first asymmetric window function as the asymmetric window function. The window function processing for The Fourier transform processing on the data after the window function processing is performed as the frequency decomposition processing, and when it is determined that the signal state of the second detection signal is the second signal state, the first Power spectrum density estimation processing for the detected signals may be performed as the frequency resolution processing.

  That is, the signal state is determined according to the signal state of the second detection signal. For example, in the case of the first signal state in which the SN ratio is lower than that of the second signal state, the frequency decomposition with a high sidelobe suppression effect is achieved. When the processing is performed and the second signal state is higher in the S / N ratio than the first signal state, it is possible to perform frequency decomposition processing that emphasizes real-time characteristics.

  This makes it possible to improve the real-time property of the frequency decomposition processing result (pulse rate) compared to the case where the window function processing is not performed or the case where the window function processing is performed using a symmetric window function.

  In the aspect of the invention, the signal state determination unit performs a process of determining an exercise state of the subject based on the second detection signal as the determination process, and the signal processing unit The frequency resolution processing may be determined based on a state determination result.

  This makes it possible to perform different frequency resolution processing depending on the motion state.

  In the aspect of the invention, the detection unit further includes a second sensor in addition to the first sensor that is the sensor, and the signal acquisition unit is a first sensor acquired by the first sensor. A first detection signal based on one sensor signal and a second detection signal based on a second sensor signal acquired by the second sensor are acquired from the detection unit as the detection signal, and the signal The processing unit may perform signal processing on the first detection signal based on the second detection signal.

  As a result, for example, it is possible to perform noise reduction processing of the first detection signal using the second detection signal.

  In the aspect of the invention, the signal processing unit may perform noise reduction processing on the first detection signal as the signal processing based on the second detection signal.

  As a result, it is possible to reduce the noise included in the first detection signal, improve the sidelobe characteristics after Fourier transform, and the like.

  Another aspect of the invention relates to an electronic device including the signal analysis device.

  Another aspect of the present invention relates to a program that causes a computer to function as each of the above-described units.

The system block diagram of this embodiment. 2A and 2B are examples of electronic devices including a pulsation detection device. The comparison figure of the shape of each window function. The comparison figure of the characteristic of each frequency decomposition processing. The flowchart explaining the flow of the whole process of 1st Embodiment. The flowchart explaining the flow of the determination process of the signal state in 1st Embodiment. FIG. 7A to FIG. 7C are explanatory diagrams when it is determined that the signal state is good. 8A to 8C are explanatory diagrams when the signal state is determined to be medium. FIG. 9A to FIG. 9C are explanatory diagrams when it is determined that the signal state is bad. FIG. 10A to FIG. 10C are comparison diagrams of the window function processing results. The flowchart explaining the flow of the whole process of 2nd Embodiment. The flowchart explaining the flow of the determination process of the signal state in 2nd Embodiment. The comparison figure of the measurement result of a pulse rate. FIG. 14A to FIG. 14C are comparative diagrams of spectra of acceleration detection signals in each motion state. FIGS. 15A and 15B are specific examples showing a spectrum of an acceleration detection signal at the start of exercise. FIG. 16A and FIG. 16B are specific examples showing the spectrum of the acceleration detection signal when the movement is stopped.

  Hereinafter, this embodiment will be described. First, an outline of the method of the present embodiment will be described, and then an example of a system configuration will be described. Then, the first embodiment and the second embodiment will be described using a flowchart and the like. Finally, another application example of the first embodiment and the second 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.

1. 2. Description of the Related Art Conventionally, electronic devices including a pulse detector such as a pulse meter have been widely used. The pulsation detection device is a device for detecting pulsation derived from the heartbeat of the human body, for example, based on a signal from a pulse wave sensor attached to an arm, palm, finger, etc. It is an apparatus for detecting a signal to be detected. Actually, when presenting information (pulsation information) related to a signal derived from a heartbeat to a user, for example, a pulse rate is calculated as the pulsation information and presented.

  In calculating the pulse information such as the pulse rate, firstly, calculating the pulse information corresponding to the latest pulse wave detection signal as much as possible, second, calculating more accurate pulse information, These two points are the main issues.

  In order to deal with such a problem, in the method described in Patent Document 1 described above, FFT (Fast Fourier Transform) processing is performed on a pulse wave detection signal for a predetermined period, and the maximum frequency component is obtained as a pulse spectrum. The frequency is converted into a pulse rate.

  However, there are the following problems when measuring the pulse rate by such an approach. First, in order to obtain the pulse rate more accurately, the resolution of the frequency resolution process may be increased. However, in order to increase the resolution of the frequency resolution process, a single pulse rate calculation process (beat information calculation process) is performed. It is necessary to lengthen the detection period of the pulse wave detection signal to be used.

  On the other hand, when the detection period of the pulse wave detection signal is lengthened, there are cases where the required real-time property is not satisfied.

  Here, the real-time property means that the worst value of the response time is guaranteed, that is, the response is returned within the required response time. Real-timeness is not a concept indicating that the response is fast or slow. The required response time varies depending on the system. Therefore, for example, if the response time (response period) required for a certain system is 32 days, even if a certain system returns a response over a full month, it can be said that the system satisfies the real-time property.

  In addition, the strictness required for the real-time property varies depending on the system. Some real-time properties include hard real-time properties that do not allow a slight delay, and soft real-time properties that do not allow a slight delay. For example, hardware real-time performance that does not allow delays in the order of microseconds is often required for processing performed by firmware that operates hardware, while weather forecasts have a delay of several minutes. However, an acceptable soft real-time property is required. These classifications represent the tolerance of response delay and are independent of the speed of response time.

  The real-time property required for the pulsation detection device is the soft real-time property described above, but the required response time is in units of a few seconds, and the pulsation detection device corresponds to the timing as close as possible to the present. It is desirable to be able to detect the pulse rate and the like.

  Further, as described above, when determining the pulse rate and the like, frequency decomposition processing is performed on the pulse wave detection signal for a predetermined period. In this frequency decomposition process, a window function process is performed on the pulse wave detection signal using a Hanning window or the like as shown in FIG. 3 described later, and then an FFT process is performed. A window function such as a Hanning window that is widely used generally has a maximum window function value in the middle (midpoint) of the period of the input signal. For this reason, the spectrum distribution drawn as a result of the frequency decomposition process is a distribution that shows the expected value in the past for half the time of the period of the input signal. Therefore, the pulse rate calculated based on the spectrum distribution is also calculated as a past expected value for ½ hour of the period of the input signal. For example, when the detection period of the pulse wave detection signal is 4 seconds, the pulse rate 2 seconds before is calculated as the expected value, and when the detection period of the pulse wave detection signal is 16 seconds, the expected value is The pulse rate 8 seconds before is calculated.

  As described above, when the detection period of the pulse wave detection signal input to the frequency resolution processing is lengthened, the resolution is improved, but the expected spectral distribution at a past time point is obtained, and the real-time property described above is obtained. Will decline. The reason why the real-time property is reduced in this way is not due to the lack of hardware processing capability or the like, but is due to the frequency resolution processing algorithm (method) described above.

  Therefore, the applicant proposes a new window function which will be described in detail below. The signal analysis device using the window function proposed by the present applicant can perform frequency resolution processing with improved real-time characteristics while maintaining the same resolution as the conventional method.

2. System Configuration Example Next, FIG. 1 shows a configuration example of a signal analyzing apparatus (beat detecting apparatus) 100 and electronic equipment including the same according to a first embodiment and a second embodiment described later.

  The signal analysis device (beat detection device) 100 includes a signal acquisition unit 110, a signal processing unit 130, and a signal state determination unit 150. Examples of the electronic device including the signal analysis device (beat detection device) 100 include a detection unit 200, a portable terminal including the display unit 70, and the like. Note that the signal analysis device (beat detecting device) 100 and the electronic device including the signal analysis device 100 are not limited to the configuration in FIG. 1, and some of these components are omitted or other components are added. Various modifications such as these are possible.

  Further, here, in order to more specifically describe the operations of the first embodiment and the second embodiment, the case where the signal analysis device 100 is used as the pulsation detection device 100 is illustrated, but the present invention is not limited thereto. The scope of the present invention is not limited to the pulsation detecting device 100, and the signal analyzing device 100 can be used for other purposes as described later.

  Next, processing performed in each unit will be described. Here, the detection unit 200 will be described first, and a signal input to the signal analysis device (beat detection device) 100 will be described, and then each unit of the signal analysis device 100 will be described.

  First, the detection unit 200 includes a pulse wave detection unit 210 and a body motion detection unit 220. Note that the detection unit 200 is not limited to the configuration shown in FIG. 1, and various modifications such as omitting some of these components or adding other components are possible.

  The pulse wave detection unit 210 outputs a pulse wave detection signal based on sensor information (pulse wave sensor signal) obtained from the pulse wave sensor 211. The pulse wave detection unit 210 can include, for example, a pulse wave sensor 211, a filter processing unit 215, and an A / D conversion unit 216. However, the pulse wave detection unit 210 is not limited to the configuration of FIG. 1, and various components such as omitting some of these components or adding other components (for example, an amplification unit that amplifies a signal) are included. Variations are possible.

  The pulse wave sensor 211 is a sensor for detecting a pulse wave sensor signal. For example, a photoelectric sensor can be considered. In addition, when using a photoelectric sensor as the pulse wave sensor 211, you may use the sensor comprised so that the signal component of external lights, such as sunlight, may be cut. This can be realized by, for example, a configuration in which a plurality of photodiodes are provided and difference information is obtained by feedback processing using these signals.

