JP2016077539A - Bio-information expectation device and bio-information expectation method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a bio-information expectation device capable of expecting user's pulse number at high-accuracy without contact by image processing of user's skin color region included in obtained image data even in a case where image data photographed by the user has drop of frame.SOLUTION: A bio-information expectation device 10 comprises: an image input part 1 inputting image data of photographed human; an extracting part extracting signals within a predetermined range of the input image data by the image input part, a plurality of filter parts outputting, with different coefficients, each of signals corresponding to different coefficients in signals within the predetermined range extracted by the extracting part; a plurality of expectation parts which are arranged correspondingly to the filter parts respectively and which expects human pulse value based on output of corresponding filter part and input interval of frames of the image data input by the image input part; an output part selecting and outputting any of a plurality of pulse values expected by the plurality of expectation parts according to output of the plurality of filter parts.SELECTED DRAWING: Figure 3

Description

  The present invention relates to a biological information estimation apparatus and a biological information estimation method for estimating human biological information.

  Conventionally, vital sensing technology that estimates human biometric information is not limited to the fields of home medical care and health management, for example, drowsiness detection during driving, acquisition of a user's psychological state during a game, detection of abnormal persons in a monitoring system, etc. Application to a wide variety of fields is expected. Currently, a device that senses biological information is mainly used in contact with a human body and needs to be worn by the user, so that the application range is limited.

  For this reason, as one approach for non-contact sensing, a technique for estimating a pulse as an example of biological information from an image obtained by imaging with a camera has been proposed. By using this technology, it is possible to sense biological information without making the user aware of it, and the application range can be expected to be expanded. For example, it is possible to detect a suspicious person whose pulse fluctuation has increased due to stress while photographing with a surveillance camera. In addition, it is a great merit that it is possible to simultaneously sense a plurality of people in one camera. Compared to the contact type, it is not necessary to prepare an individual device for each user, and the troublesomeness of attaching to the body can be reduced.

  Here, as a prior art related to a pulse estimation technique using a camera, for example, a pulse measurement device disclosed in Patent Document 1 is known. The pulse measuring device calculates a feature amount of a captured input image, detects a pulse wave peak interval from the calculated feature amount, and calculates a pulse rate from the detected pulse wave peak interval. In addition, the pulse measurement device has an estimated maximum error between a pulse rate having an effective peak interval based on an adoption rate indicating a ratio of an effective peak interval and a true pulse rate within a calculated value. The frame rate indicating the number of frames captured per unit time is controlled so that

JP 2010-51592 A

  However, in Patent Document 1 described above, the frame rate of the camera is optimally secured, but when an image input from the camera drops (in other words, the input image is lost and cannot be input), Since the frame rate fluctuates, it becomes difficult to properly measure the user's pulse rate.

  In view of the above-described situation, the present invention reduces the user's pulse rate by image processing of the user's skin color area included in the acquired image data even when a frame drop of the frame of the image data captured by the user occurs. An object of the present invention is to provide a biological information estimation device and a biological information estimation method that perform non-contact and high-precision estimation.

  The present invention provides an image input unit that inputs image data obtained by capturing a human image, an extraction unit that extracts a signal within a predetermined range of the image data input by the image input unit, and the extraction using different coefficients. A plurality of filter units for outputting signals corresponding to the different coefficients among the signals in the predetermined range extracted by the unit, and provided for each of the filter units, for one cycle of the corresponding filter unit And a plurality of estimation units for estimating a pulse value of the person based on the input interval of the frame of the image data corresponding to the output of the one cycle, and the outputs of the plurality of filter units, And an output unit that selects and outputs any one of the plurality of pulse values estimated by the plurality of estimation units.

  Further, the present invention is a biological information estimation method in a biological information estimation device, and is different from a step of inputting image data obtained by capturing a human image and a step of extracting a signal within a predetermined range of the input image data. A step of outputting each signal corresponding to the different coefficient among the extracted signals in the predetermined range using a plurality of filter units having coefficients, an output for one cycle of each filter unit, and the one cycle A step of estimating the pulse value of the person based on the input interval of the frame of the image data corresponding to the output of the minute, and from the plurality of estimated pulse values according to the output of the plurality of filter units And a step of selecting and outputting either of them.

  According to the present invention, even when a frame drop of a frame of image data captured by the user occurs, the user's pulse rate is non-contact and highly accurate by image processing of the user's skin color area included in the acquired image data. Can be estimated.

(A) The figure which shows typically an example of the relationship between the contraction | shrinkage of a human heart, and the absorbed amount in the blood vessel of a light, (B) The figure which shows an example of a time-sequential change of the intensity | strength of light. The figure which shows an example of the absorption factor for every wavelength of light in hemoglobin (A) The block diagram which shows an example of an internal structure of the biometric information estimation apparatus of 1st Embodiment, (B) The figure which shows typically an example of the operation | movement outline | summary of the biometric information estimation apparatus of each embodiment. The figure which shows an example of the pulse calculation method using the conventional camera Explanatory drawing of the problem in the pulse calculation method shown in FIG. The figure which shows an example of the pulse calculation method in the biometric information estimation apparatus of 1st Embodiment. Explanatory drawing showing an operation example of the waveform verification unit in each estimation module unit The flowchart explaining an example of the operation | movement procedure of the biometric information estimation apparatus of 1st Embodiment. The block diagram which shows an example of an internal structure of the biometric information estimation apparatus of 2nd Embodiment. Explanatory drawing of the problem when the peak of the observed waveform for one period is off the sampling position Explanatory drawing which shows the 1st example of operation | movement of the interpolation part of each estimation module part. Explanatory drawing which shows the 2nd example of operation | movement of the interpolation part of each estimation module part. Explanatory drawing which shows the 3rd example of operation | movement of the interpolation part of each estimation module part. The flowchart explaining an example of the operation | movement procedure of the biometric information estimation apparatus of 2nd Embodiment.

  Hereinafter, embodiments that specifically disclose a biological information estimation apparatus and a biological information estimation method according to the present invention will be described with reference to the drawings. The biological information estimation apparatus according to each embodiment uses, as an example of biological information, image data obtained by capturing an object (for example, a person, the same applies hereinafter) without using a contact-type dedicated pulse rate measuring device, for example. Estimate a person's pulse rate without contact.

