KR20200092859A - Method and apparatus for pre-processing ppg signal - Google Patents

Abstract

The present invention relates to a device for preprocessing a PPG signal and a method thereof. According to an embodiment of the present invention, a PPG signal preprocessing device can comprise: a PPG signal receiver for receiving a PPG signal including a pulse; a moving average filter for smoothing the PPG signal to obtain a smoothed PPG signal; a drift elimination engine for eliminating a drift from the smoothed PPG signal by using a fitted polynomial function to obtain a drift-eliminated PPG signal; and a motion artifact correction engine for correcting motion artifacts of the drift-eliminated PPG signal by replacing the pulse with the combination of a global signal pulse and a local signal pulse to obtain a rectified PPG signal.

Description

PPG signal pre-processing method and apparatus {METHOD AND APPARATUS FOR PRE-PROCESSING PPG SIGNAL}

The present invention relates to a signal processing system, and more particularly, to a method and apparatus for preprocessing a PPG (Photoplethysmography).

PPG signals have been widely used in non-invasive methods to measure human vitality such as blood pressure, blood oxygenation (SpO2), and heart rate. PPG signals are generally affected by various noises, drifts, and motion artifacts, among which motion artifacts greatly affect the quality of the PPG signal. Motion artifacts affect a person's pulse, and therefore most of the characteristics based on the shape of the pulse. Removing motion artifacts without affecting the PPG signal is an important task in the PPG signal preprocessing process. There are frequency-based and time-frequency-based methods for motion artifact removal. However, this technique tends to remove signals simultaneously as the frequency bands of motion artifacts and PPG signals overlap. Therefore, removing motion artifacts using the frequency domain method may remove the content of the PPG signal. Methods or alternatives to overcome these shortcomings are urgently emerging.

It is an object to provide a method and apparatus for preprocessing a PPG signal.

A PPG signal pre-processing method according to an embodiment for achieving the above object comprises: receiving a PPG signal including a pulse by a device; Smoothing, by the apparatus, the PPG signal using a moving average filter to obtain a smoothed PPG signal; Removing, by the apparatus, drift from the smoothed PPG signal using a fitted polynomial function to obtain a drifted PPG signal; And correcting, by the apparatus, the motion artifact of the drift-removed PPG signal in order to obtain a rectified PPG signal by replacing the pulse with a combination of a global signal pulse and a local signal pulse.

At this time, the PPG signal preprocessing method according to an embodiment includes extracting a feature from the rectified PPG signal by the apparatus.

At this time, the PPG signal pre-processing method according to an embodiment includes predicting a vital signal value based on the extracted feature by the apparatus.

At this time, the step of obtaining a drift-removed PPG signal by removing drift from the smoothed PPG signal by using the fitted polynomial function by the apparatus is performed by adding the fitted polynomial function to the predefined smoothed PPG signal. Determining with a polynomial approximation; And subtracting a fitted polynomial function from the smoothed PPG signal to obtain the drift-removed PPG signal.

At this time, the step of determining the global signal pulses includes ascending pulses and falling pulses based on the trough and peak of the drift-removed PPG signal for each pulse of the drift-removed PPG signal. (descending pulse); Obtaining an average ascending length as an average of the number of samples in each rising pulse; Acquiring an average descending length as an average of the number of samples in each falling pulse; Resampling each rising pulse to the average rising length using a linear interpolation to obtain a resampled rising pulse; Resampling each falling pulse to the average falling length using a linear interpolation method to obtain a resampled falling pulse; Obtaining an average rising pulse as an average of the resampled rising pulses; Obtaining an average falling pulse as an average of the resampled falling pulses; And concatenating the average rising pulse and the average falling pulse.

At this time, the step of determining the local signal pulse may include: separating a predefined number of local pulses of the PPG signal drift-drift around the pulse to be rectified; Dividing each local pulse into a local rising pulse and a local falling pulse; Resampling each local rise pulse to the average rise length using a linear interpolation method to obtain a resampled local rise pulse; Resampling each local falling pulse to the average falling length using a linear interpolation method to obtain a resampled local falling pulse; Obtaining an average local rising pulse as an average of the resampled local rising pulses and obtaining an average local falling pulse as an average of the resampled local falling pulses; And connecting the average local rising pulse and the average local falling pulse.