  The pulse wave sensor 211 is not limited to a photoelectric sensor, and may be a sensor using ultrasonic waves. In this case, the pulse wave sensor 211 has two piezoelectric elements, transmits one ultrasonic wave by exciting one of the piezoelectric elements, and converts the ultrasonic wave reflected by the blood flow of the living body into the other. Received by the piezoelectric element. Since the frequency change occurs in the transmitted ultrasound and the received ultrasound due to the Doppler effect of blood flow, a signal corresponding to the blood flow can be obtained in this case as well, and pulsation information can be estimated. . Further, another sensor may be used as the pulse wave sensor 211.

  The filter processing unit 215 performs high-pass filter processing on the pulse wave sensor signal from the pulse wave sensor 211. Note that the cutoff frequency of the high-pass filter may be obtained from a typical pulse rate. For example, there are very few cases where the pulse rate of a normal person falls below 30 times per minute. That is, since the frequency of the signal derived from the heartbeat is rarely 0.5 Hz or less, even if the information of the frequency band in this range is cut, the adverse effect on the signal to be acquired should be small. Therefore, about 0.5 Hz may be set as the cutoff frequency. Further, depending on the situation, a different cutoff frequency such as 1 Hz may be set. Furthermore, since it is possible to assume a typical upper limit value for the human pulse rate, the filter processing unit 215 may perform bandpass filter processing instead of high-pass filter processing. The cutoff frequency on the high frequency side can be set freely to some extent, but a value such as 4 Hz may be used.

  The A / D conversion unit 216 performs A / D conversion processing and outputs a digital signal. The processing in the filter processing unit 215 described above may be analog filter processing performed before the A / D conversion processing, or may be digital filter processing performed after the A / D conversion processing. .

  The body motion detection unit 220 outputs a signal (body motion detection signal) corresponding to the body motion based on sensor information (body motion sensor signal) of various sensors. The body motion detection unit 220 can include, for example, a body motion sensor 221, a filter processing unit 225, and an A / D conversion unit 226. However, the body motion detection unit 220 may include other sensors (for example, a gyro sensor), an amplification unit that amplifies a signal, and the like.

  As the body motion sensor, a motion sensor (acceleration sensor), a pressure sensor (contact pressure sensor), or the like is used. Further, as the body motion sensor, a plurality of types of sensors may be provided, or a configuration including one type of sensor may be used.

  The motion sensor is, for example, an acceleration sensor. The acceleration sensor is composed of an element whose resistance value is increased or decreased by an external force, for example, and detects triaxial acceleration information.

  The pressure sensor is, for example, a contact pressure sensor. The contact pressure sensor may be one that directly contacts the subject and measures the contact pressure, or may indirectly measure the contact pressure using a cuff structure or the like. That is, a piezoelectric element may be used, or an atmospheric pressure sensor or the like may be used.

  The filter processing unit 225 performs various filter processes on the body motion sensor signal. For example, the filter processing unit 225 performs high-pass filter processing, band-pass filter processing, or the like. However, the filtering process is not necessarily performed. The filter processing unit 225 may not be included.

  Since the processing performed by the A / D conversion unit 226 is substantially the same as that of the A / D conversion unit 216, the description is omitted.

  Here, we will explain the terminology once. First, pulsation refers to a movement in which peripheral blood vessels expand or contract. Further, the pulse wave is a signal that captures a change in volume caused by the inflow of blood into the body tissue as a signal. For example, the pulse wave is obtained by irradiating a body surface with an LED from a pulse wave sensor, capturing the scattered light / reflected light with a photodiode and capturing it as a signal waveform. The pulse wave captures vasomotor reaction, not the heart motion itself, and includes noise factors other than the heart motion, for example, changes in the volume of blood vessels caused by human motion and movement. In order to correctly capture the heart motion itself and the pulse rate, it is necessary to remove this noise factor.

  On the other hand, pulsation refers to a movement that occurs when the periodical contraction and relaxation of the internal organs are repeated not only in the heart, but also in the heart. The movement as a pump that sends blood is called pulsation.

  Furthermore, body movement means all the movements of the body in a broad sense, and refers to steady, periodic movements of the arm (near the site where the pulse meter is worn) associated with walking, jogging, etc. in the narrow sense. .

  Further, the pulse wave sensor signal (pulse wave sensor original signal, pulse wave signal) refers to a signal itself detected by the pulse wave sensor 211. The pulse wave sensor signal includes a pulsation component signal, a body motion noise component signal, a disturbance noise component signal, and the like. Although it is considered that the arrhythmia is strictly included in the pulsation component signal, it is assumed that the arrhythmia is included in the disturbance noise component signal as an electrical signal.

  On the other hand, the pulse wave detection signal is a signal output from the pulse wave detection unit 210. Specifically, the pulse wave detection signal is filtered by the filter processing unit 215 with respect to the pulse wave sensor signal, and the A / D conversion unit 216 performs A / D conversion on the pulse wave sensor signal after the filter processing. This refers to a signal that has undergone D conversion processing. However, as described above, the order of the filtering process and the A / D conversion process may be reversed.

  Furthermore, the body motion sensor signal (body motion sensor original signal, body motion signal) refers to a signal itself detected by the body motion sensor. Specifically, the body motion sensor signal indicates a motion sensor signal indicating the signal itself detected by the motion sensor 221, a contact pressure sensor signal indicating the signal itself detected by the contact pressure sensor, or the like.

  On the other hand, the body motion detection signal is a signal output from the body motion detection unit 220. Specifically, the body motion detection signal refers to a signal after filtering processing or A / D conversion processing is performed on the body motion sensor signal. For example, the body motion detection signal is a motion detection signal that is a signal after the above signal processing is performed on the motion sensor signal, or a pressure detection signal that is a signal after the above signal processing is performed on the pressure sensor signal. Etc. However, as described above, the order of the filtering process and the A / D conversion process may be reversed.

  The pulse wave sensor signal and the body motion sensor signal are collectively referred to as a sensor signal, the pulse wave detection signal and the body motion detection signal are collectively referred to as a detection signal.

  Next, processing performed in each unit of the signal analysis device (beat detecting device) 100 will be described.

  First, the signal acquisition unit 110 acquires a detection signal from the detection unit 200. Further, the signal acquisition unit 110 of the pulsation detection device 100 acquires a pulse wave detection signal from the pulse wave detection unit 210 and a body motion detection signal from the body motion detection unit 220. The signal acquisition unit 110 may be a communication unit (I / F unit, antenna unit, or the like) that transmits and receives data when connected to the detection unit 200 wirelessly or by wire.

  Next, the signal processing unit 130 performs frequency decomposition processing on the sampling data of the detection signal to obtain a spectrum. For example, as the frequency decomposition process, a window function process is performed on the sampling data of the detection signal and a Fourier transform process is performed on the data after the window function process. The signal processing unit 130 of the pulsation detection device 100 performs frequency resolution processing based on the pulse wave detection signal from the pulse wave detection unit 210 and the body motion detection signal from the body motion detection unit 220, and generates a pulse spectrum. Ask.

  In addition, the signal processing unit 130 can include a noise reduction processing unit 131, a window function processing unit 133, a Fourier transform processing unit 135, and a power spectrum density estimation processing unit 137. Details of the operations of these processing units will be described later.

  Then, the signal state determination unit 150 performs signal state determination processing on the detection signal acquired by the signal acquisition unit 110. The signal state determination unit 150 of the pulsation detection device 100 performs signal state determination processing on at least one of the pulse wave detection signal and the body movement detection signal.

  Further, the signal state determination unit 150 can include a spectrum value calculation unit 151 and an exercise state determination unit 153. Details of the operations of these processing units will be described later.

  Furthermore, when the signal analysis device 100 is the pulsation detection device 100, a pulsation information calculation unit 170 can be included as shown in FIG. The pulsation information calculation unit 170 calculates pulsation information based on the result of the frequency decomposition processing obtained by the signal processing unit 130. Here, the pulsation information refers to information representing the pulsation by a numerical value or the like, for example, the pulse rate.

  For example, the pulsation information calculation unit 170 performs a process of converting to a pulse rate while adding a smoothing process or the like to the analysis result (pulse frequency value) of the signal processing unit 130. When calculating the pulse rate, the pulsation information calculation unit 170 may perform a process of setting a representative frequency in the pulse spectrum calculated by the signal processing unit 130 as a pulse frequency. In this case, a value obtained by multiplying the obtained pulse frequency by 60 is a commonly used pulse rate (heart rate).

  The pulsation information is not limited to the pulse rate, and may be, for example, other information indicating the pulse rate (heartbeat frequency, period, etc.). Moreover, the information which represents the state of pulsation may be sufficient, for example, the value showing blood flow itself (or its fluctuation | variation) is good also as pulsation information. However, since the relationship between the blood flow rate and the signal value of the pulse wave sensor signal has individual differences for each user, when the blood flow rate or the like is used as pulsation information, correction processing is performed to deal with the individual differences. It is desirable.

  When the signal analysis device 100 is not the pulsation detection device 100, a processing unit for calculating other information may be included instead of the pulsation information calculation unit 170.

  The display unit 70 (output unit in a broad sense) is for displaying various display screens used for presenting the calculated pulsation information and the like, and can be realized by, for example, a liquid crystal display or an organic EL display.

  Specific examples in the case where the electronic device described above is a pulse meter are shown in FIGS. 2 (A) and 2 (B). FIG. 2A shows an example of a wristwatch-type pulse meter. The base unit 400 including the pulse wave sensor 211 and the display unit 70 is attached to the left wrist of the subject (user) by a holding mechanism 300 (for example, a band). FIG. 2B shows an example of a finger wearing type. A pulse wave sensor 211 is provided at the bottom of a ring-shaped guide 302 for insertion into the fingertip of the subject. However, in the case of FIG. 2B, since there is no space to provide the display unit 70, the display unit 70 (and a device corresponding to the pulsation detecting device 100 if necessary) should be provided elsewhere. Is assumed. However, the electronic device including the signal analysis device (beat detecting device) 100 is not limited to the pulse meter, and may be a pedometer, a smartphone, or the like.