  More specifically, the biological information estimation apparatus according to each embodiment inputs a frame of image data obtained by capturing a person and extracts a signal (pixel value) within a predetermined range (for example, a skin color region) of the input image data. Then, each filter unit outputs each signal corresponding to a different coefficient among the extracted signals in the predetermined range using different filter coefficients. In addition, the biological information estimation device uses each filter unit to calculate a human pulse value based on an output signal for at least one cycle of each filter unit and an input interval of frames of image data corresponding to the output signal for at least one cycle. Is estimated by the estimation module unit corresponding to, and one of a plurality of pulse values estimated by the plurality of estimation module units is selected and output according to the output signal of each filter unit.

  The biological information estimation apparatus according to each embodiment is a data communication terminal such as a desktop or laptop PC (Personal Computer), a smartphone, a mobile phone, a tablet terminal, or a PDA (Personal Digital Assistant), for example. It may have a camera function for imaging a person.

  First, the estimation principle of the pulse rate in the biological information estimation apparatus of each embodiment will be described with reference to FIGS. 1 (A) and 1 (B). FIG. 1A is a diagram schematically illustrating an example of the relationship between the contraction of a human heart and the amount of light absorbed in a blood vessel. FIG. 1B is a diagram illustrating an example of time-series changes in light intensity.

  FIG. 1A shows that the volume of a blood vessel changes in synchronization with the systole of a human heart. When the volume of the blood vessel increases in accordance with the contraction of the heart, the amount of absorption of light (for example, light in a specific wavelength region shown in FIG. 2) increases, so that the light intensity (Light intensity) also decreases (FIG. 1B). reference). The pulse wave indicates a wave motion when the pressure change in the blood vessel generated when blood is pushed out to the aorta due to the contraction of the heart is transmitted in the peripheral direction.

  In FIG. 1B, the horizontal axis represents time, and the vertical axis represents the intensity of a signal (photoelectric pulse wave) obtained by a change in the amount of absorbed light. That is, in FIG. 1B, when the peak appears, the amount of light absorption is small, so the volume of the blood vessel is not increased, and when the minimum value appears, the amount of light absorption is large. Therefore, the volume of the blood vessel is increasing. It should be noted that although there is a slight delay due to the distance between the heart and the peripheral portion, a slight delay is seen, but the heart contraction and the change in the intensity of the photoelectric pulse wave basically fluctuate in synchronization.

  FIG. 2 is a diagram illustrating an example of an absorptance for each wavelength of light in hemoglobin. FIG. 2 shows that, for example, hemoglobin (blood) easily absorbs a wavelength of 400 nm (that is, green). In each of the following embodiments, the description will be made using the fact that the green light component has a high absorptance, but for example, the reflectance of the red light component (for example, a wavelength exceeding 1000 nm) is high. It may be explained using.

(First embodiment)
Hereinafter, the configuration of the biological information estimation apparatus of the first embodiment will be described with reference to FIG. FIG. 3A is a block diagram illustrating an example of an internal configuration of the biological information estimation apparatus 10 according to the first embodiment. FIG. 3B is a diagram schematically illustrating an example of an operation outline of the biological information estimation apparatuses 10 and 10A according to each embodiment.

  The biological information estimation apparatus 10 illustrated in FIG. 3A includes a camera CM, an image input unit 1, a skin color extraction unit 2, a plurality of (for example, three) filters F1, F2, and F3, and a plurality (for example, three). ) Estimation module units M1, M2, and M3, a reliability determination unit 5, a selector 6, and a pulse output unit 7. The estimation module unit M1 is provided corresponding to the filter F1, the estimation module unit M2 is provided corresponding to the filter F2, and the estimation module unit M3 is provided corresponding to the filter F3.

  The camera CM images a person who is an object at a predetermined frame rate (for example, 10 fps: frame per second) and outputs the image to the image input unit 1. Note that the camera CM may not be included in the biological information estimation apparatus 10, and may be connected to the biological information estimation apparatus 10 via a network, for example. The network is the Internet or an intranet connected with a wireless network or a wired network as an interface. The wireless network is, for example, a wireless LAN (Local Area Network), a wireless WAN (Wide Area Network), 3G, LTE (Long Term Evolution), or WiGig (Wireless Gigabit). The wired network is, for example, IEEE 802.3 or Ethernet (registered trademark) (ETHERNET (registered trademark)).

  The image input unit 1 continuously inputs (acquires) frames of image data captured by the camera CM at a predetermined frame rate from the camera CM and outputs the frames to the skin color extraction unit 2. When the camera CM and the biological information estimation device 10 are configured separately, the image input unit 1 continuously receives frames of image data transmitted from the camera CM.

  The skin color extraction unit 2 as an example of the extraction unit is a signal (for example, pixel value, luminance) of a predetermined range (for example, the skin color region FL1 shown in FIG. 3B) in each frame of the image data input by the image input unit 1. Value) is extracted using the number of frames for at least one period (the period is known). Here, the time-series signal extracted by the skin color extraction unit 2 is an original signal including a noise signal shown on the lower side of the human face shown in FIG. 3B, for example. The skin color extraction unit 2 outputs the extracted signal in the predetermined range to each of the filter units F1 to F3.

  The filter units F1, F2, and F3 are configured using band pass filters using different filter coefficients, and the operations are common except that the filter coefficients are different. Therefore, the operation of the filter unit F1 is exemplified. explain. The filter unit F1 averages the signal (pixel value) within a predetermined range output from the skin color extraction unit 2, thereby removing a noise signal included when the camera CM captures an image, for example.

  Although the human pulse wave can be extracted by this averaging, there is a high possibility that a human body motion component and a noise residual component are still included, so the filter unit F1 uses a filter coefficient corresponding to the filter unit F1. Then, the frequency components other than the fundamental frequency of the pulse wave are cut. The output of the filter unit F1 is input to the estimation module unit M1.

  The filter coefficient of the filter unit F1 is set in advance so that, for example, a signal of 30 to 60 bpm passes through the filter unit F1. The filter coefficient of the filter unit F2 is set in advance so that, for example, a signal of 50 to 90 bpm passes through the filter unit F2. The filter coefficient of the filter unit F3 is set in advance so that, for example, a signal of 70 to 120 bpm passes through the filter unit F3.

  In general, the pulse of an adult at rest is 60 to 80 bpm. However, when the subject has bradycardia or tachycardia, and the influence at the time of exercise or tension is taken into consideration, each filter unit F1 to F3 has a target range of 30 to 30 It is set to 120 bpm. Further, since the band of the output signal of each filter unit F1 to F3 is divided into three using the three filter units F1 to F3, the output signal of each filter unit F1 to F3 approaches a sine wave ( (See each waveform of the multiband filter shown in FIG. 3B), and the estimation accuracy of the pulse wave using the outputs of the filter units F1 to F3 can be improved.

