KR102627743B1 - METHOD AND APPARATUS FOR ESTIMATING AGE USING photoplethysmography - Google Patents

Abstract

A method and device for age estimation using photoplethysmography (PPG) are provided. According to one embodiment, the method measures PPG for the target area of the estimated subject to obtain a PPG signal, and performs differentiation on the PPG signal to obtain a second derivative of PPG (SDPPG) signal and 3 Obtaining a third derivative of PPG (TDPPG) signal, determining a difference variable based on the second peak values of the SDPPG signal, and determining a cumulative variable based on the third peak values of the TDPPG signal, It may include steps of executing an age estimation model based on the difference variable and the accumulation variable to obtain a model output, and estimating the age of the estimated subject based on the model output.

Description

Age estimation method and device using photoplethysmography {METHOD AND APPARATUS FOR ESTIMATING AGE USING photoplethysmography}

The examples below relate to a method and device for age estimation using photoplethysmography.

As the aging of society progresses rapidly, the need for health care for the elderly is being emphasized. In particular, ICT-based healthcare technology that measures biosignals for cardiovascular and musculoskeletal functions, which are evaluated as vital signs, has become advanced, and daily life monitoring technology can record non-invasive wearable devices or life logs. are being commercialized. Among various daily life monitoring technologies, a representative technology is photoplethysmography (PPG), which represents cardiovascular function. A PPG pulse can represent the absorption of infrared radiation across a body part (eg, finger, earlobe) with a pulsatile component and a constant component. It can represent cardiovascular flow, volume, blood vessel wall movement, and arrangement of arteries and the heart. This cardiovascular function can be used for indirect evaluation of aging, high blood pressure, diabetes, etc. Prior art includes Patent Publication No. 10-2008-0030189.

According to one embodiment, the age estimation method includes the steps of measuring photoplethysmography (PPG) on a target area of an estimated subject to obtain a PPG signal; performing differentiation on the PPG signal to obtain a second derivative PPG (second derivative of PPG, SDPPG) signal and a third derivative PPG (third derivative of PPG, TDPPG) signal; determining a difference variable based on secondary peak values of the SDPPG signal; determining a cumulative variable based on third peak values of the TDPPG signal; executing an age estimation model based on the difference variable and the cumulative variable to obtain model output; and estimating the age of the estimated subject based on the model output.

The secondary peak values of the SDPPG signal are a first secondary peak value and a second secondary peak value sequentially corresponding to inflection points of the PPG signal in one cycle including one systole and one diastole of the PPG signal. It may include a peak value, a third secondary peak value, a fourth secondary peak value, and a fifth secondary peak value, and the difference variable is the second secondary peak value to the third secondary peak value, It may correspond to a result of dividing the remainder after subtracting the fourth secondary peak value and the fifth secondary peak value by the first secondary peak value.

The 3rd peak values of the TDPPG signal are a 1st 3rd peak value, a 2nd 3rd peak value, and a 3rd 3rd peak value that sequentially correspond to inflection points of the SDPPG signal in the one cycle of the PPG signal. , may include a fourth tertiary peak value, and a fifth tertiary peak value, and the cumulative variable is the first tertiary peak value, the second tertiary peak value, the third tertiary peak value, and the It may correspond to the root mean square (rms) of the TDPPG signal according to the fourth third peak value and the fifth third peak value.

The estimated subject may be a woman over 65 years of age, and the target area may be an earlobe.

Obtaining the PPG signal may include measuring the PPG for a predetermined period of time through a wearable device fixed to the target area of the estimated subject.

The age estimation model may be derived through a neural network model pre-trained based on input samples including a training set, validation set, and test set, and the neural network model may include a single hidden layer including a plurality of neurons, and It may include an output layer containing two layers of a feed forward neural network (FFNN).

The model output can be obtained through the following equation.

In the above equation, Age is the model output, AI is the difference variable, and may represent the cumulative variable.

According to one embodiment, the age estimation device includes a processor; and a memory including instructions executable by the processor, and when the instructions are executed by the processor, the processor measures photoplethysmography (PPG) on the target area of the estimated subject to obtain a PPG signal, , perform differentiation on the PPG signal to obtain a second derivative PPG (second derivative of PPG, SDPPG) signal and a third derivative PPG (third derivative of PPG, TDPPG) signal, and add to the second peak values of the SDPPG signal determine a difference variable based on the third peak values of the TDPPG signal, determine a cumulative variable based on the third peak values of the TDPPG signal, run an age estimation model based on the difference variable and the cumulative variable to obtain a model output, and obtain a model output. Based on this, the age of the subject is estimated.