3. Asymmetric Window Function Next, a window function used in the signal analysis device (beat detecting device) 100 of the first embodiment and the second embodiment described later will be described with reference to FIG. FIG. 3 is a graph comparing the shapes of the window functions when the number of samples is 256. The vertical axis represents the window function value, and the horizontal axis represents the sample number. The window function value is a weight by which each sample is multiplied.

  First, the Hanning window function described above has a symmetrical shape on the graph shown in FIG. In addition to the Hanning window function, commonly used window functions such as the Blackman window function are the same. As described above, generally, a window function having a symmetrical shape on the graph of FIG. 3 (hereinafter referred to as a symmetric window function) is used to improve sidelobe characteristics or to adjacent power spectrum. This is to make it easier to see the difference in power between the two. Note that the side lobe characteristic refers to the degree to which the amplitude of a frequency other than the main component is detected.

  In the first place, the Fourier transform is to represent a signal in an arbitrary time domain in the frequency domain on the premise that any periodic function can be represented by a superposition of a sine wave and a cosine wave. That is, as a premise, it is important that the Fourier transform is performed on a periodic function.

  Here, only a single frequency is detected in the spectrum obtained when a completely periodic time signal is Fourier transformed. That is, the sidelobe characteristics described above are good and the main component can be easily detected. On the other hand, in a spectrum obtained when Fourier transform is performed on a non-periodic time signal, a plurality of frequencies are detected in a wide range from a low frequency to a high frequency. That is, the sidelobe characteristics are poor and it is difficult to determine which frequency the main component is. Since the purpose of performing the Fourier transform is often to detect a representative frequency (principal component) of a time signal, it is better that the sidelobe characteristics are better.

  However, when actually performing a Fourier transform, it is not possible to take an infinite number of samples from a certain time signal. Therefore, taking a finite number of samples from a limited range (period) of a certain time signal as sampling data, This is used as an input signal. In particular, when performing FFT, there is a restriction that the number of samples of the input signal must be a power of 2 such as 64, 128, 256, or the like. However, it is extremely unlikely that sampling data consisting of a finite number of samples can be represented by a periodic function. Moreover, the frequency often changes with time.

  Furthermore, when performing discrete Fourier transform, if the continuity between the start point and the end point of sampling data consisting of a finite number of samples is not ensured (secured), the side lobe that appears in the frequency decomposition result becomes large.

  Therefore, a window function process is performed on the sampling data to ensure (secure) the continuity of the sampling data consisting of a finite number of samples. There are several methods for allowing the sampling data to be represented by a periodic function when it is repeated. As the most easily understood method, there is a method in which the values at both ends of the sampling data are set to 0 and the slopes at both ends are set to 0. . In this case, the sampling data can be represented by a periodic function having a sampling period of one period. Since a window function such as a Hanning window employs such an approach, it has a shape as shown in the graph of FIG.

  In addition, the frequency resolution processing for performing window function processing using such a window function before Fourier transform processing is performed in the field of communication, for example. In that case, for example, a signal is received by wireless communication or the like, frequency decomposition processing or the like is performed, and the original information converted into the signal is analyzed. Therefore, if the waveform of the signal transmitted by the transmission source is greatly changed, the original information converted into the signal may not be analyzed. Therefore, in the window function processing, it is necessary to maintain the waveform of the original signal as much as possible while processing it into a periodic signal. Therefore, a window function such as a Hanning window is designed to be symmetrical in the graph of FIG.

  However, in the field of pulsation detection (measurement), when the frequency resolution processing is performed using the above-described symmetric window function, there is a problem that the real-time property decreases when the frequency resolution is increased as described above. is there. This problem is caused by the window function value being maximized in the middle of the sampling period in the target window function such as the Hanning window. In other words, the longer the sampling period is, the more the timing at which the window function value becomes maximum becomes in the past, and the shorter the required response time, the more difficult it is to satisfy real-time characteristics.

  On the other hand, it can be said that the information desired to be acquired by some signal analysis devices such as a pulsation detection device is the main frequency (principal component) of the time signal. That is, the signal waveform (signal amplitude) itself is not important, and there is no problem even if the waveform collapses. The main component of the frequency of the time signal is important. More specifically, in the case of a pulsation detecting device, it is desired to know the pulse rate of how much the pulse was beaten in one minute, but information on how strong the pulse was beaten is not always necessary.

  Therefore, the present applicant has devised a window function in which the timing at which the window function value becomes maximum is intentionally shifted from the middle of the sampling period. As can be seen from the above example, the frequency detected as a result of the frequency resolution processing has a strong correlation with the actual frequency at the timing at which the window function value is maximized, and the expected frequency at the timing at which the window function value is maximized. It can also be said.

  Therefore, a window function is used in which the timing at which the window function value is maximum is shifted from the middle of the sampling period toward the current timing. This increases the weight for the latest sample and compresses the signal level for the past sample. Thereby, compared with the case where window function processing is not performed, or the case where window function processing is performed using a symmetric window function, the real-time property of the frequency decomposition processing result (pulse rate) is improved.

  As a specific example, a window function such as Type 1 to Type 3 shown in FIG.

  The window functions of Type 1 to Type 3 are based on a trigonometric function and are configured in a frequency band of 0.5 Hz or less, and the frequency of the window function itself deviates from the frequency band of a pulse that can be taken by a normal human body. . Therefore, in the calculated spectrum, the pulse spectrum (0.5 Hz or more) and the window function spectrum are not confused.

  In the Type 3 window function, no signal level compression is performed on recent signal samples. As shown in the table comparing the characteristics of each frequency decomposition process in FIG. 4, when the values at both ends of the FFT interval (sampling interval) are 0 and the slope is 0 (Hanning window function, Type 2 window function described later, etc.) In comparison, the sidelobe characteristics are sacrificed, and the response to the case where the pulse wave detection signal contains a lot of noise is reduced. However, on the other hand, the real-time property can be further improved.

  In the Type 2 window function, the signal level is compressed so that the values at both ends of the input signal are 0 and the slope is 0. As shown in FIG. 4, it cannot be said that there is no effect of shifting the timing at which the window function value is maximized, but the pulse wave detection signal has no noise because it has better sidelobe characteristics than the Type 3 window function. Suitable for use when many are included.

  In the Type 1 window function, the signal level is compressed so that the values at both ends of the input signal become zero. However, as shown in FIG. 4, the sidelobe performance is slightly reduced as compared to Type 2 because the inclination of both ends is not zero. On the other hand, since the degree of compression of the signal is smaller than Type 2, even if the level of the pulse wave detection signal is low, the power spectrum value indicating pulsation tends to be clearer. A specific expression of the window functions of Type 1 to Type 3 will be described later.

4). First embodiment 4.1. Method Next, the method of the first embodiment will be described. The signal analysis device (beat detection device) 100 according to the first embodiment selects a frequency decomposition processing procedure according to the signal-noise ratio state of the pulse wave detection signal.

  First, when it is determined that the S / N ratio is good, that is, when it is determined that there is little mixing of body motion noise or sudden disturbance, power spectrum density estimation processing is performed as frequency resolution processing. Do. In this case, the window function processing using the asymmetric window function described above is not performed. This is because it is considered that sufficiently accurate frequency decomposition processing can be performed without performing window function processing. Specific methods of the power spectrum density estimation process include an AR (Auto-Regressive) method and a maximum entropy method.

  According to this, since the frequency distribution can be obtained based on the pulse wave detection signal waveform immediately before the observation timing, the real-time property can be further improved.

  Second, when it is determined that the state of the SN ratio is inferior, that is, when the proportion of body motion noise is high, or when sudden and large disturbance is detected, the frequency decomposition process is performed as described above. The window function processing is performed using the asymmetric window function of Type 1 or Type 2, and FFT is performed on the data after the window function processing.

  According to this, compared with the case where window function processing is not performed or the case where window function processing is performed using a general window function, the real-time property of the frequency decomposition processing result (pulse rate) is improved. Furthermore, compared with the case where the window function processing is performed using a Type 3 window function described later, the side lobes can be further suppressed and the side lobe characteristics can be improved. Therefore, it is easy to specify the pulsation spectrum even if the pulsation waveform has a small amplitude.

  Third, when it is determined that the S / N ratio state is intermediate between the above-described good state and the poor state, the window function using the above-described asymmetric window function of Type 3 is used as the frequency decomposition process. Processing is performed, and FFT is performed on the data after the window function processing.

  According to this, not only when the window function processing is not performed, but also when the window function processing is performed using a general window function or an asymmetric window function of Type 1 or Type 2, the frequency decomposition processing result (pulse rate). Real-time performance is improved. On the other hand, the side lobe suppression effect is lower than that of using Type 1 or Type 2 asymmetric window functions, but it is not a problem because it is applied when the SN ratio is not so bad in the first place.

4.2. Process Flow A specific process flow when the pulsation detecting device 100 measures the pulse rate as the pulsation information is shown in the flowchart of FIG. The detection unit 200 connected to the pulsation detection device 100 has an acceleration sensor as the body motion sensor 221 in addition to the pulse wave sensor 211.

  First, an acceleration detection signal is acquired as a pulse wave detection signal and a body motion detection signal from the detection unit 200 (S101). Based on the pulse wave detection signal, the signal state of the pulse wave detection signal is determined (S102). A specific method for determining the signal state will be described later.