  The estimation module unit M1 includes a waveform verification unit M1a and a pulse estimation unit M1b. Hereinafter, since the configuration of the estimation module units M1, M2, and M3 is the same, the operation of the estimation module unit M1 will be described as an example.

  The waveform verification unit M1a as an example of the verification unit receives at least one cycle of the output signal of the filter unit F1, and detects at least one input of the noise signal segment that could not be cut by the filter unit F1. It is determined whether or not there is a section of a signal (that is, a noise signal) that satisfies a predetermined condition (that is, Expression (1), Expression (2)) among the output signals for the period (see FIG. 7). The waveform verification unit M1a eliminates the interval of the output signal determined to satisfy the predetermined condition as an invalid interval. FIG. 7 is an explanatory diagram illustrating an operation example of the waveform verification unit in each of the estimation module units M1, M2, and M3.

  FIG. 7 shows, for example, time-series changes in the amplitude of the output signal for at least one cycle of the filter unit F1, and P1, P2, P3, P4, P5, P6, and P7 are, for example, waveform verification units M1a. Are sample values that have been AD-converted by an ADC (Analog Digital Converter) included. MPA (Mean peak to peak amplitude) indicates the average value of the amplitude between the peaks of the sample values P1 to P7. PPA (Peak to peak amplitude) indicates the maximum value of the amplitude between the peaks of the sample values P1 to P7.

  According to Equation (1), the waveform verification unit M1a can eliminate signals whose amplitude of the output signal of at least one cycle of the filter unit F1 is extremely larger than a predetermined value (for example, zero). Further, according to Equation (2), the waveform verification unit M1a can exclude a signal whose amplitude of the output signal for at least one cycle of the filter unit F1 is extremely smaller than a predetermined value (for example, zero). When the waveform verification unit M1a determines that there is a signal section that satisfies a predetermined condition (that is, Formula (1), Formula (2)), the waveform verification unit M1a outputs a signal for at least one cycle excluding the corresponding signal section. It outputs to the pulse estimation part M1b and the reliability determination part 5. FIG.

  On the other hand, when the waveform verification unit M1a determines that there is no section of the signal that satisfies the predetermined condition (that is, the formulas (1) and (2)), the output signal for at least one cycle of the filter unit F1 is used as it is. It outputs to the pulse estimation part M1b and the reliability determination part 5. FIG. The processing of the waveform verification unit M1a uses the knowledge that the amplitude of a person's pulse wave changes slowly within a certain range, and a signal that satisfies Equations (1) and (2) may be disturbance noise. The nature will be high. A similar output signal is also input to the reliability determination unit 5 from waveform verification units (not shown) in the other estimation module units F2 and F3.

  The pulse estimation unit M1b as an example of the estimation unit is based on the output signal for at least one cycle of the filter unit F1 or the output signal for at least one cycle of the waveform verification unit M1a according to Equation (3). (Pulse rate) is calculated and output to the selector 6 (see FIG. 6). FIG. 6 is a diagram illustrating an example of a pulse calculation method in the biological information estimation apparatus 10 according to the first embodiment.

  Here, the subject of the pulse calculation method using the conventional camera is demonstrated with reference to FIG.4 and FIG.5. FIG. 4 is a diagram illustrating an example of a pulse calculation method using a conventional camera. FIG. 5 is an explanatory diagram of problems in the pulse calculation method shown in FIG.

  In FIG. 4, the horizontal axis indicates time, and the vertical axis (not shown) indicates a signal WV0 (that is, pixel value) for one cycle of green in the skin color area of an image captured by the camera, for example, 8 bits. In this case, 0 to 255 is indicated. 4 and 5, for example, using the calculation result of the difference between the sample values (that is, the difference from the immediately preceding sample value), the zero at the next monotonic increase from the time when the zero point is passed through the monotonic increase. The signal up to the point passing point is extracted as a signal WV0 for one cycle.

  As an example of calculating the pulse rate shown in FIG. 4, the frame rate of the camera is 10 fps, the frame count corresponding to the signal WV0 for one cycle (that is, the number of frames of image data necessary to obtain the signal WV0 for one cycle). ) Is 9, the pulse rate is 10 × 60 ÷ 9≈67 bpm.

  In FIG. 5, as in FIG. 4, the horizontal axis represents time, and the vertical axis (not shown) represents the signal WV0 (that is, the pixel value) for one green period of the flesh-colored area of the image captured by the camera. For example, in the case of 8 bits, 0 to 255 are indicated. Here, when a frame drop of the frame of the image data passed from the camera (that is, reception loss or frame loss on the frame acquisition side) occurs, the frame count corresponding to the signal WV0 for one cycle becomes 8. An error occurs because a pulse rate different from the pulse rate that should be calculated is calculated. That is, when a frame drop occurs, the pulse rate becomes 10 × 60 ÷ 8 = 75 bpm, and a difference (8 bpm) from 67 bpm occurs as an error.

  Therefore, the biological information estimation apparatuses 10 and 10A according to the respective embodiments including the present embodiment can obtain the signal WV0 for one cycle without using the frame count corresponding to the signal WV0 for one cycle in the pulse estimation unit M1b. The pulse rate is calculated according to Equation (3) using the sum of the input intervals (for example, the input interval from the camera CM or the reception interval) of the frame of the image data input by the image input unit 1 so far.

In FIG. 6, as in FIG. 4, the horizontal axis indicates time, and the vertical axis (not shown) indicates a signal extracted by the skin color extraction unit 2. In other words, the signal shown in FIG. 6 indicates a signal WV0 (that is, a pixel value) for one cycle of green in the skin color area of the image captured by the camera CM. Indicated. The pulse estimator M1b has input intervals t ₀ , t ₁ , t ₂ , t ₃ , t ₄ , t ₅ , t ₆ , and t ₇ of image data frames necessary for obtaining a signal WV0 for one cycle. The sum (= 0.01 + 0.01 + 0.02 + 0.01 + 0.01 + 0.01 + 0.01 + 0.01) is used as the PWI (Pulse Wave Interval) shown in Equation (3) above.

Therefore, in the case where a frame drop of the third frame in FIG. 6 occurs, the number of frames of image data necessary to obtain the signal WV0 for one cycle is 8, but the pulse estimation unit M1b comprising an input interval t ₂ until entering the second frame by entering from the next frame (fourth frame shown in FIG. 6), necessary image for one cycle of signal WV0 obtain By using the sum of the data frame input intervals t ₀ , t ₁ , t ₂ , t ₃ , t ₄ , t ₅ , t ₆ , and t ₇ , as in the case where no frame drop occurs in FIG. An appropriate pulse rate can be calculated (see Formula (3)).