Figure 1 schematically shows an age estimation process according to one embodiment.
Figure 2 shows PPG, SDPPG, and TDPPG signals according to one embodiment.
Figure 3 shows a neural network model according to one embodiment.
Figure 4 shows linear fitting results for candidate variables according to one embodiment.
Figure 5 shows estimation results for a test set of selection variables according to one embodiment.
Figure 6 shows mse of each epoch by training, verification, and test data according to one embodiment.
Figure 7 shows training, validation, testing, regression results with all data, error histogram, and Brandt-Altman plot according to one embodiment.
Figure 8 shows an age estimation device and a measurement device according to an embodiment.
9 shows an electronic device according to one embodiment.

Specific structural or functional descriptions of the embodiments are disclosed for illustrative purposes only and may be changed and implemented in various forms. Accordingly, the actual implementation form is not limited to the specific disclosed embodiments, and the scope of the present specification includes changes, equivalents, or substitutes included in the technical idea described in the embodiments.

Terms such as first or second may be used to describe various components, but these terms should be interpreted only for the purpose of distinguishing one component from another component. For example, a first component may be named a second component, and similarly, the second component may also be named a first component.

When a component is referred to as being “connected” to another component, it should be understood that it may be directly connected or connected to the other component, but that other components may exist in between.

Singular expressions include plural expressions unless the context clearly dictates otherwise. In this specification, terms such as “comprise” or “have” are intended to designate the presence of the described features, numbers, steps, operations, components, parts, or combinations thereof, and are intended to indicate the presence of one or more other features or numbers, It should be understood that this does not exclude in advance the possibility of the presence or addition of steps, operations, components, parts, or combinations thereof.

Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by a person of ordinary skill in the art. Terms as defined in commonly used dictionaries should be interpreted as having meanings consistent with the meanings they have in the context of the related technology, and unless clearly defined in this specification, should not be interpreted in an idealized or overly formal sense. No.

Hereinafter, embodiments will be described in detail with reference to the attached drawings. In the description with reference to the accompanying drawings, identical components will be assigned the same reference numerals regardless of the reference numerals, and overlapping descriptions thereof will be omitted.

Figure 1 schematically shows an age estimation process according to one embodiment.

The age estimation device 100 may obtain a PPG signal 111 by measuring photoplethysmography (PPG) on a target area of the estimated subject 101. For example, PPG can be measured through a wearable device. The PPG signal 111 may correspond to a pulse wave. PPG pulse can be used as pulse variability, such as heart rate variability to evaluate the sympathetic and autonomic nervous systems, and can obtain various indicators related to the cardiovascular system through its own morphology analysis.

Since recognition of the inflection point is not as clear as in an ECG record, differential values such as SDPPG or TDPPG can be used to enable simpler and more objective evaluation. The age estimation device 100 performs differentiation on the PPG signal to obtain a second derivative PPG (second derivative of PPG, SDPPG) signal 112 and a third derivative PPG (third derivative of PPG, TDPPG) signal 113. can do. The SDPPG signal 112 and TDPPG signal 113 can provide more detailed information about the inflection point of the pulse wave. SDPPG can also be used as an indicator of arterial stiffness.

Generally, among the five inflection points of up and down waves in SDPPG measured at the finger, the first rising part can mean an increase until early systolic, the second is a decrease, the third is a re-increase after contraction, and the fourth is an increase after contraction. Re-reduction, the last may mean early diastolic. In other words, cardiovascular characteristics based on contraction and relaxation, which are difficult to accurately confirm from the PPG pulse, can be confirmed through the differential form.

It is reported that these PPG pulses have differences and effects depending on the body part measured and age. PPG pulses measured simultaneously from the ear, wrist, arm, and fingers are all different, and in particular, it may be difficult to determine the location of dicrotic notch and relaxation for values measured from the earlobe. Age can affect these difficulties. Confirming age-related changes in PPG pulse characteristics, it can be seen that the PPG of elderly people has difficulty in determining the location of overlap pulse notches and relaxation compared to that of young adults, such as pulse values measured from the earlobe. In this way, the PPG pulse can reflect the characteristics of the measurement location and age group, and measurement and analysis that takes this into account may be required.