  Next, body motion noise removal filter processing is performed on the pulse wave detection signal using the acceleration detection signal (S103). Then, based on the determined signal state, frequency resolution processing to be performed on the pulse wave detection signal is determined (S104). Then, frequency decomposition processing is actually performed on the pulse wave detection signal (S105), and the pulse frequency is specified based on the result of the frequency decomposition processing (S106). Then, the pulse frequency is converted into a pulse rate (S107), and the pulse rate is displayed on the display unit 70 (S108). Finally, it is determined whether or not to continue the measurement (S109). If it is determined that the measurement is not continued, the process is terminated. On the other hand, if it is determined that the measurement is to be continued, the processes of steps S100 to S110 are repeated.

  Next, a specific example of the signal state determination method in step S102 will be described with reference to the flowchart of FIG.

  First, pulse wave detection signals and body movement detection signals (acceleration detection signals) for 256 samples are acquired for 16 seconds (S200). Then, normal FFT processing is performed on the acquired pulse wave detection signal and body motion detection signal (acceleration detection signal) (S201), and a table as shown in FIG. 7A is created for each spectrum. In FIG. 7A, adr represents a spectrum number and val represents a spectrum value. Further, as shown in FIG. 7B, the spectrum of the pulse wave detection signal is sorted in descending order (S202).

  Then, the pulsation signal spectrum ratio is calculated (S203). Specifically, the size of the maximum spectrum and a predetermined spectrum (for example, the fifth or the tenth largest spectrum, but not limited to the fifth or the tenth) among the other spectra. Calculate the ratio. If this ratio is small, it is determined that there is little unsteady noise and the pulse wave detection signal is composed of limited frequency components. That is, it is determined that the S / N ratio is high (good state) because it does not include an extra frequency component and there is little noise. On the other hand, when this ratio is large, it is determined that the non-stationary noise is large or the frequency fluctuation is large within the observation period. That is, it is determined that the SN ratio is low (poor state).

  Specifically, whether or not the ratio pr5 between the largest spectrum and the fifth largest spectrum is smaller than 0.5, which is the first threshold, and the ratio pr10 between the largest spectrum and the tenth largest spectrum is It is determined whether or not the second threshold value is smaller than 0.2 (S204).

  The ratio pr5 between the largest spectrum and the fifth largest spectrum is smaller than 0.5, which is the first threshold, and the ratio pr10 between the largest spectrum and the tenth largest spectrum is more than 0.2, which is the second threshold. If it is determined that the SNR is also small, it is determined (estimated) that the S / N ratio is good (S / N ratio is high), and it is determined that the power spectrum density estimation process is performed as the frequency resolution process (S205).

  For example, in the example of FIGS. 7A to 7C, pr5 is calculated as pr5 = 1033.663 / 2973.406 = 0.376360, and pr10 = 503.500 / 297.406 = 0.1693344 Is calculated as Since pr5 is smaller than 0.5 and pr10 is smaller than 0.2, it is determined that the state of the SN ratio is good. The time waveform of the pulse wave detection signal at this time and the FFT result are as shown in FIG.

  On the other hand, if the determination condition is not satisfied in step S204, whether or not the ratio pr5 between the maximum spectrum and the fifth largest spectrum is smaller than 0.7, which is the third threshold, and the maximum spectrum It is determined whether or not the ratio pr10 between the spectrum and the tenth largest spectrum is smaller than the fourth threshold value 0.5 (S206).

  The ratio pr5 between the largest spectrum and the fifth largest spectrum is smaller than 0.7, which is the third threshold, and the ratio pr10 between the largest spectrum and the tenth largest spectrum is more than 0.5, which is the fourth threshold. Is also determined (estimated) as the signal-to-noise ratio is inferior (the signal-to-noise ratio is low), and it is determined to perform the window function processing and FFT using the asymmetric window function of Type 2 as the frequency resolution processing. (S207). Specifically, the examples of FIGS. 9A to 9C correspond.

  On the other hand, if the determination condition is not satisfied in step S206, it is determined (estimated) that the state of the S / N ratio is medium, and the window function processing using the asymmetric window function of Type 3 and FFT are performed as the frequency resolution processing. Determine (S208). Specifically, the examples of FIGS. 8A to 8C correspond.

  Moreover, the determination method of a signal state is not limited to said method, You may carry out by arbitrary methods.

The signal analysis device 100 of the present embodiment described above includes a signal acquisition unit 110 that acquires a detection signal from the detection unit 200 having a sensor, and sampling data D _{0 to} D _N (N is a positive integer of 2 or more) of the detection signal. And a signal processing unit 130 that performs a frequency decomposition process on the window function process for and the Fourier transform process on the data after the window function process. Here, the j-th sampling data D _j among the sampling data D _{0 to DN} is data sampled at a timing later than the timing at which the _i- th sampling data D _i is sampled (integer i, j is an integer satisfying 0 ≦ i <N / 2 <j ≦ N). The signal processing unit 130 then sets the window function value h (j) corresponding to the j-th sampling data D _j to the maximum value, and the window function value h (i) corresponding to the _i- th sampling data D _i is the window function. Window function processing is performed using an asymmetric window function that is smaller than the value h (j). Note that the window function processing is performed by the window function processing unit 133 in the signal processing unit 130, and the Fourier transform processing is performed by the Fourier transform processing unit 135 in the signal processing unit 130.

  Here, the frequency decomposition process is a process of decomposing (analyzing) the frequency included in the time signal, and there are various methods. In this embodiment, as the frequency decomposition process, a window function process and a Fourier transform process are performed. Do. Further, as described above, power spectrum density estimation processing may be performed as frequency decomposition processing instead of window function processing and Fourier transform processing.

The sampling data D ₀ to D _N of the detection signals is that sampled from the detection signal of the (N + 1) samples. N may be a positive integer greater than or equal to 2, but when performing FFT as described above, the number of samples needs to be a power of two. In this case, for example, N + 1 = 2, 4, 8, 16, 32, 64, 128, 192, 256, 512, 1024, 2048, 4096, etc. However, when performing Fourier transform other than FFT, the number of samples may not be a power of two.

In this embodiment, the window function value h (j) corresponding to the j-th sampling data D _j is the maximum value, and the window function value h (i) corresponding to the _i- th sampling data D _i is the window function value. An asymmetric window function having a value smaller than h (j) is used. The integers i and j are integers satisfying 0 ≦ i <N / 2 <j ≦ N.

In other words, in the present embodiment, the window function value h (i) corresponding to the _i th sampling data D _i is made smaller than the window function value h (j) corresponding to the _{j th} sampling data D _j , The signal value of the sample of 0 ≦ i <N / 2 is compressed larger than the sample of N / 2 <j ≦ N.

  For example, in the example of FIG. 3, N + 1 = 256, the integer i is an integer that satisfies 0 ≦ i ≦ 127, and the integer j is an integer that satisfies 127 ≦ j ≦ 255. Therefore, the window function value h (j) is maximized on the right side of the graph of FIG. 3, and the window function value h (i) on the left side of the graph is always smaller than h (j). All the window functions of Type 1 to Type 3 described above satisfy this condition.

  Thereby, in a window function process, the past sample can be compressed compared with the recent sample. Therefore, the real-time property of the frequency decomposition processing result (pulse rate) can be improved as compared with the case where the window function processing is not performed or the case where the window function processing is performed using the symmetric window function. In addition, if the window function processing is not performed or if the specification is sufficient with the same real-time characteristics as when the window function processing is performed using a symmetric window function, the sampling period can be made longer. A frequency distribution with high frequency resolution can be obtained.

  As described above, since the sampling data is closer to a signal waveform that can be expressed by a periodic function if the values at both ends are 0, the sidelobe characteristics can be improved when Fourier transform is performed.

Therefore, the signal processing unit 130, the asymmetric window function when the sampling data D ₀ to D window function corresponding to the N sampled data D _N of the _N values h (N) is the minimum value, performing window function processing May be.

  That is, in the example of FIG. 3, the window function value h (255) = 0 for the 255th sample number. In this case, the signal value multiplied by the window function value h (255) = 0 is 0. The window functions of Type 1 and Type 2 in FIG. 3 have such characteristics.

  As a result, it is possible to improve the side lobe characteristics compared to the case where the window function processing is not performed. In addition, the real-time property of the frequency resolution processing result (pulse rate) can be improved as compared with the case where the window function processing is performed using a symmetric window function. Specific application results will be described later with reference to FIGS. 10 (A) to 10 (C).

  Next, a specific expression of the asymmetric window function of Type 1 is illustrated.

The signal processing unit 130 calculates the p-th sampling data D _p among the sampling data D _{0 to DN} (integers p, M and N are integers satisfying 0 ≦ p ≦ 3M−1 and N = 4M−1). Is subjected to window function processing according to the following expression (1), and q-th sampling data D _q (integers q, M and N are integers satisfying 3M ≦ q ≦ 4M−1 and N = 4M−1) ) May be subjected to window function processing according to the following equation (2).

That is, in the example of FIG. 3 described above, because it is N + 1 = 256, M = 64, relative to the 0 ≦ p ≦ 191 sampling data _{D p,} performs window function processing by the formula (1), 192 ≦ q for ≦ 255 sampling data _{D q,} performs window function processing by the following equation (2).

  In addition, the asymmetric window function of Type 1 has a smaller signal compression degree than the window function of Type 2. Therefore, even when the level of the pulse wave detection signal is low, the power spectrum value indicating pulsation is easily clarified, and the pulse frequency can be easily specified.

  Next, a specific expression of the asymmetric window function of Type 2 is illustrated.

The signal processing unit 130 calculates the p-th sampling data D _p among the sampling data D _{0 to DN} (integers p, M and N are integers satisfying 0 ≦ p ≦ 3M−1 and N = 4M−1). Is subjected to window function processing according to the following expression (3), and q-th sampling data D _q (integers q, M and N are integers satisfying 3M ≦ q ≦ 4M−1 and N = 4M−1) ) May be subjected to window function processing according to the following equation (4).