  The reliability determination unit 5 as an example of the determination unit includes each estimation module in which sections of the output signals of the filter units F1, F2, and F3 that satisfy a predetermined condition (Equation (2) and Equation (3)) are excluded. Output signals for at least one cycle from the units M1, M2, and M3, or output signals of the filter units F1, F2, and F3 that do not satisfy the predetermined condition (in other words, output signals of the filter units F1, F2, and F3) Are output from the respective estimation module units M1, M2, and M3 as they are, and an output signal for at least one cycle from the estimation module unit having the smallest invalid section is determined. The reliability determination unit 5 outputs a determination result (that is, information on any estimation module unit) to the selector 6.

  For example, the reliability determination unit 5 assigns reliability “100” when there is no invalid section as the reliability, and the number of invalid sections that satisfy Formula (1), the number of invalid sections that satisfy Formula (2), or In accordance with the number of invalid sections that satisfy both formula (1) and formula (2), a predetermined calculation method (for example, formula (1) and formula (2) based on which the number of invalid sections and invalid sections are determined). Reliability is given according to a method of multiplying by using a predetermined weight coefficient (known) for each and deducting points from “100”.

  Based on the determination result output from the reliability determination unit 5, the selector 6 as an example of the output unit outputs from each estimation module unit M1, M2, M3 (that is, each estimation module unit M1, M2, M2). Among the information regarding each pulse rate estimated in M3), the output of the estimation module unit (information regarding the pulse rate) corresponding to the determination result of the reliability determination unit 5 is selected and output to the pulse output unit 7.

  For example, as shown in FIG. 3B, the pulse rate estimated by the estimation module unit M1 is “70” and the reliability determined by the reliability determination unit 5 is “60”, and the estimation module unit M2 performs estimation. The reliability determined by the reliability determination unit 5 is “80” and the pulse rate estimated by the estimation module unit M3 is “78” and is determined by the reliability determination unit 5 It is assumed that the reliability is “70”. In this case, the selector 6 outputs the pulse rate “76” at which “80” having the highest reliability is obtained to the pulse output unit 7.

  The pulse output unit 7 as an example of the output unit outputs information on the pulse rate selected by the selector 6. For example, when the biological information estimation apparatus 10 has a display (not shown), the pulse output unit 7 displays information on the pulse rate selected by the selector 6 on the display. For example, when the biological information estimation apparatus 10 includes a speaker (not shown), the pulse output unit 7 outputs information on the pulse rate selected by the selector 6 from the speaker.

  Next, an example of an operation procedure of the biological information estimation apparatus 10 of the present embodiment will be described with reference to FIG. FIG. 8 is a flowchart illustrating an example of an operation procedure of the biological information estimation apparatus 10 according to the first embodiment.

  In FIG. 8, the image input unit 1 continuously inputs (acquires) frames of image data in which a person is captured by the camera CM at a predetermined frame rate and outputs the frames to the skin color extraction unit 2 (ST1). . The skin color extracting unit 2 receives at least 1 signal (for example, a pixel value and a luminance value) of a predetermined range (for example, a skin color region FL1 shown in FIG. 3B) in each frame of the image data input by the image input unit 1. Extraction is performed using the number of frames for a period (the period is known) (ST2). The skin color extraction unit 2 outputs the extracted signal in the predetermined range to each of the filter units F1 to F3.

  In step ST3, filter coefficients in the filter units F1 to F3 are selected (ST3). As described above, the filter coefficient is set in advance in the filter unit F1 so that, for example, a signal of 30 to 60 bpm passes through the filter unit F1. In the filter unit F2, for example, a filter coefficient is set in advance so that a signal of 50 to 90 bpm passes through the filter unit F2. Similarly, a filter coefficient is set in advance in the filter unit F3 so that, for example, a signal of 70 to 120 bpm passes through the filter unit F3. In FIG. 3, for ease of explanation, the processing from step ST <b> 3 to step ST <b> 7 is shown to be time series. However, in the biological information estimation apparatus 10, these processes are performed in parallel. It has been broken.

  As an example of the filter processing, the filter unit F1 averages the signals (pixel values) in a predetermined range output from the skin color extraction unit 2, thereby removing, for example, a noise signal included during imaging by the camera CM (ST4). ). Moreover, the filter part F1 cuts frequency components other than the fundamental frequency of a pulse wave using the filter coefficient corresponding to the filter part F1 as another example of a filter process (ST4). The output of the filter unit F1 is input to the estimation module unit M1.

  The waveform verification unit M1a, based on the output signal for at least one cycle of the filter unit F1, detects the section of the noise signal that could not be cut by the filter unit F1, and outputs the input output signal for at least one cycle. Of these, the waveform is verified by determining whether or not there is a signal section that satisfies a predetermined condition (that is, Formula (1), Formula (2)) (ST5).

  The waveform verification unit M1a excludes a section of the output signal that is determined to satisfy the predetermined condition as an invalid section, and at least one period of the signal that excludes the section of the corresponding signal is used as the pulse estimation unit M1b and the reliability determination. Output to unit 5. On the other hand, when the waveform verification unit M1a determines that there is no section of the signal that satisfies the predetermined condition (that is, the formulas (1) and (2)), the output signal for at least one cycle of the filter unit F1 is used as it is. It outputs to the pulse estimation part M1b and the reliability determination part 5. FIG.

  The pulse estimation unit M1b calculates a human pulse rate (Pulse rate) according to the equation (3) based on the output signal for at least one cycle of the filter unit F1 or the output signal for at least one cycle of the waveform verification unit M1a. Then, it is output to the selector 6 (ST6).

  In step ST7, when each process is not complete | finished in each filter part F1-F3 and each estimation module part M1-M3 (ST7, NO), each filter part F1-F3 and each estimation module part M1-M3 In step ST3, until the processes from step ST3 to step ST7 are completed, the operation of the biological information estimation apparatus 10 is in a standby state.

  On the other hand, when each process is complete | finished in each filter part F1-F3 and each estimation module part M1-M3 (ST7, YES), the reliability determination part 5 is a predetermined condition (Formula (2), Formula (3). )) Satisfying the output signal sections of the filter units F1, F2, and F3 are excluded, and output signals for at least one period from the estimation module units M1, M2, and M3, Based on the output signals of the filter units F1, F2, and F3, an output signal for at least one cycle from the estimation module unit having the smallest invalid section is determined (ST8). The reliability determination unit 5 outputs a determination result (that is, information on any estimation module unit) to the selector 6.