When using PPG pulses that reflect the characteristics of elderly people, it is possible to develop indicators that can quantitatively evaluate various physical functional states of elderly people. In particular, by developing an indicator to check vascular status with age, a standard that can be used for personal health management through non-contact measurement can be established. This can be used to quantitatively determine the degree of vascular aging, and an aging index (AI) can be developed using SDPPG. AI that can evaluate cardiovascular aging from SDPPG can be defined as the difference between the first inflection point of SDPPG and the remaining four inflection points. Existing models can predict the age of most elderly people to a high level.

However, age is predicted based on the PPG pulse measured from the finger, and standards for assessing the level of aging using additional variables other than AI are not proposed. The age estimation device 100 derives new AI from the TDPPG pulse, which can represent the inflection state of the pulse more specifically than SDPPG, and provides more advanced age-dependent blood vessels through linear regression using deep learning technology along with AI according to SDPPG. Status indicators can be presented.

The age estimation device 100 can estimate the age of the subject 101 using variables derived in advance for the SDPPG signal 112 and the TDPPG signal 113. The age estimation device 100 determines a difference variable based on the second peak values of the SDPPG signal, determines an accumulation variable based on the third peak values of the TDPPG signal, and creates an age estimation model based on the difference variable and the accumulation variable. The model output can be obtained by executing (120). The age estimation device 100 may estimate the age of the subject based on the model output. As will be explained in detail below, the difference variable may correspond to AI, and the cumulative variable may correspond to cumulative TDPPG. The cumulative variable is an additional variable derived from TDPPG along with AI that can estimate age, and the age estimation model 120 can be derived by combining variables with significant correlation, such as difference variables and cumulative variables.

Variables and age estimation models can be derived under specific conditions, and those conditions can also be applied when estimating age. For example, the variables and age estimation model can be derived from PPG measured from the earlobe of an elderly (e.g., 65 years or older) woman, and thus the subject 101 is an elderly (e.g., 65 years or older) or older woman. It may be a woman, and the target area may be the earlobe. For example, PPG can be measured in advance for a sample group, and AI and cumulative variables can be calculated from SDPPG and TDPPG. For example, the sample group could be 84 elderly Korean women (age: 71.19 ± 6.97 years). Accordingly, an age estimation model 120 that indexes the age of a female elderly person can be derived. For example, the age estimation model 120 can predict aging of elderly women based on PPG-based cardiovascular function measured in a body part (eg, earlobe).

Figure 2 shows PPG, SDPPG, and TDPPG signals according to one embodiment.

PPG can be measured in advance for a sample group. For example, the sample group could be 84 elderly women (age: 71.19 ± 6.97 years) with normal cognitive and musculoskeletal functions and no clinical signs of arrhythmia. The characteristics of the sample group can be shown in Table 1.

characteristic value Age (year) 71.19 ± 6.97 Weight (Kg) 58.94 ± 9.33 Height (cm) 153.73 ± 8.11

With all subjects in a sample group at rest, the heart rate pulse can be measured from the PPG of a part of the subject's body (e.g., earlobe). For example, a measurement sensor can use visible light of 640 nm and have a relative sensitivity of 60%. The sensor may include a transmitting unit using a RED LED and a receiving unit using a photo-diode. A photo-diode can output current according to the intensity of received light. The sensor circuit may convert the current into voltage, amplify and filter the signal, and provide a final output signal. The frequency response of the receiver including the photo-diode may correspond to the -3dB band in the 0.3~5Hz range in the characteristic curve (gain with frequency). This can correspond to the frequency response of the filter that removes unnecessary noise. The time delay of the output signal relative to the input signal may be 72.0 ms.

PPG data can be measured for a certain period of time (e.g. 5 minutes) at a certain sampling frequency (e.g. 250 Hz), and trends or drifts can be removed through detrending in the initial preprocessing stage. To remove low- and high-frequency noise of DC components such as movement noise and breathing rhythm, a filter (e.g., 4-th order band-pass zero phase Butterworth filter) considering phase delay and distortion with a cutoff frequency of 0.4 to 9 Hz is applied. You can. Morphologically, the pulse-to-pulse interval can be defined by setting the point where the notch is minimized after passing the overlapping pulse notch in the volume pulse trace as the end point and start point of the pulse. You can. In order to obtain a representative single pulse value from the values of the interval between pulses, the representative pulse value can be obtained as the average value at each time point based on the starting point of the pulse. Second- and third-order differentiation can be performed to obtain SDPPG and TDPPG from each subject's single-pulse PPG signal. SDPPG and TDPPG signals in the second and third differential forms in continuous pulse data are shown in Figure 2. As shown in Figure 2, SDPPG has a-wave (rise toward contraction), b-wave (early contraction peak), c-wave (late contraction and double pulse depression), d-wave (relaxation), An e-wave (pulse followed by a decrease) can be defined.