That is, in the example of FIG. 3 described above, since N + 1 = 256 and M = 64, the window function processing is performed on the sampling data D _p with 0 ≦ p ≦ 191 according to Equation (3), and 192 ≦ q for ≦ 255 sampling data _{D q,} performs window function processing by the following equation (4).

  As a result, in the Nth sampling data, not only the signal value but also the slope can be set to zero.

  10A to 10C show the results of each frequency decomposition process. The upper diagram of each figure shows the time waveform of the pulse wave detection signal for 16 seconds, and the lower diagram of each figure shows the FFT result of the pulse wave detection signal shown in the upper diagram. FIG. 10A shows an example in the case where window function processing is not performed, and FIG. 10B shows an example in which window function processing is performed using an asymmetric window function of Type 3 to be described later, and FIG. Shows an example in which window function processing is performed using an asymmetric window function of Type2.

  As shown by the FFT results in FIGS. 10A to 10C, when the FFT is applied to the signal after the window function processing by the asymmetric window function, the distribution of the power spectrum is narrowed. That is, the side lobe is suppressed and the side lobe characteristics are enhanced. This is an effect obtained by compressing the past signal portion. As a result, it becomes easier to specify the frequency indicating the pulse. Furthermore, when this result is compared, it can be seen that the result when the Type 2 asymmetric window function is used (FIG. 10C) has the best sidelobe characteristics among the three FFT results.

  Therefore, if the asymmetric window function of Type 2 is used, the sidelobe characteristics can be further improved, and the pulse frequency can be specified even in a state where the noise is severe.

  Further, as described above, for example, when it is determined that the state of the SN ratio is good, that is, when it is determined that there is little body motion noise or sudden disturbance, side lobes are not generated so much. It is not always necessary to improve the lobe characteristics. In this case, it is desirable to perform window function processing that emphasizes real-time characteristics rather than improving the sidelobe characteristics.

Therefore, the signal processing unit 130, the asymmetric window function when the sampling data D ₀ to D window function corresponding to the N sampled data D _N of the _N values h (N) is the maximum value, performs window function processing May be.

  For example, if the window function value h (N) = 1, the signal value of the Nth sample data is not compressed. Even if h (N) = 1, the compression rate can be made the lowest for other sample data.

  According to this, not only when performing Fourier transform without performing window function processing as shown in FIG. 10 (A), but also by performing window function processing using a general window function or an asymmetric window function of Type 1 or Type 2. Compared to the case, it is possible to improve the real-time property of the frequency decomposition processing result (pulse rate).

  Further, not only the Nth sampling data but also sampling data that is traced back by a predetermined number from the Nth sampling data, if the window function value is increased and the compression rate is lowered, the frequency decomposition processing result (pulse rate) Number) real-time characteristics are improved.

Therefore, the signal processing unit 130 converts the sampling data D _{0 to DN} to the k-th sampling data D _k to the N-th sampling data D _N (the integer k is an integer satisfying N / 2 <k <N). The window function processing may be performed by an asymmetric window function in which the corresponding window function values h (k) to h (N) are all the maximum values.

  For example, in the example of the asymmetric window function of Type 3 shown in FIG. 3, the window functions h (192) to h (255) corresponding to the 192nd to 255th sampling data are all set to 1, which is the maximum value. .

  As a result, not only the Nth sampling data but also sampling data that is traced a predetermined number of times from the Nth sampling data can be maximized to reduce the compression rate. Therefore, compared with the case where window function processing is not performed, the case where window function processing is performed using a symmetric window function, or the case where window function processing is performed using an asymmetric window function of Type 1 or Type 2, frequency decomposition processing is performed. It is possible to improve the real-time result (pulse rate).

  Next, a specific expression of the Type 3 asymmetric window function is illustrated.

The signal processing unit 130 calculates the p-th sampling data D _p among the sampling data D _{0 to DN} (integers p, M and N are integers satisfying 0 ≦ p ≦ 3M−1 and N = 4M−1). Is subjected to window function processing according to the following equation (5), and q-th sampling data D _q (integers q, M and N are integers satisfying 3M ≦ q ≦ 4M−1 and N = 4M−1) ) May be subjected to window function processing according to the equation h (q) = 1.

That is, in the example of FIG. 3 described above, because it is N + 1 = 256, M = 64, relative to the 0 ≦ p ≦ 191 sampling data _{D p,} performs window function processing by the equation (5), 192 ≦ q for ≦ 255 sampling data _{D q,} performs window function processing by the formula h (q) = 1.

  As a result, it is possible to realize not only the Nth sampling data but also the maximum window function value for the sampling data that is traced a predetermined number from the Nth sampling data. For example, in the graph shown in the upper diagram of FIG. 10B, it can be seen that the signal values of 64 samples on the right side of the graph are not compressed from the original waveform shown in the upper diagram of FIG.

  The window function value may not be calculated every time the window function process is performed, but a value calculated in advance may be held in the array variable. Thereby, the amount of calculation can be reduced.

  Now, it is desirable that the above asymmetric window functions from Type 1 to Type 3 are properly used according to the signal state as described above. Further, the frequency decomposition process is not limited to the window function process and the FFT process, and it may be better to perform the power spectrum density estimation process.

  Therefore, the pulsation detection device 100 of the present embodiment is a signal processing unit that performs frequency resolution processing based on the pulse wave detection signal from the pulse wave detection unit 210 and the body motion detection signal from the body motion detection unit 220. 130, a pulsation information calculation unit 170 that calculates pulsation information based on the result of the frequency decomposition process, and a signal state determination that performs a signal state determination process on at least one of the pulse wave detection signal and the body movement detection signal Part 150 may be included. Then, the signal processing unit 130 performs different frequency resolution processing according to the determination result of the signal state by the signal state determination unit 150.

  This makes it possible to perform different frequency resolution processing depending on the signal state.

  In addition, since the window function processing described above should switch the window function to be used depending on whether or not the pulse wave detection signal contains a lot of noise, can it be estimated that the SN ratio is high or low? The signal state may be determined based on such a determination.

  That is, when it is determined that the pulse wave detection signal is in the first signal state, the signal processing unit 130 performs the first frequency decomposition process, and the pulse wave detection signal is SN more than the first signal state. When it is determined that the second signal state has a high ratio, the second frequency decomposition process may be performed.

  As a result, it is possible to perform different frequency resolution processing according to signal states having different SN ratios.

  Further, depending on the signal state, power spectrum density estimation processing or the like may be performed as frequency resolution processing without performing window function processing as in the above-described example.

  That is, the signal analysis apparatus 100 according to the present embodiment may include a signal state determination unit 150 that performs a signal state determination process on the detection signal acquired by the signal acquisition unit 110. When it is determined that the signal state is the first signal state, the signal processing unit 130 performs frequency resolution processing on window function processing using an asymmetric window function and Fourier transform processing on the data after the window function processing. If the signal state is determined to be the second signal state, the power spectrum density estimation process for the detection signal may be performed as a frequency decomposition process. It is assumed that the power spectrum density estimation process is performed by the power spectrum density estimation processing unit 137 in the signal processing unit 130.

  That is, if the signal analysis device 100 is limited to the pulsation detection device 100, in other words, the signal processing unit 130 performs the first frequency decomposition when it is determined that the pulse wave detection signal is in the first signal state. As processing, when performing window function processing on the sampling data of the pulse wave detection signal and Fourier transform processing on the data after the window function processing, and determining that the pulse wave detection signal is in the second signal state, A power spectrum density estimation process may be performed as the second frequency resolution process.

  Here, the power spectrum density estimation processing is performed when FFT processing is performed on a long-term observation signal even when only a short-term observation signal is obtained, compared with a case where the frequency distribution is calculated by normal FFT. This is a calculation algorithm capable of obtaining the same frequency resolution and frequency distinction. In fields other than pulse detection, it is often used in, for example, earthquake occurrence cycle analysis and ocean current analysis.

  Specifically, when performing the power spectrum density estimation process, for example, it is possible to calculate the pulse rate using only the signal waveform about 2 seconds before the present time. In that case, the frequency component included in the pulse wave detection signal in the past can be ignored than two seconds ago.

  As a result, when the S / N ratio is the first signal state lower than the second signal state, the frequency decomposition process with a high sidelobe suppression effect is performed, and the S / N ratio is higher than that of the first signal state. In the case of the signal state, it is possible to perform frequency decomposition processing that emphasizes real-time characteristics.

Further, when the signal processing unit 130 determines that the pulse wave detection signal is in the first signal state, the window function value h (j) corresponding to the _j-th sampling data D _j is the maximum value, The window function processing may be performed by an asymmetric window function in which the window function value h (i) corresponding to the i-th sampling data D _i is smaller than the window function value h (j).

  This makes it possible to improve the real-time property of the frequency decomposition processing result (pulse rate) compared to the case where the window function processing is not performed or the case where the window function processing is performed using a symmetric window function.

  Further, when it is determined that the signal state is the first signal state, the signal processing unit 130 performs window function processing using the first asymmetric window function, and the signal state is the second signal state. If it is determined, a window function process using a second asymmetric window function different from the first asymmetric window function may be performed.

  That is, if the signal analysis device 100 is limited to the pulsation detection device 100, in other words, the signal processing unit 130 performs the first frequency decomposition when it is determined that the pulse wave detection signal is in the first signal state. As the processing, the window function processing for the sampling data of the pulse wave detection signal using the first window function and the Fourier transform processing for the first data after the window function processing are performed, and the pulse wave detection signal becomes the second If it is determined that the signal state is present, a window function process for sampling data of the pulse wave detection signal using a second window function different from the first window function as the second frequency resolution process; You may perform the Fourier-transform process with respect to the 2nd data after a function process.