  Based on the determination result output from the reliability determination unit 5, the selector 6 outputs from each estimation module unit M1, M2, M3 (that is, each estimated in each estimation module unit M1, M2, M3). Among the information on the pulse rate), the output of the estimation module unit (information on the pulse rate) corresponding to the determination result of the reliability determination unit 5 is selected and output to the pulse output unit 7 (ST9). The pulse output unit 7 outputs information on the pulse rate selected by the selector 6 (ST9).

  As described above, the biological information estimation apparatus 10 according to the present embodiment extracts a signal of a predetermined range (for example, human skin color region FL1) of input image data, and extracts it using a different coefficient (for example, a filter coefficient). Each signal corresponding to a different coefficient among the signals in the predetermined range is output in the plurality of filter units F1 to F3. The biological information estimation apparatus 10 uses the estimation module units M1 to M3 provided corresponding to the filter units F1 to F3 to output an output signal for at least one cycle and an output signal for at least one cycle of the corresponding filter unit. Based on the input interval of the frame of the corresponding image data, the person's pulse value is estimated, and one of the estimated pulse values is selected according to the output signals of the plurality of filter units F1 to F3. Output.

  Thereby, the biometric information estimation device 10 performs image processing (for example, noise removal processing using a predetermined filter coefficient) of the user's skin color area included in the image data even when the frame of the image data captured by the user drops. The pulse rate of the user is estimated using the input intervals of a plurality of frames of image data corresponding to the signal of at least one cycle obtained by (1) and the property that blood absorbs light in a specific wavelength range. The user's pulse rate can be estimated with high accuracy in a non-contact manner without using a dedicated pulse rate measuring device of the type.

  In addition, the biological information estimation apparatus 10 excludes an output signal section of a filter unit that satisfies a predetermined condition as an invalid section in an estimation module unit corresponding to each filter unit, and an output signal of at least one cycle from which the invalid section is excluded And a pulse value of the person is estimated based on the input interval of the frame of the image data corresponding to the output signal of at least one cycle.

  Thereby, the biological information estimation apparatus 10 can exclude the section of the output signal of the filter unit that satisfies the predetermined condition (for example, the section of the signal whose amplitude is extremely large or extremely small compared to the predetermined value) as the invalid section. Therefore, for example, the influence of disturbance noise can be suppressed, and the pulse value of a person can be obtained with high accuracy using the input interval of the frame of image data corresponding to the output signal for at least one period from which the invalid interval is excluded. Can be estimated.

  In addition, the biological information estimation apparatus 10 outputs the output signal for at least one cycle with the least number of invalid intervals based on the output signal for at least one cycle in which the section of the output signal of each filter unit that satisfies the predetermined condition is excluded. A determination is made, and the estimated pulse value of the person corresponding to the determined output signal for at least one period is selected.

  Thereby, the biometric information estimation apparatus 10 can select a person's pulse value using an output signal for at least one cycle having the smallest invalid section (in other words, the least affected by disturbance noise).

(Second Embodiment)
FIG. 9 is a block diagram illustrating an example of an internal configuration of the biological information estimation apparatus 10A according to the second embodiment. FIG. 10 is an explanatory diagram of a problem when the peak of the observed waveform for one period is off the sampling position. In the biometric information estimation apparatus 10A shown in FIG. 9, the same components as those in the biometric information estimation apparatus 10 shown in FIG. 3A are denoted by the same reference numerals, and description thereof will be simplified or omitted, and different contents will be described. .

  The biological information estimation apparatus 10A illustrated in FIG. 9 includes a camera CM, an image input unit 1, a skin color extraction unit 2, a plurality (for example, three) of filters F1, F2, and F3, and a plurality (for example, three) of estimation. The module unit includes M1R, M2R, and M3R, a reliability determination unit 5, a selector 6, and a pulse output unit 7. The estimation module unit M1R is provided corresponding to the filter F1, the estimation module unit M2R is provided corresponding to the filter F2, and the estimation module unit M3R is provided corresponding to the filter F3.

  The estimation module unit M1R includes an interpolation unit M1c, a waveform verification unit M1a, and a pulse estimation unit M1b. The interpolation unit M1c receives at least one cycle of the output signal of the filter unit F1, and based on the output signal of at least one cycle of the filter unit F1, the output signal of at least one cycle of the filter unit F1 has a predetermined value ( For example, the difference (time difference) of the position (time) that is zero) is interpolated (for example, linear interpolation), and output signals for at least one cycle after the interpolation are output to the waveform verification unit M1a and the pulse estimation unit M1b. In the present embodiment, the waveform verification unit M1a and the pulse estimation unit M1b use the output of the interpolation unit M1c. Hereinafter, since the configuration of the estimation module units M1R, M2R, and M3R is the same, the operation of the estimation module unit M1R will be described as an example.

  Here, with reference to FIG. 10, the problem when the peak of the observed waveform for one period is off the sampling position will be described. In FIG. 10, the horizontal axis indicates time, and the vertical axis (not shown) indicates the signal WV0 extracted by the skin color extraction unit 2. In other words, the signal shown in FIG. 10 indicates a signal WV0 (that is, a pixel value) for one cycle of green in the flesh-color area FL1 of an image captured by the camera CM. Indicated by

  In FIG. 10, S1, S2, S3, S4, S5, S6, and S7 are sample values AD-converted by the ADC included in the waveform test unit M1a, for example, and T1, T2, T3, T4, T5, T6 T7 is the sampling time AD-converted by the ADC. To be precise, one cycle of the signal WV0 shown in FIG. 10 is the time between the peak of the signal WV0 and the next peak (see the straight line AS1), but the peak of the signal WV0 is the sampling position (sample at the sampling time). If the value is not within the range, an error will occur. Specifically, one cycle of the signal WV0 shown in FIG. 10 is the time between the sampling time T2 of the sample value S2 that is the peak value and the sampling time T7 of the sample value S7 that is the next peak value (see the straight line NT1). Therefore, an error corresponding to the difference between the straight line AS1 and the straight line NT1 occurs, and an error occurs in the signal input to the filter units F1 to F3. Will deteriorate.

  Therefore, in the present embodiment, each of the estimation module units M1R to MR3 is input to each of the estimation module units M1R to MR3 in order to more accurately obtain the PWI (pulse wave interval) shown in Equation (3). The signals WV0 for one cycle are derived by interpolating the received signals (that is, the output signals of the filter units F1 to F3) (see FIGS. 11, 12, and 13).