Through multiple linear regression analysis, a model that estimates the dependent variable, age, can be derived. b/a, c/a, d/a, e/a as the ratio of a, b, c, d, and e waves of SDPPG, Tab, Tac, Tad, Tae as the time between waves, and vascular aging index. The variable AI (e.g. AI=(bcde)/a)) and the cumulative TDPPG (e.g. ∑TDPPG(*10 ⁴ )) variable according to the root mean square (rms) of TDPPG can be set as independent variables. there is. Regression analysis may include a stepwise method. A data set of a randomly selected portion (e.g., 68 people, approximately 80%) of the total subjects (e.g., 84 people) can be used in the model. The randomly selected remaining data (e.g., 16 people, approximately 20%) can form a test set to verify the derived model.

Figure 3 shows a neural network model according to one embodiment.

A neural network that estimates age can be derived through SDPPG and TDPPG variables. For example, four variables may be considered as input to a neural network, and actual age may be the target.

A neural network can initially have arbitrary weights, and the results of this example may appear slightly different in each run of the neural network. To avoid this randomness, a random seed can be set.

A neural network that learns age estimation may include hidden layers and output layers. For example, the hidden layer may be a single hidden layer containing multiple neurons (e.g., 15), and the output layer may be two layers of a feed forward neural network (FFNN). FFNN may not form connection cycles between units. In this network, information moves in only one direction - from the input layer to the hidden layer, so there can be no cycles or loops in the network. Learning the FFNN model can minimize the error in the output value through repeated updates of weights. For example, the error can be calculated through back propagation using Levenberg-Marquardt weights.

In embodiments, input samples may be divided into a training set, validation set, and test set. The training set is used for training the neural network, and the validation set can be used from the training set data during training until the desired accuracy is reached. The test set can be divided into measures that are completely independent from the validation and training sets to check the accuracy of the neural network. According to one embodiment, the Levenberg-Marquardt method, which combines the Gauss-Newton method and the gradient descent method, can be used as an optimization function when learning a neural network model. The performance improvement of a neural network during training can be confirmed by measuring the mean squared error (MSE). MSE can be expressed in a logarithmic scale. Performance can be expressed for training, validation, and test sets respectively. The network that showed the best performance on the validation set may be the final network. The MSE of a trained neural network can be measured on test samples. A regression across all data can be displayed to check the level of fit of the final neural network.

When analyzing the correlation between age and all variables, the C/A, TAC, AI, and cumulative TDPPG variables appear to have a significant correlation with age, as shown in Table 2. Figure 4 shows linear fitting results for candidate variables according to one embodiment. Figure 4 can be derived through first-order linear fitting of four variables correlated with age: C/A, Tac, AI, and cumulative TDPPG.

mean±SD r p b/a -1.10±0.07 0.31 N.S. c/a 0.14±0.06 -0.54 .00* d/a 0.03±0.06 0.12 N.S. e/a 0.21±0.05 0.00 N.S. T _ab (ms) 44.21±2.57 -0.05 N.S. T _ac (ms) 88.21±6.48 0.33 0.1* T _ad (ms) 108.07±5.03 0.08 N.S. A.I. -1.49±0.14 0.53 .00* ∑TDPTG * 10 ⁴ 100.18±0.05 -0.52 .00*

As a result of multiple linear regression analysis using the stepwise method with four variables correlated with the dependent variable age as independent variables, an age estimation model containing only two variables, AI and cumulative TDPPG, can be derived. The adjusted-r2 value is 0.46 (p=0.00), and the comprehensive results of the model can be shown in Table 3. The model developed using the estimated values can be expressed as Equation 1. Figure 5 shows estimation results for a test set of selection variables according to one embodiment.

In Equation 1, Age is the estimation result (model output), AI is the difference variable, can represent a cumulative variable.