Further, when the signal processing unit 130 determines that the pulse wave detection signal is in the first signal state, the window function value h (j) corresponding to the _j-th sampling data D _j is the maximum value, The window function processing is performed using the asymmetric window function in which the window function value h (i) corresponding to the i-th sampling data D _i is smaller than the window function value h (j) as the first window function, and the pulse wave When it is determined that the detection signal is in the second signal state, the window function processing may be performed using a second window function that is an asymmetric window function different from the first window function.

  That is, when the signal state is the first signal state and when the signal state is the second signal state, the window function processing may be performed using different asymmetric window functions.

  For example, when the signal state is the first signal state, the window function processing is performed using the above-described asymmetric window function of Type 3 as the first window function, and the signal state is the second signal state. Can perform window function processing using the Type 2 asymmetric window function described above as the second window function. In some cases, a symmetric window function may be used.

  This makes it possible to perform different window function processing depending on the signal state. Furthermore, it becomes possible to change the asymmetric window function used for the window function processing according to the signal state.

  Further, when determining the signal state based on the state of the SN ratio as described above, it is necessary to obtain the SN ratio, but since it is difficult to directly obtain the SN ratio, the SN ratio is obtained from other information. It is necessary to estimate.

  Therefore, the signal state determination unit 150 includes at least one second spectrum other than the first spectrum and the first spectrum indicating the maximum spectrum value among the frequency spectra obtained by the frequency resolution processing of the pulse wave detection signal. A signal state determination process may be performed by calculating a ratio of a spectrum value to a spectrum. Note that the spectrum value calculation process is performed by the spectrum value calculation unit 151 of the signal state determination unit 150.

  Specifically, it is as described with reference to the flowchart of FIG.

  Thereby, it is possible to estimate the signal-to-noise ratio and obtain the signal state.

5. Second Embodiment 5.1. Method Next, the method of the second embodiment will be described. As described above, in order to increase the resolution of the frequency resolution processing, it is necessary to lengthen the detection period (sampling period) of the pulse wave detection signal used for one pulse rate calculation process. When the pulse does not change during this detection period, the specific spectrum clearly shows the peak value, and therefore it is relatively easy to specify the pulse frequency.

  On the other hand, when the pulse rate has changed greatly within the detection period, the spectrum is dispersed over a wide band, so that it is difficult to specify the peak spectrum. For example, if jogging is started from the state of sitting in the above-described detection period (pulse 60) and the pulse rises to about 150, the pulse spectrum is dispersed in the interval of 1 to 2.5 Hz. Since the magnitudes of the individual frequency spectra are also arranged side by side, the difficulty of condition determination increases which frequency should be determined as the pulse frequency.

  Therefore, the signal analyzing apparatus (beat detecting apparatus) 100 according to the second embodiment determines an exercise state based on the body movement detection signal, and selects a frequency resolution processing procedure according to the exercise state.

  When it is determined that the motion state is a resting state, that is, when only a minute motion is detected in a body motion detection signal such as an acceleration detection signal, a power spectrum density estimation process is performed as a frequency resolution process.

  Next, when it is determined that the motion state is a steady motion state, the window function processing is performed using the Type 3 window function as the frequency resolution processing, and the FFT is performed on the data after the window function processing. Do. Specifically, the steady motion state refers to a state where the subject is walking, a state where the subject is jogging, or the like, and refers to a state where the subject moves his / her body at a substantially constant cycle (rhythm). In other words, the steady motion state refers to a state of performing a motion having periodicity.

  When it is determined that the motion state is an unsteady motion state, the window function processing is performed using the window function of Type 1 or Type 2 as the frequency resolution processing, and the data after the window function processing is performed. Perform FFT. Specifically, the unsteady state of motion refers to a state of physical exercise or a state of playing a ball game, etc., and a state in which the subject is moving in a disjoint cycle (unspecified cycle). Say. That is, the non-stationary movement state refers to a state where the movement is not periodic.

  Furthermore, when it is detected that the motion state has changed from a resting state to a steady motion state or an unsteady motion state, that is, when it is detected that the subject has started motion, Window function processing is performed using the window functions of Type 1 and Type 2, and FFT is performed on the data after the window function processing.

  On the other hand, when it is detected that the motion state has changed from a steady motion state or an unsteady motion state to a resting state, that is, when it is detected that the subject has stopped moving, power is used as a frequency decomposition process. Perform spectral density estimation.

5.2. Process Flow A specific process flow when the pulsation detecting device 100 measures the pulse rate as the pulsation information is shown in the flowchart of FIG. The detection unit 200 connected to the pulsation detection device 100 has an acceleration sensor as the body motion sensor 221 in addition to the pulse wave sensor 211.

  The basic processing flow is the same as the processing flow shown in FIG. 5 of the first embodiment, but step S302 corresponding to step S102 is different. In the present embodiment, the motion state of the subject is determined based on the acceleration detection signal (S302).

  Next, a specific example of the determination method of the exercise state in step S302 will be described using the flowchart of FIG.

  First, based on the acceleration detection signal, it is determined whether or not the user is moving (S400).

  When it is determined that the user has a motion, frequency resolution processing is performed on the acceleration detection signal data (S401). Note that the frequency decomposition processing performed here is not frequency decomposition processing performed on the pulse wave detection signal. Then, it is determined whether or not it is in steady motion (S402), and when it is determined that it is in a steady motion state, the steady motion state is set as a temporary determination result of the motion state (S403).

  On the other hand, when it is determined that it is not a steady motion state, the unsteady motion state is set as a temporary determination result of the motion state (S404).

  Next, after step S403 or S404, it is determined whether or not the exercise state at the previous determination was a resting state (S405).

  If the exercise state at the time of the previous determination is a resting state, it is determined that the start of exercise has been detected (S406), and the exercise state determination process ends.

  On the other hand, if it is determined that the exercise state at the time of the previous determination is not a resting state, it is determined whether or not the exercise state at the time of the previous determination matches the provisional exercise state determined in step S403 (S407). That is, it is determined here whether or not the exercise has changed.

  If the exercise state at the time of the previous determination coincides with the provisional exercise state, it is considered that the exercise state has not changed and the same exercise is continued, so the provisional exercise state is determined as the determination result (S408). The exercise state determination process is terminated.

  On the other hand, if the exercise state at the time of the previous determination does not match the provisional exercise state, it is considered that the same exercise is not continued, so it is determined that the exercise state is changing (S409), and the exercise state determination process is performed. finish.

  Furthermore, when it is determined in step S400 that the user does not move, the provisional determination result is set to a resting state (S410). If the exercise state at the time of the previous determination is not a resting state, it is determined that an exercise stop has been detected (S412), and the exercise state determination process ends.

  On the other hand, if the exercise state at the time of the previous determination is a resting state, it is determined that there is no movement in the user, it is determined that the user is in a resting state (S413), and the exercise state determination process ends. The above is the flow of the motion state determination process.

  Next, FIG. 13 shows a graph comparing the measurement results of the pulse rate when the present embodiment is applied and when the present embodiment is not used.

  The measurement result shown in FIG. 13 shows the time transition of the pulse rate when jogging for about 1 minute from the walking state (around 300 seconds) and walking again from around 360 seconds. By applying this embodiment, it can be said that the rise and fall of the pulse rate can be observed at an earlier timing with respect to the elapsed time.

  In the above embodiment, the detection unit 200 may further include a second sensor in addition to the first sensor that is a sensor. At this time, the signal acquisition unit 110 performs the second detection based on the first detection signal based on the first sensor signal acquired by the first sensor and the second sensor signal acquired by the second sensor. The signal may be acquired from the detection unit 200.

  And the signal state determination part 150 may perform the process which determines the signal state of a 2nd detection signal as determination processing. Further, when it is determined that the signal state of the second detection signal is the first signal state, the signal processing unit 130 performs the window function processing using the first asymmetric window function as the asymmetric window function. When it is determined that the signal state of the second detection signal is the second signal state, the second asymmetric window function different from the first asymmetric window function is used. Thus, the window function processing may be performed on the first detection signal.

  Accordingly, it is possible to change the asymmetric window function used for the window function processing according to the signal state of the second detection signal different from the first detection signal, and perform the frequency decomposition processing.

  In addition, when the signal processing unit 130 determines that the signal state of the second detection signal is the first signal state, the first detection signal using the first asymmetric window function as the asymmetric window function. When the window function processing for and the Fourier transform processing for the data after the window function processing are performed as frequency decomposition processing, and it is determined that the signal state of the second detection signal is the second signal state, the first The power spectrum density estimation process for the detected signal may be performed as a frequency resolution process.

  As a result, the signal state is determined according to the signal state of the second detection signal. For example, when the first signal state is lower than the second signal state in the S / N ratio, the frequency having a high sidelobe suppression effect is obtained. When the decomposition process is performed and the second signal state is higher than the first signal state in the SN ratio, it is possible to perform a frequency decomposition process that emphasizes real-time characteristics.

  Further, the signal state determination unit 150 may perform a process of determining the motion state of the subject based on the second detection signal as the determination process. And the signal processing part 130 may determine the process performed with respect to a 1st detection signal based on the determination result of an exercise state. Note that the exercise state determination process is performed by the exercise state determination unit 153 included in the signal state determination unit 150.

  That is, in other words, limiting the signal analysis device 100 to the pulsation detection device 100, the signal state determination unit 150 performs a process of determining the exercise state of the subject based on the body movement detection signal as the signal state determination process. You may go. And the signal processing part 130 may perform different frequency decomposition processing according to an exercise state.

  As described above, specific motion states include a rest state, a steady motion state, and an unsteady motion state. Here, a specific example of the determination process of the motion state when an acceleration sensor is used as the body motion sensor will be described with reference to FIGS. 14 (A) to 14 (C). FIG. 14A to FIG. 14C show spectra of acceleration detection signals in each motion state. The upper part of each figure shows the time waveform of the acceleration detection signal for 16 seconds, and the lower part of each figure shows the FFT result of the acceleration detection signal shown in the upper part.