At this time, correcting the motion artifact of the drift-removed PPG signal to obtain a rectified PPG signal includes replacing each local signal pulse with a convex combination of the local signal pulse and the global signal pulse. Obtaining an intermediate rectified PPG signal; And resampling each pulse of the intermediate rectified PPG signal with the number of samples in the corresponding pulse of the drift-removed PPG signal.

At this time, resampling each pulse of the intermediate rectified PPG signal with the number of samples in the corresponding pulse of the drift-removed PPG signal in order to obtain a rectified PPG signal may include re-sampling each pulse of the intermediate rectified PPG signal. Dividing the number of samples into a rising pulse corresponding to the average rising length and a falling pulse whose number of samples corresponds to the average falling length; Resampling each rising pulse with the number of samples of the rising pulse of the corresponding pulse in the drift-removed PPG signal to obtain a rectified rising pulse; Resampling each falling pulse with the number of samples of the falling pulse of the corresponding pulse in the drift-removed PPG signal to obtain a rectified falling pulse; And connecting each rectified rising pulse with a corresponding rectified falling pulse to obtain the rectified PPG signal.

At this time, the trough and peak represent sample indices corresponding to the trough and peak of the smoothed PPG signal.

At this time, the pulse to be replaced is determined by adjusting the peak and trough so that peaks occurring before the first trough and after the last trough are eliminated, and as a result, only pulses starting at the trough and ending at the trough are considered to be rectified.

At this time, the features include at least one of Kaiser Teager Energy (KTE) features, Spectral Entropy features and Spectral Energy Logarithmic features. do.

In addition, the PPG signal pre-processing apparatus includes a PPG signal receiver for receiving a PPG signal including a pulse; A moving average filter that smoothes the PPG signal to obtain a smoothed PPG signal; A drift removal engine that removes drift from the smoothed PPG signal using a fitted polynomial function to obtain a drift-removed PPG signal; It consists of a motion artifact correction engine that corrects the motion artifacts of the drift-removed PPG signal by replacing the pulses with a combination of global signal pulses and local signal pulses to obtain a rectified PPG signal.

At this time, the PPG signal pre-processing apparatus according to an embodiment further includes a vital signal prediction engine that extracts features from the rectified PPG signal and predicts a vital signal based on the extracted features.

At this time, the drift removal engine determines the fitted polynomial function as a polynomial approximation of a predefined smoothed PPG signal, and calculates a fitted polynomial function from the smoothed PPG signal to obtain the drift-removed PPG signal. do.

In this case, the global signal pulse is divided into rising pulses and falling pulses based on the trough and peak of the drift-removed PPG signal, and each pulse of the drift-removed PPG signal. To obtain the average rising length as the average of the number of samples, to obtain the average falling length as the average of the number of samples from each falling pulse, and to obtain the resampled rising pulse, each rising pulse is converted to the average rising length using a linear interpolation method. Resampling, resampling each falling pulse to the average falling length using a linear interpolation method to obtain a resampled falling pulse, obtaining an average rising pulse as the average of the resampled rising pulses, and obtaining an average of the resampled falling pulses It is determined by acquiring an average falling pulse, and connecting the average rising pulse and the average falling pulse.

At this time, the local signal pulse separates a predefined number of local pulses of the drift-removed PPG signal around the pulse to be rectified, divides each local pulse into a local rising pulse and a local falling pulse, and resamples the local rise. Each local rising pulse is resampled to an average rising length using a linear interpolation method to obtain a pulse, and each local falling pulse is resampled to an average falling length using a linear interpolation method to obtain a resampled local falling pulse, and the resampling is performed. It is determined by obtaining an average local rising pulse as an average of the local rise pulses obtained, obtaining an average local falling pulse as an average of the resampled local falling pulses, and connecting the average local rising pulse and the average local falling pulses. .

At this time, the motion artifact correction engine acquires an intermediate rectified PPG signal by replacing each local signal pulse with a convex combination of the local signal pulse and the global signal pulse, and the number of samples in each pulse of the drift-removed PPG signal. Each pulse of the PPG signal intermediately rectified with is resampled.