Intercept A.I. ∑TDPTG * 10 ⁴ Estimate 109.65 22.16 -0.05 Standard Error(SE) 6.56 4.46 0.01 t-Statistics(Estimate/SE) 16.73 4.97 -4.81 P 0.00 0.00 0.00 adjusted-r ² 0.46

Figure 6 shows mse of each epoch by training, verification, and test data according to one embodiment.

As a result of FFNN training, MSE decreased rapidly as the network was trained. Performance on training, validation, and test sets is shown in Figure 6. The MSE of the optimal performance of the training model is 16.023. This corresponds to 3 epochs out of a total of 9 epochs. The MSE was lowest in Epoch 3, but after that the MSE increased on the validation and test sets.

Figure 7 shows training, validation, testing, regression results with all data, error histogram, and Brandt-Altman plot according to one embodiment.

Training, validation, testing, and regression results for all data are shown in Figure 7a. The r of the training data results for the regression model derived through FFNN is 0.855, the r of the validation data results is 0.701, the r of the test data results is 0.859, and the r of all data results is 0.839. The error histogram in Figure 7b shows the error of the estimation result. The Bland-Altman plot can be used to compare actual age with the output values of a learning model.

Figure 8 shows an age estimation device and a measurement device according to an embodiment.

The age estimation device 810 includes a processor 811 and a memory 812. The memory 812 is connected to the processor 811 and can store instructions executable by the processor 811, data to be operated by the processor 811, or data processed by the processor 811. Memory 812 may include non-transitory computer-readable media, such as high-speed random access memory and/or non-volatile computer-readable storage media (e.g., one or more disk storage devices, flash memory devices, or other non-volatile solid-state memory devices). It can be included.

The processor 811 may execute instructions to perform one or more operations described with reference to FIGS. 1 to 7 and FIG. 9 . For example, the processor 811 measures PPG for the target area of the estimated subject to obtain a PPG signal, performs differentiation on the PPG signal to obtain an SDPPG signal and a TDPPG signal, and obtains the secondary peak values of the SDPPG signal. Determine the difference variable based on , determine the accumulation variable based on the third peak values of the TDPPG signal, run the age estimation model based on the difference variable and the accumulation variable to obtain model output, and based on the model output The age of the subject can be estimated.

The age estimation device 810 can measure PPG from the target area of the estimated subject through the measurement device 820. For example, the measurement device 820 may correspond to a wearable device fixed to the target area of the estimated subject. The age estimation device 810 may measure PPG through the measurement device 820 for a predetermined period of time. The PPG signal may repeatedly include systole and diastole depending on the measurement time, and one cycle including one systole and one diastole is extracted, or a representative waveform corresponding to one cycle is extracted through averaging. This can be derived. Second peak values and third peak values may be determined in one extracted cycle or one cycle of the representative waveform.

Specifically, the secondary peak values of the SDPPG signal are the first secondary peak value and the second secondary peak sequentially corresponding to the inflection points of the PPG signal in one cycle including one systolic phase and one diastolic phase of the PPG signal. value, the third secondary peak value, the fourth secondary peak value, and the fifth secondary peak value, and the difference variable is the second secondary peak value, the third secondary peak value, and the fourth secondary peak value. It may correspond to the result of dividing the peak value and the remainder after subtracting the fifth secondary peak value by the first secondary peak value. In addition, the 3rd peak values of the TDPPG signal are the 1st 3rd peak value, the 2nd 3rd peak value, the 3rd 3rd peak value, and the 4th 3rd peak value, which sequentially correspond to the inflection points of the SDPPG signal in one cycle of the PPG signal. It may include a tertiary peak value, and a fifth tertiary peak value, and the cumulative variable may include a first tertiary peak value, a second tertiary peak value, a third tertiary peak value, a fourth tertiary peak value, and It may correspond to the root mean square (rms) of the TDPPG signal according to the fifth third peak value.

As shown in Figure 2, the five inflection sections of the SDPPG signal may include a, b, c, d, and e. a, b, c, d, and e may correspond to the first secondary peak value, the second secondary peak value, the third secondary peak value, the fourth secondary peak value, and the fifth secondary peak value. there is. a, b, c, d, and e are values corresponding to the amplitude of the peak, and indicators for the ratio of amplitude or duration can be used as evaluation variables. In general, a and b-waves may correspond to the early systolic component with little influence from reflection waves, c-waves correspond to mid-systolic contraction, and d is after contraction. systolic), e may correspond to the position after the dicrotic notch wave or the diastolic component. Depending on the elderly or the measurement location, the wave may be gentle, and in this case, detection of the overlap pulse notch and the subsequent relaxation component may be difficult. Therefore, when differentiating PPG data showing gentle waves, the position may appear delayed.