  First, in the case of FIG. 14A, since the amplitude of the acceleration detection signal is equal to or less than a predetermined threshold value, it is determined as “rest state”. In the case of FIG. 14B, since the amplitude of the acceleration detection signal is greater than or equal to a predetermined threshold, it is determined that “there is motion”. In addition, since there is no relatively large power spectrum in a specific frequency range, it is determined as an “unsteady motion state”. Further, in the case of FIG. 14C, as in the case of FIG. 14B, the amplitude of the acceleration detection signal is greater than or equal to a predetermined threshold value, and therefore it is determined that “there is motion”. In addition, since there is a relatively large power spectrum in a specific frequency range, it is determined as a “steady motion state”.

  This makes it possible to perform different frequency resolution processing depending on the motion state.

  Specifically, when it is determined that the motion state is a steady motion state, the signal processing unit 130 performs a first frequency decomposition process, and the motion state is determined to be an unsteady motion state. In such a case, the second frequency decomposition process may be performed.

  Thereby, it is possible to perform different frequency resolution processing, for example, when the pulse wave detection signal has periodicity and when it does not have periodicity.

  Furthermore, when it is determined that the motion state is a steady motion state, the signal processing unit 130 uses the first window function as a first frequency decomposition process and performs sampling on the pulse wave detection signal sampling data. When the window function process and the Fourier transform process on the first data after the window function process are performed and it is determined that the motion state is an unsteady motion state, the first frequency decomposition process is performed as the first frequency decomposition process. A window function process for the sampling data of the pulse wave detection signal and a Fourier transform process for the second data after the window function process may be performed using a second window function different from the above window function.

  This makes it possible to perform window function processing using different window functions, for example, when the pulse wave detection signal has periodicity and when it does not have periodicity.

Here, the j-th sampling data D _j among the sampling data D _{0 to} D _N (N is a positive integer of 2 or more) of the pulse wave detection signal is based on the timing at which the _i- th sampling data D _i is sampled. (Integers i and j are integers satisfying 0 ≦ i <N / 2 <j ≦ N).

At this time, if the signal processing unit 130 determines that the motion state is a steady motion state, the window function value h (j) corresponding to the _jth sampling data Dj becomes the maximum value, and the i th The window function processing is performed using the asymmetric window function in which the window function value h (i) corresponding to the sampling data D _i is smaller than the window function value h (j) as the first window function, and the motion state is not When it is determined that the stationary motion state is present, the window function processing may be performed using a second window function that is an asymmetric window function different from the first window function.

  That is, the window function used when the motion state is a steady motion state and the window function used when the motion state is an unsteady motion state may be different from each other.

  This makes it possible to perform window function processing using different asymmetric window functions, for example, when the pulse wave detection signal has periodicity and when it does not have periodicity.

  Further, when the exercise state is a resting state, the pulse wave detection signal hardly includes noise due to body movement or the like.

  Therefore, when it is determined that the motion state is a resting state, the signal processing unit 130 performs the first frequency decomposition process, and determines that the motion state is a steady motion state or an unsteady motion state. In such a case, the second frequency decomposition process may be performed.

  Accordingly, it is possible to perform different frequency decomposition processing, for example, when the motion state is a resting state and when the motion state is a steady motion state or an unsteady motion state.

  Specifically, when it is determined that the exercise state is the resting state, the signal processing unit 130 performs the power spectrum density estimation process as the first frequency decomposition process, and the exercise state is the in-exercise state. If it is determined, as the second frequency decomposition process, a window function process for the sampling data of the pulse wave detection signal and a Fourier transform process for the data after the window function process may be performed.

  As a result, when the motion state is a steady motion state or an unsteady motion state, frequency decomposition processing with a high side lobe suppression effect is performed, and when the motion state is a resting state, real-time characteristics are further improved. It is possible to perform an important frequency decomposition process.

Further, when it is determined that the motion state is a steady motion state or an unsteady motion state, the signal processing unit 130 has a maximum window function value h (j) corresponding to the _j-th sampling data D _j. The window function processing is performed using the asymmetric window function in which the window function value h (i) corresponding to the _i th sampling data D _i is smaller than the window function value h (j) as the first window function. May be.

  Thus, when the motion state is a steady motion state or an unsteady motion state, it is possible to perform a window function process using an asymmetric window function and perform a frequency resolution process with improved real-time characteristics.

  Further, it may be determined whether or not exercise has started based on the spectrum of the acceleration detection signal, and the frequency resolution processing to be executed may be switched based on the determination result. FIG. 15A shows the spectrum of the acceleration detection signal before the start of exercise, and FIG. 15B shows the spectrum of the acceleration detection signal immediately after the start of exercise. A specific determination method is as shown in the flowchart of FIG. In this way, by observing the waveform of the acceleration detection signal, it can be determined whether or not exercise has started.

  Further, it may be determined whether or not the exercise is stopped based on the spectrum of the acceleration detection signal, and the frequency resolution processing to be executed may be switched based on the determination result. FIG. 16A shows the spectrum of the acceleration detection signal during exercise, and FIG. 16B shows the spectrum of the acceleration detection signal immediately after stopping the exercise. A specific determination method is as shown in the flowchart of FIG. Thus, by observing the waveform of the acceleration detection signal, it is possible to determine whether or not the exercise has been stopped.

  Now, the use of the second detection signal described above is not limited to the determination of the signal state.

  That is, the signal processing unit 130 may perform signal processing on the first detection signal based on the second detection signal.

  Specifically, the signal processing unit 130 (noise reduction processing unit 131) may perform the noise reduction processing on the first detection signal as the signal processing described above based on the second detection signal.

  Here, in the pulsation detection device 100, it is assumed that various noises are mixed in the pulse wave sensor signal acquired from the pulse wave sensor (or the pulse wave detection signal acquired therefrom). Therefore, if noise reduction processing that reduces body movement noise components included in the pulse wave sensor signal is not performed, the detected pulsation information will be affected by noise, so the value of the pulsation information is attached May not accurately represent the person's actual beat. In this case, if the pulsation information is simply output, the user's judgment based on the pulsation information may be mistaken.

  Therefore, the noise reduction processing unit 131 may perform processing for reducing noise (body motion noise component) due to body motion from the pulse wave detection signal using the body motion detection signal that is the first detection signal. .

  Here, the body motion noise component (body motion noise component signal) is a component signal included in the pulse wave sensor signal, which is a blood vessel generated due to human steady motion / motion (body motion) or the like. A component signal indicating a change in volume. For example, in the case of a pulse meter attached to an arm or a finger, the volume of the blood vessel changes in accordance with the rhythm of the arm swing due to the effect of the arm swing while walking or jogging. When a human performs a steady motion in this way, the body motion noise component becomes a periodic signal, and becomes a component signal having the frequency of the motion. In addition, the body motion noise component has a feature that it has a high correlation with the waveform of a signal output from the acceleration sensor mounted in the vicinity of the site where the pulse wave sensor is mounted.

  The pulse wave sensor signal acquired from the pulse wave sensor 211 includes a component caused by body movement in addition to a component caused by heartbeat. The same applies to the pulse wave detection signal (pulse wave sensor signal after the DC component cut) used for the calculation of pulsation information. Of these, components useful for the calculation of pulsation information are components caused by heartbeats, and components caused by body movements obstruct the calculation. Therefore, a signal (body motion detection signal) resulting from body motion is acquired using a body motion sensor, and a signal component correlated with the body motion detection signal (referred to as an estimated body motion noise component) is removed from the pulse wave detection signal. By doing so, the body movement noise component contained in the pulse wave detection signal is reduced. However, even if the body motion noise component in the pulse wave detection signal and the body motion detection signal from the body motion sensor are signals resulting from the same body motion, the signal level is not necessarily the same. . Therefore, for example, an estimated body motion noise component is calculated by performing filter processing in which a filter coefficient is adaptively determined for the body motion detection signal, and only the pulse wave detection signal and the calculated estimated body motion noise component are included. The difference from the signal is taken.

  As a result, it is possible to reduce the noise included in the first detection signal, improve the sidelobe characteristics after Fourier transform, and the like.

  Signal processing such as noise reduction processing can also be performed in the first embodiment. If noise can be reduced by the noise reduction processing, the signal state of the first embodiment described above changes, so that it is possible to perform frequency decomposition processing that places more emphasis on real-time performance.

6). Other Application Examples The signal analysis device (beat detection device) 100 described above can be used not only for pulse measurement but also for other measurement (analysis). For example, it can be used in any field as long as various sensor signals are measured and the period, speed, frequency, and the like are calculated based on the frequency decomposition result. As a result, it is possible to calculate the period, speed, frequency, etc. with improved real-time characteristics while maintaining the same resolution as the conventional method.

  More specifically, the signal analysis apparatus 100 can be used for pitch measurement for measuring the number of steps per unit time, measurement of a repetition period of movement based on an acceleration detection signal, and the like. In addition, heart rate variability analysis (autonomic nerve analysis) that estimates the function of sympathetic and parasympathetic nerves by frequency analysis of heart rate interval fluctuations, respiratory cycle analysis, blood pressure regulation cycle analysis, self-power generation included in brain waves It can be used for many measurements and analyses, such as measurement of position, climate change analysis, and periodic analysis of sunspot activity.

  In addition, the signal analysis device (beat detecting device) 100 and the electronic device or the like of the first embodiment or the second embodiment described above may realize part or most of the processing by a program.

  As a result, the processing of the first embodiment or the second embodiment can be realized by a program. The program may be a program that is read and executed by a processing unit (for example, a DSP) of a device such as a smartphone.