At this time, when the motion artifact correction engine resamples each pulse of the intermediate rectified PPG signal with the number of samples in each pulse of the drift-removed PPG signal, each pulse of the intermediate rectified PPG signal is sampled. Is divided by the rising pulse corresponding to the average rising length and the number of samples falling into the falling pulse corresponding to the average falling length, and to obtain the rectified rising pulse, the number of samples of the rising pulse of the corresponding pulse in the drift-removed PPG signal. Resample each rising pulse, resample each falling pulse from the drift-removed PPG signal to the number of samples of the falling pulse of the corresponding pulse to obtain a rectified falling pulse, and obtain each rectified PPG signal. The rectified rising pulse is connected to the corresponding rectified falling pulse.

At this time, the trough and peak represent the sample index corresponding to the trough and peak of the smoothed PPG signal.

At this time, the pulse to be replaced is determined by adjusting the peak and trough so that peaks occurring before the first trough and after the last trough are removed, and as a result, only pulses starting at the trough and ending at the trough are rectified.

In this case, the feature includes at least one of Kaizer Teager Energy (KTE) features, Spectral Entropy features and Spectral Energy Logarithmic features. do.

Through the pre-processing process of the PPG signal, it is possible to extract features from the rectified PPG signal and predict the vital signal value based on this.

1 is a block diagram of a PPG signal pre-processing apparatus according to an embodiment.
2 is a block diagram of a motion artifact correction engine according to an embodiment.
3 is a flowchart of a method of preprocessing a PPG signal and extracting features for predicting a vital signal according to an embodiment.
4 is a flow chart for removing drift from a smoothed PPG signal using a fitted polynomial function to obtain a drift-removed PPG signal according to an embodiment.
5 is a flowchart of correcting a motion artifact of a drift-removed PPG signal to obtain a rectified PPG signal according to an embodiment.
6 is a flow diagram for determining a global signal pulse according to one embodiment.
7 is a flow diagram for determining a local signal pulse according to one embodiment.
8 is a graph showing a smoothed PPG signal after applying a moving average filter.
9 is a graph showing a signal that has been removed from drift.
10 is a graph showing that the average rising pulse and the average falling pulse are calculated and connected to obtain global/local signal pulses.
FIG. 11 is a graph showing the original signal rectified and the rectified PPG signal when α=0.3 and M=3.
FIG. 12 is a graph showing the original signal rectified and the rectified PPG signal when α=0.5 and M=5.
13 is a graph showing characteristics of a PPG signal for defining a quality evaluation metric.

Specific details of other embodiments are included in the detailed description and drawings. The advantages and features of the described technology, and how to achieve them will become apparent with reference to the embodiments described below in detail in conjunction with the drawings. The same reference numerals refer to the same components throughout the specification.

Terms such as first and second may be used to describe various components, but the components should not be limited by the terms. The terms are only used to distinguish one component from other components. Singular expressions include plural expressions unless the context clearly indicates otherwise. Also, when a part “includes” a certain component, this means that other components may be further included instead of excluding other components, unless otherwise specified. In addition, terms such as “part” and “module” described in the specification mean a unit that processes at least one function or operation, which may be implemented in hardware or software, or a combination of hardware and software.

In addition, as is common in the art, functions may be performed by blocks in an embodiment, and blocks referred to as "units" and "modules" may include logic gates, integrated circuits, microprocessors, microcontrollers, and memory circuits. , Physically implemented by analog or digital circuitry such as passive electronic components, active electronic components, optical components, circuits embedded in the hardware, and can optionally be driven by firmware and software.

Further, the circuit may be implemented with one or more semiconductor chips or on a substrate support such as a printed circuit board. The circuit constituting the block may be implemented by dedicated hardware or a processor, or may be performed by a combination of dedicated hardware for performing some functions of the block and a processor for performing other functions of the block.

The PPG signal pre-processing method according to an embodiment includes: receiving a PPG signal including a pulse by a device, and smoothing the PPG signal by using a moving average filter to obtain a smoothed PPG signal by the device. , By the device, removing drift from the smoothed PPG signal using a fitted polynomial function to obtain a drift-depleted PPG signal, and by the device, the pulse is generated from global and local signal pulses. And correcting motion artifacts of the drift-removed PPG signal to obtain a rectified PPG signal by replacing with a combination.

Unlike general methods and systems, the proposed method can be used to correct the motion artifacts of the drift-removed PPG signal without using additional sensors or reference signals.