(First interpolation example)
FIG. 11 is an explanatory diagram illustrating a first example of the operation of the interpolation unit M1c of each estimation module unit. In FIG. 11, the horizontal axis represents time, and the vertical axis (not shown) represents the signal WV1 extracted by the skin color extraction unit 2. In other words, the signal shown in FIG. 11 indicates a signal WV1 (that is, a pixel value) for one green period of the flesh-color area FL1 of an image captured by the camera CM. Indicated by

  In the first interpolation example shown in FIG. 11, the interpolation unit M1c, for example, at least two positions (time) on the horizontal axis where the signal WV1 when monotonically increasing or monotonically decreasing passes a predetermined value (for example, zero). The difference (time difference) is interpolated (that is, linear interpolation processing), and the signal WV1 for at least one period is estimated. Since the straight line AS1 and the straight line NT1 shown in FIG. 11 are the same as the straight line AS1 and the straight line NT1 shown in FIG. 10, the description thereof is omitted, and the same applies to FIG. 12 and FIG.

  More specifically, the interpolation unit M1c estimates, for example, a difference (see a straight line PR1) between a position where the signal WV1 when monotonically increasing passes zero and a position where the signal WV1 when monotonically increasing next passes zero. Therefore, a position where the straight line passing through at least two positions (point A0 (0, a0) and point A1 (1, a1)) of the first monotonically increasing signal WV1 passes zero (straight line PR1 (The left end) and a position where a straight line passing through at least two positions (point B0 (0, b0) and point B1 (1, b1)) of the second monotonically increasing signal WV1 passes zero (straight line PR1) To the right). a0, a1, b0 and b1 are positions on the signal WV1 and sample values. In FIG. 11, description is made using points on the signal WV1 at the time of monotonic increase, but the same applies to points on the signal WV1 at the time of monotonic decrease.

  The distance p0 between the position on the horizontal axis of the point A0 and the left end of the straight line PR1 is expressed by the equation (4) by linear interpolation of the interpolation unit M1c. Similarly, the distance p1 between the position on the horizontal axis of the point B0 and the right end of the straight line PR1 is expressed by Equation (5) by linear interpolation of the interpolation unit M1c. As a result, the interpolation unit M1c can calculate the length of the straight line PR1, and can reduce the error with the straight line AS1 compared to the straight line NT1 with a large error with the straight line AS1, so the estimation of the pulse rate. Accuracy can be improved.

(Second interpolation example)
FIG. 12 is an explanatory diagram illustrating a second example of the operation of the interpolation unit M1c of each estimation module unit. In FIG. 12, the horizontal axis indicates time, and the vertical axis (not shown) indicates the signal WV1 extracted by the skin color extraction unit 2. In other words, the signal shown in FIG. 11 indicates a signal WV1 (that is, a pixel value) for one green period of the flesh-color area FL1 of an image captured by the camera CM. Indicated by

  In the second interpolation example shown in FIG. 12, the interpolation unit M1c, for example, inverts the straight line passing through two points on the signal WV1 at the time of monotonic increase or monotonic decrease where the first peak is obtained, and the slope of the straight line. The intersection of the slope and the straight line passing through one point on the signal WV1 when monotonously decreasing or monotonically increasing is obtained. Further, the interpolation unit M1c similarly applies a straight line passing through two points on the signal WV1 at the time of monotonic increase or decrease when the next peak is obtained, a slope obtained by inverting the slope of the straight line, and a monotonic decrease or monotonic. An intersection point with a straight line passing through one point on the signal WV1 at the time of increase is obtained. The interpolation unit M1c performs interpolation (that is, equiangular straight line fitting) using these two intersections, and estimates the signal WV1 for at least one period.

  More specifically, the interpolation unit M1c, for example, includes a straight line L0 passing through two points (point A0 (0, a0) and point A1 (1, a1)) on the signal WV1 when monotonically increasing, and the straight line L0. The intersection of the slope obtained by inverting the slope and the straight line L1 passing through one point (point A2 (2, a2)) on the signal WV1 when monotonously decreasing is obtained. Further, similarly, the interpolation unit M1c, for example, a straight line M0 passing through two points (point B0 (0, b0) and point B1 (1, b1)) on the signal WV1 when monotonously increasing, and the straight line M0 The intersection of the slope obtained by inverting the slope and the straight line M1 passing through one point (point B2 (2, b2)) on the signal WV1 when monotonously decreasing is obtained. a0, a1, a2, b0, b1, b2 are positions on the signal WV1 and sample values. In FIG. 12, the straight lines L0 and MO are explained using points on the signal WV1 when monotonically increasing, and the straight lines L1 and M1 are monotonically decreasing. But the same is true.

  The distance p0 between the position on the horizontal axis of the point A1 and the position on the horizontal axis of the intersection of the straight line L0 and the straight line L1 is expressed by the equation (6) by the equiangular straight line fitting of the interpolation unit M1c. Similarly, the distance p1 between the position on the horizontal axis of the intersection of the straight line M0 and the straight line M1 and the position on the horizontal axis of the point B1 is expressed by the equation (7) by the equiangular straight line fitting of the interpolation unit M1c. It is. Thereby, the interpolation unit M1c can calculate the length of the straight line PR2, and can reduce the error from the straight line AS1 compared to the straight line NT1 having a large error from the straight line AS1, so the estimation of the pulse rate. Accuracy can be improved.

(Third interpolation example)
FIG. 13 is an explanatory diagram illustrating a third example of the operation of the interpolation unit M1c of each estimation module unit. In FIG. 13, the horizontal axis indicates time, and the vertical axis (not shown) indicates the signal WV1 extracted by the skin color extraction unit 2. In other words, the signal shown in FIG. 13 indicates a signal WV1 (that is, a pixel value) for one green period of the flesh-color area FL1 of an image captured by the camera CM. Indicated by

  In the third interpolation example shown in FIG. 13, the interpolation unit M1c includes a vertex of a quadratic curve passing through three points on the signal WV1 from which the first peak is obtained and three points on the signal WV1 from which the next peak is obtained. Is interpolated (ie, parabolic fitting) using a vertex of a quadratic curve passing through, and the signal WV1 for at least one period is estimated.

  More specifically, for example, the interpolation unit M1c uses three points (point A0 (0, a0), point A1 (1, a1), and point A2 (2, a2)) on the signal WV1 from which the first peak is obtained. The position of the vertex of the passing quadratic curve R0 on the horizontal axis and the three points (point B0 (0, b0), point B1 (1, b1) and point B2 (2, b2)) and the position of the vertex of the quadratic curve R1 on the horizontal axis. a0, a1, a2, b0, b1, b2 are positions on the signal WV1 and sample values.