According to embodiments, in the case of a and b-waves, it may appear in a similar position to the general initial contraction component, and in the case of d and e-waves, it may appear after the overlapping pulse notch. This may be due to the gentle nature of the curve as the measurement location is the earlobe and the target is relatively elderly people in their 70s and 80s. When measuring the fingers of typical young subjects, the results can be interpreted as a, b, c, and d waves corresponding to contraction and e-waves corresponding to relaxation. In cases where characteristic PPG waves appear as in the embodiments, the inflection points of a and b-waves are similar, but it can be considered that c and d-waves correspond to overlap pulse depression and relaxation rather than mid-contraction or post-contraction. there is.

The index divided by the sum of c and d by a may mean the level of blood transfer ability of the peripheral vascular system. If you are young or have good blood flow, the c-wave may be high and the d-wave may be low, and if the blood circulation is poor in the elderly, the c-wave may be low. The double pulse notch, which refers to cardiac load in a static state between contraction and relaxation, can be clearly detected in the finger PPG waves of young adults. In this case, c and d refer to the section after contraction and immediately before the double pulse notch, and the sections c and d in the embodiments are the double pulse notch and contraction section, so the cardiovascular state of cardiac load and relaxation in a static state. The better the c-wave, the larger the c-wave can appear.

According to the embodiments, in the correlation between chronological age (actual age) and the variables, there is a significant correlation in the variable cumulative TDPPG (∑TDPPG(*104)), which is the sum of the c/a and AI of SDPPG and the rms of TDPPG. You can confirm that this exists.

c-waves can occur under the influence of b and d. The ratio of c/a may reflect a decrease in arterial stiffness. The c/a ratio may decrease with age. The b-wave can influence the c-wave. The b-wave may correspond to the first vascular response upon blood ejection from the left ventricle and may reflect aorta tensibility. b/a can indicate whether aortic blood volume increases rapidly during cardiac ejection, and this indicator can increase when the corresponding vascular load decreases. Conversely, in the case of large arteries, stiffness may decrease.

The AI variable is the difference between b, c, d, and e divided by a, and can reflect chronological age (actual age). AI can also be used as an indicator of functional aging of blood vessels and severe arteriosclerosis. This value may also increase during vascular aging in normal people.

c' of the third order derivative may be defined as the initial contraction component (p1), and d' may refer to the late contraction or early relaxation component (p2). When using third-order differentiation, the inflection point of second-order differentiation can be seen more clearly. From the variables calculated from TDPPG and the evaluation of arterial stiffness according to pulse, it can be confirmed that mental stress has a significant impact on arterial stiffness. Accordingly, the cumulative TDPPG variable, which represents the change trend of the entire pulse as the sum of the areas of the inflection point center wave, can also be used as a quantitative value that can overall represent the elasticity of blood vessels corresponding to the overall blood flow.

You can create a regression model to derive age using three highly correlated variables. As a result of confirming the accuracy, the c/a variable can be removed by the stepwise method during regression analysis, and the AI and cumulative TDPPG (∑TDPPG(*104)) variables can be selected. The r of the developed model can have an accuracy of 0.68, and the adjusted-r2 value can have an accuracy of 0.46. In the results of a randomly selected test-set, r may be lowered to 0. The r value can be improved compared to when AI and cumulative TDPPG variables are used independently.

Additionally, FFNN can be used to improve the accuracy of the estimation model. As a result, when performing multiple linear regression analysis, an r value of 0.859 can be derived, which is higher than the r value of 0.308 in the test set. Out of a total of 9 epochs, the model learning results conducted in the 3rd epoch can produce the best results. In order to improve the model results derived from multiple linear regression analysis, it can be seen that the accuracy of the model can be improved even without increasing the number of subjects.

According to embodiments, additional variables in addition to AI can be derived, and a model that estimates chronological age based on cardiovascular function can be developed by combining meaningful variables. If the measurement location for PPG is the earlobe rather than the typical fingertip, it may have greater significance if the target is female elderly people. The waveform may appear differently depending on the measurement location. Embodiments can provide a basis for developing various indicators that can monitor physical functions and provide personalized diagnoses in the daily lives of elderly people, taking into account measurement locations in a non-invasive manner.