  Further, the pulse wave detection device worn by the user may include a pulse wave sensor 211 and a communication unit that communicates a pulse wave sensor signal from the pulse wave sensor 211 wirelessly or by wire. In that case, the program of the first embodiment or the second embodiment is provided as a separate body from the pulse wave detection device, and a processing unit (for example, an information processing system that receives a pulse wave sensor signal from the communication unit described above (for example, CPU) and the like are executed. This information processing system may not be assumed to be worn by a user such as a PC, or may be assumed to be worn (mobile) by a user such as a smartphone. Further, a server system or the like connected via a network such as the Internet may be used as the information processing system.

  When the pulse wave detection device and the information processing system on which the program is executed are separate, the display unit 70 used for presenting pulsation information to the user is provided at an arbitrary location. For example, you may display on the display part 70 of an information processing system, the display part 70 may be provided in a pulse wave detection device, and the pulsation information output from the information processing system may be displayed. Moreover, you may display on the display part 70 of a different apparatus (for example, arbitrary client apparatuses, etc. when a server system is used as an information processing system).

  The above program is recorded on an information storage medium. Here, as the information recording medium, various recording media that can be read by an information processing system such as an optical disk such as a DVD or a CD, a magneto-optical disk, a hard disk (HDD), a memory such as a nonvolatile memory or a RAM can be assumed. .

  As described above, the first embodiment and the second embodiment have been described in detail. However, it is easy for those skilled in the art that many modifications that do not substantially depart from the novel matters and effects of the present invention are possible. You can understand. Accordingly, all such modifications are intended to be included in the scope of the present invention. For example, a term described at least once together with a different term having a broader meaning or the same meaning in the specification or the drawings can be replaced with the different term in any part of the specification or the drawings. Further, the configuration and operation of the signal analysis device (beat detecting device) 100, the electronic device, the program, and the like are not limited to those described in the first embodiment and the second embodiment, and various modifications can be made. is there.

70 display unit, 100 signal analysis device (beat detection device), 110 signal acquisition unit,
130 signal processing units, 131 noise reduction processing units, 133 window function processing units,
135 Fourier transform processing unit, 137 power spectrum density estimation processing unit,
150 signal state determination unit, 151 spectrum value calculation unit, 153 exercise state determination unit, 170 pulsation information calculation unit, 200 detection unit, 210 pulse wave detection unit,
211 pulse wave sensor, 215 filter processing unit, 216 A / D conversion unit,
220 body motion detector, 221 body motion sensor (motion sensor),
225 filter processing unit, 226 A / D conversion unit, 300 holding mechanism,
302 guide, 400 base

Claims (16)

A signal acquisition unit for acquiring a detection signal from a detection unit having a sensor;
A signal processing unit that performs window function processing on the sampling data D _{0 to} D _N (N is a positive integer of 2 or more) of the detection signal and Fourier transform processing on the data after the window function processing as frequency decomposition processing;
Including
Sampling data _{D j} of the j-th of the sampling data _D 0 to D _N, the
Data sampled at a timing later than the timing at which the _i- th sampling data D _i is sampled (integers i and j are integers satisfying 0 ≦ i <N / 2 <j ≦ N),
The signal processing unit
The window function value h (j) corresponding to the j-th sampling data D _j is the maximum value, and the window function value h (i) corresponding to the _i- th sampling data D _i is the window function value h (j). A signal analysis apparatus that performs the window function processing using an asymmetric window function having a smaller value. In claim 1,
The signal processing unit
By the asymmetric window function the sampling data D ₀ to D window function corresponding to the N sampled data D _N of the _N values h (N) is the minimum value, and performs the window function processing Signal analysis device. In claim 1,
The signal processing unit
By the sampling data D ₀ to D the asymmetric window function N-th sampling data D corresponding to the _N window function value h (N) is the maximum value of _N, and performs the window function processing Signal analysis device. In claim 3,
The signal processing unit
Sampling data _{D N} of the sampling data _{D k} ~ N-th of the first k of the sampling data _D 0 to D _N (integer k is, N / 2 <k <integer satisfying N) window function corresponding to the value h ( The signal analysis apparatus characterized in that the window function processing is performed by the asymmetric window function in which k) to h (N) are all maximum values. In claim 2,
The signal processing unit
For the p-th sampling data D _p among the sampling data D _{0 to DN} (integers p, M and N are integers satisfying 0 ≦ p ≦ 3M−1 and N = 4M−1), The window function processing is performed by the following equation (1),

For the q-th sampling data D _q (integers q, M and N are integers satisfying 3M ≦ q ≦ 4M−1 and N = 4M−1), the following equation (2):

A signal analysis apparatus that performs the window function processing. In claim 2,
The signal processing unit
For the p-th sampling data D _p among the sampling data D _{0 to DN} (integers p, M and N are integers satisfying 0 ≦ p ≦ 3M−1 and N = 4M−1), The window function processing is performed by the following equation (3),

For the q-th sampling data D _q (integers q, M and N are integers satisfying 3M ≦ q ≦ 4M−1 and N = 4M−1),

A signal analysis apparatus that performs the window function processing. In claim 2,
The signal processing unit
For the p-th sampling data D _p among the sampling data D _{0 to DN} (integers p, M and N are integers satisfying 0 ≦ p ≦ 3M−1 and N = 4M−1), The window function processing is performed by the following equation (5),

For the q-th sampling data D _q (integers q, M and N are integers satisfying 3M ≦ q ≦ 4M−1 and N = 4M−1), the window is given by the equation h (q) = 1. A signal analysis apparatus characterized by performing function processing. In any one of Claims 1 thru | or 7,
A signal state determination unit that performs a signal state determination process on the detection signal acquired by the signal acquisition unit;
The signal processing unit
When it is determined that the signal state is the first signal state, the window function processing by the first asymmetric window function is performed,
When it is determined that the signal state is the second signal state, the signal processing apparatus performs the window function processing using a second asymmetric window function different from the first asymmetric window function. In any one of Claims 1 thru | or 7,
A signal state determination unit that performs a signal state determination process on the detection signal acquired by the signal acquisition unit;
The signal processing unit
When it is determined that the signal state is the first signal state, the window function processing by the asymmetric window function and the Fourier transform processing on the data after the window function processing are performed as the frequency decomposition processing. ,
When it is determined that the signal state is the second signal state, a power spectrum density estimation process for the detection signal is performed as the frequency decomposition process. In any one of Claims 1 thru | or 7,
A signal state determination unit that performs a signal state determination process on the detection signal acquired by the signal acquisition unit;
The detector is
In addition to the first sensor that is the sensor, it further has a second sensor,
The signal acquisition unit
A first detection signal based on a first sensor signal acquired by the first sensor and a second detection signal based on a second sensor signal acquired by the second sensor are transmitted from the detection unit. Acquired,
The signal state determination unit
As the determination process, a process of determining the signal state of the second detection signal,
The signal processing unit
When it is determined that the signal state of the second detection signal is the first signal state, the window function processing is performed using the first asymmetric window function as the asymmetric window function. To the detection signal,
When it is determined that the signal state of the second detection signal is the second signal state, the window function processing is performed using a second asymmetric window function different from the first asymmetric window function. A signal analyzing apparatus, wherein the signal analyzing apparatus performs the first detection signal. In any one of Claims 1 thru | or 7,
A signal state determination unit that performs a signal state determination process on the detection signal acquired by the signal acquisition unit;
The detector is
In addition to the first sensor that is the sensor, it further has a second sensor,
The signal acquisition unit
A first detection signal based on a first sensor signal acquired by the first sensor and a second detection signal based on a second sensor signal acquired by the second sensor are transmitted from the detection unit. Acquired,
The signal state determination unit
As the determination process, a process of determining the signal state of the second detection signal,
The signal processing unit
When it is determined that the signal state of the second detection signal is the first signal state, the window function for the first detection signal using the first asymmetric window function as the asymmetric window function Performing the processing and the Fourier transform processing on the data after the window function processing as the frequency decomposition processing,
When it is determined that the signal state of the second detection signal is the second signal state, a power spectrum density estimation process for the first detection signal is performed as the frequency resolution process. Signal analysis device. In claim 10 or 11,
The signal state determination unit
As the determination process, a process of determining the motion state of the subject based on the second detection signal is performed,
The signal processing unit
The signal analysis apparatus, wherein the frequency resolution processing is determined based on the determination result of the motion state. In any one of Claims 1 thru | or 7,
The detector is
In addition to the first sensor that is the sensor, it further has a second sensor,
The signal acquisition unit
A first detection signal based on a first sensor signal acquired by the first sensor and a second detection signal based on a second sensor signal acquired by the second sensor; Obtained from the detection unit as
The signal processing unit
A signal analyzing apparatus that performs signal processing on the first detection signal based on the second detection signal. In claim 13,
The signal processing unit
A signal analysis apparatus that performs noise reduction processing on the first detection signal as the signal processing based on the second detection signal.   An electronic apparatus comprising the signal analysis device according to claim 1. A signal acquisition unit for acquiring a detection signal from a detection unit having a sensor;
As a signal processing unit that performs window function processing on sampling data D _{0 to} D _N (N is a positive integer of 2 or more) of the detection signal and Fourier transform processing on the data after the window function processing as frequency decomposition processing,
Make the computer work,
Sampling data _{D j} of the j-th of the sampling data _D 0 to D _N, the
Data sampled at a timing later than the timing at which the _i- th sampling data D _i is sampled (integers i and j are integers satisfying 0 ≦ i <N / 2 <j ≦ N),
The signal processing unit
The window function value h (j) corresponding to the j-th sampling data D _j is the maximum value, and the window function value h (i) corresponding to the _i- th sampling data D _i is the window function value h (j). The window function processing is performed by an asymmetric window function having a smaller value.

Images