  A distance p0 between the position on the horizontal axis of the point A1 and the position on the horizontal axis of the vertex of the quadratic curve R0 is expressed by Equation (8) by parabolic fitting of the interpolation unit M1c. Similarly, the distance p1 between the position of the vertex of the quadratic curve R1 on the horizontal axis and the position of the point B1 on the horizontal axis is expressed by Equation (9) by equiangular straight line fitting by the interpolation unit M1c. As a result, the interpolation unit M1c can calculate the length of the straight line PR3, and can reduce the error from the straight line AS1 compared to the straight line NT1 having a large error from the straight line AS1. Accuracy can be improved.

  FIG. 14 is a flowchart illustrating an example of an operation procedure of the biological information estimation apparatus 10A according to the second embodiment. In FIG. 14, the content different from the operation procedure of the biological information estimation apparatus 10 of the first embodiment will be described, and the same operation will be given the same step number and the description will be simplified or omitted.

  In FIG. 14, after step ST4, the interpolation unit M1c receives at least one cycle of the output signal of the filter unit F1, and based on the output signal of at least one cycle of the filter unit F1, at least one of the filter units F1. Interpolate (eg, linear interpolation) the difference (time difference) between the positions (time) at which the output signal for the period becomes a predetermined value (for example, zero), and use the waveform verification unit M1a and the pulse estimation for the output signal for at least one period after the interpolation. The data is output to the part M1b (ST10). The waveform verification unit M1a and the pulse estimation unit M1b use the output of the interpolation unit M1c.

  For example, the waveform verification unit M1a receives at least one cycle of the output signal from the interpolation unit M1c, and detects at least one cycle of the input signal in order to detect a section of the noise signal that could not be cut by the filter unit F1. It is determined whether or not there is a section of a signal (that is, a noise signal) that satisfies a predetermined condition (that is, Expressions (1) and (2)) among the output signals (ST5). In addition, the pulse estimation unit M1b is based on the pulse rate (Pulse rate) of the person according to Equation (3) based on the output signal for at least one cycle from the interpolation unit M1c or the output signal for at least one cycle of the waveform verification unit M1a. ) And output to the selector 6 (ST6). The details of the processing in step ST5 and step ST6 are the same as those in the first embodiment, and thus description thereof is omitted. Since the process after step ST6 is the same as that of FIG. 8, description is abbreviate | omitted.

  As described above, the biological information estimation apparatus 10A according to the present embodiment is based on the output signals for at least one cycle of the filter units F1, F2, and F3 corresponding to the estimation module units M1R, M2R, and M3R. The time difference at which the output signal for at least one cycle of F2 and F3 becomes a predetermined value (for example, zero) is interpolated, and the output signal for at least one cycle of the filter unit after interpolation is output.

  Accordingly, the biological information estimation apparatus 10A can obtain the output signal for at least one cycle of the filter units F1, F2, and F3 with high accuracy, and thus the peak of the output signal for one cycle is off the sampling position. Even in this case, it is possible to minimize the error between the time difference between the peaks of the accurate output signal that should originally be obtained and the time difference between the peaks of the output signal of the filter section that is actually obtained. The value can be estimated.

  Finally, configurations, actions, and effects of the biological information estimation apparatus and biological information estimation method according to the present invention will be listed respectively.

  One embodiment of the present invention uses different coefficients for an image input unit that inputs image data obtained by imaging a person, and an extraction unit that extracts a signal within a predetermined range of the image data input by the image input unit. A plurality of filter units for outputting signals corresponding to the different coefficients among the signals in the predetermined range extracted by the extraction unit, and the corresponding filter units provided corresponding to the filter units. A plurality of estimation units for estimating a pulse value of the person based on an output signal for at least one period and an input interval of frames of the image data corresponding to the output signal for at least one period; And an output unit that selects and outputs one of the plurality of pulse values estimated by the plurality of estimation units according to an output signal of the filter unit.

  In this configuration, the biological information estimation apparatus extracts a signal in a predetermined range (for example, a human skin color region) of input image data, and uses a different coefficient (for example, a filter coefficient) to extract the signal in the extracted predetermined range. Each signal corresponding to a different coefficient is output from a plurality of filter units. The biological information estimation apparatus is configured to input an output signal for at least one cycle of the corresponding filter unit and an image data frame corresponding to the output signal for at least one cycle in each estimation unit provided corresponding to each filter unit. Based on the interval, the person's pulse value is estimated, and one of the estimated plurality of pulse values is selected and output according to the output signals of the plurality of filter units.

  Thereby, the biometric information estimation device performs image processing (for example, noise removal processing using a predetermined filter coefficient) of the user's skin color area included in the image data even when a frame of the image data captured by the user drops. Since the pulse rate of the user is estimated using the input interval of a plurality of frames of image data corresponding to the signal of at least one cycle obtained by the above and the property that blood absorbs light in a specific wavelength range, for example, contact type The user's pulse rate can be estimated with high accuracy in a non-contact manner without using a dedicated pulse rate measuring device.

  In one embodiment of the present invention, the estimation unit defines an interval of the output signal of the filter that satisfies a predetermined condition among output signals for at least one cycle of the filter unit corresponding to the estimation unit as an invalid interval. A test unit to be excluded, an output signal for at least one cycle of the test unit from which the invalid interval is excluded, and an input interval of the frame of the image data corresponding to the output signal for at least one cycle of the test unit. And a pulse estimation unit that estimates the pulse value of the person.

  In this configuration, the biological information estimation apparatus excludes the output signal section of the filter unit that satisfies the predetermined condition as an invalid section in the estimation unit corresponding to each filter unit, and outputs an output signal of at least one cycle from which the invalid section is excluded. And a pulse value of the person is estimated based on the input interval of the frame of the image data corresponding to the output signal of at least one cycle.

  Thereby, the biological information estimation apparatus can exclude the section of the output signal of the filter unit satisfying the predetermined condition (for example, the section of the signal whose amplitude is extremely large or extremely small compared to the predetermined value) as the invalid section. Therefore, for example, the influence of disturbance noise can be suppressed, and the pulse value of a person is estimated with high accuracy using the input interval of the frame of the image data corresponding to the output signal for at least one period from which the invalid interval is excluded. can do.

  In addition, according to an embodiment of the present invention, based on an output signal for at least one cycle of each of the test units from which the section of the output signal of each filter that satisfies the predetermined condition is excluded, A determination unit that determines an output signal for at least one cycle of the test unit, wherein the output unit corresponds to the output signal for at least one cycle of the test unit with the least number of invalid intervals. It is a biological information estimation apparatus which selects the said person's pulse value estimated by the estimation part.