9 shows an electronic device according to one embodiment.

The electronic device 900 may include a processor 910, memory 920, camera 930, storage device 940, input device 950, output device 960, and network interface 970, They can communicate with each other via communication bus 980. For example, electronic device 900 may include mobile devices such as mobile phones, smart phones, PDAs, netbooks, tablet computers, laptop computers, etc., wearable devices such as smart watches, smart bands, smart glasses, etc., computing devices such as desktops, servers, etc. , may be implemented as at least part of a vehicle such as home appliances such as televisions, smart televisions, refrigerators, etc., security devices such as door locks, autonomous vehicles, smart vehicles, etc. The electronic device 900 may structurally and/or functionally include age estimation devices 100 and 810 .

The processor 910 executes functions and instructions for execution within the electronic device 900. For example, the processor 910 may process instructions stored in the memory 920 or the storage device 940. The processor 910 may perform one or more operations described with reference to FIGS. 1 to 8 . Memory 920 may include a computer-readable storage medium or computer-readable storage device. The memory 920 may store instructions for execution by the processor 910 and may store related information while software and/or applications are executed by the electronic device 900 .

Camera 930 may take photos and/or video. Storage device 940 includes a computer-readable storage medium or computer-readable storage device. The storage device 940 can store a larger amount of information than the memory 920 and can store the information for a long period of time. For example, storage device 940 may include a magnetic hard disk, optical disk, flash memory, floppy disk, or other forms of non-volatile memory known in the art.

The input device 950 may receive input from the user through traditional input methods such as a keyboard and mouse, and new input methods such as touch input, voice input, and image input. For example, input device 950 may include a keyboard, mouse, touch screen, microphone, or any other device capable of detecting input from a user and transmitting the detected input to electronic device 900. The output device 960 may provide the output of the electronic device 900 to the user through visual, auditory, or tactile channels. Output device 960 may include, for example, a display, touch screen, speaker, vibration generating device, or any other device capable of providing output to a user. The network interface 970 can communicate with external devices through a wired or wireless network.

The embodiments described above may be implemented with hardware components, software components, and/or a combination of hardware components and software components. For example, the devices, methods, and components described in the embodiments may include, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, and a field programmable gate (FPGA). It may be implemented using a general-purpose computer or a special-purpose computer, such as an array, programmable logic unit (PLU), microprocessor, or any other device capable of executing and responding to instructions. The processing device may execute an operating system (OS) and software applications running on the operating system. Additionally, a processing device may access, store, manipulate, process, and generate data in response to the execution of software. For ease of understanding, a single processing device may be described as being used; however, those skilled in the art will understand that a processing device includes multiple processing elements and/or multiple types of processing elements. It can be seen that it may include. For example, a processing device may include multiple processors or one processor and one controller. Additionally, other processing configurations, such as parallel processors, are possible.

Software may include a computer program, code, instructions, or a combination of one or more of these, which may configure a processing unit to operate as desired, or may be processed independently or collectively. You can command the device. Software and/or data may be used on any type of machine, component, physical device, virtual equipment, computer storage medium or device to be interpreted by or to provide instructions or data to a processing device. , or may be permanently or temporarily embodied in a transmitted signal wave. Software may be distributed over networked computer systems and stored or executed in a distributed manner. Software and data may be stored on a computer-readable recording medium.

The method according to the embodiment may be implemented in the form of program instructions that can be executed through various computer means and recorded on a computer-readable medium. A computer-readable medium may store program instructions, data files, data structures, etc., singly or in combination, and the program instructions recorded on the medium may be specially designed and constructed for the embodiment or may be known and available to those skilled in the art of computer software. there is. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes, optical media such as CD-ROMs and DVDs, and magnetic media such as floptical disks. -Includes optical media (magneto-optical media) and hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, flash memory, etc. Examples of program instructions include machine language code, such as that produced by a compiler, as well as high-level language code that can be executed by a computer using an interpreter, etc.

The hardware devices described above may be configured to operate as one or multiple software modules to perform the operations of the embodiments, and vice versa.

As described above, although the embodiments have been described with limited drawings, those skilled in the art can apply various technical modifications and variations based on this. For example, the described techniques are performed in a different order than the described method, and/or components of the described system, structure, device, circuit, etc. are combined or combined in a different form than the described method, or other components are used. Alternatively, appropriate results may be achieved even if substituted or substituted by an equivalent.