  In this configuration, the biological information estimating apparatus outputs at least one cycle of output signals having the least number of invalid sections based on the output signal of at least one cycle in which the section of the output signal of each filter unit that satisfies the predetermined condition is excluded. And the estimated pulse value of the person corresponding to the determined output signal for at least one period is selected.

  Thereby, the biometric information estimation apparatus can select a person's pulse value using an output signal for at least one period with the least number of invalid sections (in other words, the least affected by disturbance noise).

  Further, according to an embodiment of the present invention, the estimation unit has an output signal for at least one cycle of the filter unit based on an output signal for at least one cycle of the filter unit corresponding to the estimation unit. The biological information estimation apparatus further includes an interpolation unit that interpolates the time difference to be output and outputs an output signal for at least one period of the interpolated filter unit.

  In this configuration, the biological information estimation apparatus has a predetermined value (for example, zero) of the output signal for at least one cycle of the filter unit based on the output signal for at least one cycle of each filter unit corresponding to each estimation unit. The time difference is interpolated, and an output signal for at least one cycle of the interpolated filter unit is output.

  Thereby, since the biological information estimation apparatus can obtain the output signal for at least one cycle of the filter unit with high accuracy, it can be obtained even when the peak of the output signal for one cycle deviates from the sampling position. Since it is possible to minimize the error between the time difference between the peaks of the exact output signal and the time difference between the peaks of the output signal of the filter section actually obtained, it is possible to estimate a more accurate human pulse value. it can.

  One embodiment of the present invention is a biometric information estimation method in a biometric information estimation device, the step of inputting image data obtained by imaging a person, and extracting a signal in a predetermined range of the input image data. Using a plurality of filter units having different coefficients, outputting each signal corresponding to the different coefficients among the extracted signals in the predetermined range, and at least one cycle of each of the filter units Based on the output signal and the input interval of the frame of the image data corresponding to the output signal for at least one period, estimating the pulse value of the person, and according to the output signals of the plurality of filter units, Selecting and outputting any one of the plurality of estimated pulse values.

  In this method, the biological information estimation apparatus extracts a signal in a predetermined range (for example, human skin color region) of input image data, and uses a different coefficient (for example, a filter coefficient) to extract the signal in the extracted predetermined range. Each signal corresponding to a different coefficient is output from a plurality of filter units. The biological information estimation apparatus is configured to input an output signal for at least one cycle of the corresponding filter unit and an image data frame corresponding to the output signal for at least one cycle in each estimation unit provided corresponding to each filter unit. Based on the interval, the person's pulse value is estimated, and one of the estimated plurality of pulse values is selected and output according to the output signals of the plurality of filter units.

  Thereby, the biometric information estimation device performs image processing (for example, noise removal processing using a predetermined filter coefficient) of the user's skin color area included in the image data even when a frame of the image data captured by the user drops. Since the pulse rate of the user is estimated using the input interval of a plurality of frames of image data corresponding to the signal of at least one cycle obtained by the above and the property that blood absorbs light in a specific wavelength range, for example, contact type The user's pulse rate can be estimated with high accuracy in a non-contact manner without using a dedicated pulse rate measuring device.

  As mentioned above, although various embodiment was described with reference to drawings, it cannot be overemphasized that this indication is not limited to this example. It will be apparent to those skilled in the art that various changes and modifications can be made within the scope of the claims, and these are naturally within the technical scope of the present disclosure. Understood.

  The present invention estimates the user's pulse rate with high accuracy in a non-contact manner by performing image processing of the user's skin color area included in the acquired image data even when frames of the image data captured by the user are dropped. It is useful as a biological information estimation device and a biological information estimation method.

DESCRIPTION OF SYMBOLS 1 Image input part 2 Skin color extraction part 5 Reliability determination part 6 Selector 7 Pulse output part 10, 10A Biometric information estimation apparatus CM Camera F1, F2, F3 Filter FL1 Skin color area | region M1, M2, M3, M1R, M2R, M3R estimation module M1a Waveform verification unit M1b Pulse estimation unit M1c Interpolation unit

Claims (5)

An image input unit for inputting image data of a person imaged;
An extraction unit for extracting a signal within a predetermined range of the image data input by the image input unit;
A plurality of filter units that output signals corresponding to the different coefficients among the signals in the predetermined range extracted by the extraction unit using different coefficients;
Based on an output signal for at least one cycle of the corresponding filter unit and an input interval of frames of the image data corresponding to the output signal for at least one cycle, provided corresponding to each filter unit, A plurality of estimation units for estimating the person's pulse value;
An output unit that selects and outputs one of the plurality of pulse values estimated by the plurality of estimation units according to output signals of the plurality of filter units;
Biological information estimation device. The biological information estimation apparatus according to claim 1,
The estimation unit is configured to exclude, as an invalid interval, an interval of the output signal of the filter unit that satisfies a predetermined condition among the output signals of at least one cycle of the filter unit corresponding to the estimation unit;
Based on the output signal of at least one cycle of the verification unit from which the invalid interval has been excluded and the input interval of the frame of the image data corresponding to the output signal of at least one cycle of the verification unit, the pulse of the person A pulse estimator for estimating a value,
Biological information estimation device. The biological information estimation apparatus according to claim 2,
Based on the output signal of at least one cycle of each of the test units from which the section of the output signal of each filter unit that satisfies the predetermined condition is excluded, at least one cycle of the test unit having the least number of invalid sections A determination unit for determining the output signal of
The output unit selects the pulse value of the person estimated by the pulse estimation unit corresponding to an output signal for at least one cycle of the test unit with the least number of invalid intervals.
Biological information estimation device. The biological information estimation apparatus according to claim 1,
The estimation unit interpolates a time difference at which an output signal of at least one cycle of the filter unit becomes a predetermined value based on an output signal of at least one cycle of the filter unit corresponding to the estimation unit, and after the interpolation An interpolation unit that outputs an output signal for at least one cycle of the filter unit;
Biological information estimation device. A biological information estimation method in a biological information estimation device,
Inputting image data of a person imaged;
Extracting a signal within a predetermined range of the input image data;
Outputting each signal corresponding to the different coefficient among the extracted signals in the predetermined range using a plurality of filter units having different coefficients;
Estimating a pulse value of the person based on an output signal of at least one period of each filter unit and an input interval of frames of the image data corresponding to the output signal of at least one period;
Selecting and outputting one of the plurality of estimated pulse values according to the output signals of the plurality of filter units, and
Biological information estimation method.

Images