Therefore, other implementations, other embodiments, and equivalents of the claims also fall within the scope of the claims described below.

Claims (11)

In the age estimation method,
Obtaining a PPG signal by measuring photoplethysmography (PPG) on the target area of the estimated subject;
performing differentiation on the PPG signal to obtain a second derivative PPG (second derivative of PPG, SDPPG) signal and a third derivative PPG (third derivative of PPG, TDPPG) signal;
determining a difference variable based on secondary peak values of the SDPPG signal;
determining a cumulative variable based on third peak values of the TDPPG signal;
executing an age estimation model based on the difference variable and the cumulative variable to obtain model output; and
Estimating the age of the estimated subject based on the model output
Including,
The secondary peak values of the SDPPG signal are a first secondary peak value and a second secondary peak value sequentially corresponding to inflection points of the PPG signal in one cycle including one systole and one diastole of the PPG signal. a peak value, a third secondary peak value, a fourth secondary peak value, and a fifth secondary peak value,
The difference variable is a result of subtracting the third secondary peak value, the fourth secondary peak value, and the fifth secondary peak value from the second secondary peak value and dividing the remainder by the first secondary peak value. Corresponding to,
How to estimate age. delete According to paragraph 1,
The 3rd peak values of the TDPPG signal are a 1st 3rd peak value, a 2nd 3rd peak value, and a 3rd 3rd peak value that sequentially correspond to inflection points of the SDPPG signal in the one cycle of the PPG signal. , a fourth tertiary peak value, and a fifth tertiary peak value,
The cumulative variable is the TDPPG signal according to the first tertiary peak value, the second tertiary peak value, the third tertiary peak value, the fourth tertiary peak value, and the fifth tertiary peak value. Corresponding to the root mean square (rms),
How to estimate age. According to paragraph 1,
The estimated target is a woman over 65 years of age,
The target area is the earlobe,
How to estimate age. According to paragraph 1,
The step of acquiring the PPG signal is
Comprising the step of measuring the PPG for a predetermined time through a wearable device fixed to the target area of the estimated subject,
How to estimate age. According to paragraph 1,
The age estimation model is
Derived through a pre-trained neural network model based on input samples including a training set, validation set, and test set,
The neural network model includes a single hidden layer including a plurality of neurons and an output layer including two layers of a feed forward neural network (FFNN),
How to estimate age. According to paragraph 1,
The model output is
Obtained through the following equation,

In the above equation, Age is the model output, AI is the difference variable, and represents the cumulative variable,
How to estimate age. A computer program combined with hardware and stored in a computer-readable recording medium to execute the method of any one of claims 1 and 3 to 7. In the age estimation device,
processor; and
Memory containing instructions executable by the processor
Including,
When the instructions are executed on the processor, the processor
Obtain a PPG signal by measuring photoplethysmography (PPG) on the target area of the estimated subject,
Perform differentiation on the PPG signal to obtain a second derivative PPG (second derivative of PPG, SDPPG) signal and a third derivative PPG (third derivative of PPG, TDPPG) signal,
Determine a difference variable based on the secondary peak values of the SDPPG signal,
Determine a cumulative variable based on the third peak values of the TDPPG signal,
Running an age estimation model based on the difference variable and the cumulative variable to obtain a model output,
Estimate the age of the estimated subject based on the model output,
The secondary peak values of the SDPPG signal are a first secondary peak value and a second secondary peak value sequentially corresponding to inflection points of the PPG signal in one cycle including one systole and one diastole of the PPG signal. a peak value, a third secondary peak value, a fourth secondary peak value, and a fifth secondary peak value,
The difference variable is a result of subtracting the third secondary peak value, the fourth secondary peak value, and the fifth secondary peak value from the second secondary peak value and dividing the remainder by the first secondary peak value. Corresponding to,
Age estimation device. delete According to clause 9,
The 3rd peak values of the TDPPG signal are a 1st 3rd peak value, a 2nd 3rd peak value, and a 3rd 3rd peak value that sequentially correspond to inflection points of the SDPPG signal in the one cycle of the PPG signal. , a fourth tertiary peak value, and a fifth tertiary peak value,
The cumulative variable is the TDPPG signal according to the first tertiary peak value, the second tertiary peak value, the third tertiary peak value, the fourth tertiary peak value, and the fifth tertiary peak value. Corresponding to the root mean square (rms),
Age estimation device.

Images