The proposed method can be used to calculate the average pulse by dividing each pulse into rising pulses and falling pulses. As a result, an accurate pulse average can be generated to obtain better motion artifact correction results.

In addition, the proposed method can be used to rectify each pulse by a convex combination of a local signal pulse and a global signal pulse, which helps to prevent the rectified signal from degrading significantly when motion artifacts exist for a longer period. .

In addition, this method can be used to resample the rectified pulses to regain period information lost during the average calculation.

In addition, the proposed time series-based pre-processing method can remove noise effects by focusing on the correction of motion artifacts.

In addition, the proposed method considers local signal pulses together with global signal pulses and can be used to individually estimate the average rising and falling pulses.

In addition, the intermediate rectified pulse is resampled to the original number of samples, so that the pulse period information is not damaged, and as a result, the PPG signal can be accurately preprocessed.

Hereinafter, a PPG signal preprocessing method and apparatus according to an embodiment will be described in detail with reference to the accompanying drawings.

1 is a block diagram of a PPG signal pre-processing apparatus according to an embodiment.

The PPG signal pre-processing device 100 may be, for example, a cellular phone, a smart phone, a Personal Digital Assistant (PDA), a tablet computer, a laptop computer, a smart watch, a smart band, a smart ring, and the like.

The PPG signal preprocessing apparatus 100 includes a PPG signal receiver 110, a moving average filter 120, a drift removal engine 130, a motion artifact correction engine 140, a vital signal prediction engine 150, and a memory 160 , A communication unit 170 and a processor 180.

The PPG signal receiver 110 receives a PPG signal including a pulse, and after receiving the PPG signal, the moving average filter 120 smooths the PPG signal. In one embodiment, the first graph of FIG. 8 represents the PPG signal. The PPG signal in the box is the second graph, and the third graph represents the smoothed signal after applying the moving average filter 120.

Smoothing of the PPG signal is performed using the moving average filter 120. For example, the moving average filter 120 of Equation 1 below has a window size 21 having the same weight (all coefficients are 1).

[Equation 1]

To obtain a drift-removed PPG signal based on the smoothed PPG signal, the drift removal engine 130 removes drift from the smoothed PPG signal using a fitted polynomial function. In one embodiment, the drift removal engine 130 determines the fitted polynomial function as a polynomial approximation of a predefined smoothed PPG signal, and uses the fitted polynomial function from the smoothed PPG signal to obtain a drift-depleted PPG signal. Subtract. Here, the polynomial order depends on the amount of drift and the length of the PPG signal.

In one embodiment, drift removal of the PPG signal is performed using a polynomial fit of degree = 30 for the entire PPG signal. The fitted polynomial is removed from the signal to eliminate drift according to Equation 2 below.

[Equation 2]

에 대한 오차 손실의 제곱을 최소화 하는 다항식 차수 "

"를 제공한다. 반환값 P는 degree 0에서 특정 degree까지에 대한 다항식 변수의 계수에 해당하는 degree+1 실수의 벡터이다. 폴리벌(polyval)은

에 해당하는 개별 샘플링 포인트에서 다항식 팻(Polynomial Pat)을 평가한다.In Equation 2, the polyfit function is

Polynomial order to minimize squared error loss for "

". The return value P is a vector of real numbers of degree+1 corresponding to the coefficient of the polynomial variable for degree 0 to a specific degree. Polyval

The polynomial pat is evaluated at individual sampling points corresponding to.

In FIG. 9, the dark gray line is a polynomial fit of the PPG signal, and the second graph of FIG. 9 shows the drift-removed signal.

Based on the drift-removed PPG signal, the motion artifact correction engine 140 replaces the pulse with a combination of a global signal pulse and a corresponding local signal pulse to obtain a rectified PPG signal, and thus the motion artifact of the drift-removed PPG signal. Correct it.

In addition, the motion artifact correction engine 140 acquires an intermediate rectified PPG signal by replacing each local signal pulse with a convex combination of the local signal pulse and the global signal pulse.

In addition, the motion artifact correction engine 140 divides each pulse of the intermediate rectified PPG signal into rising and falling pulses, where the number of samples of the rising pulse is the average rising length and the number of samples of the falling pulse corresponds to the average falling length. do. The average rising length is obtained as the average of the number of samples of each rising pulse, and the average falling length is obtained as the average of the number of samples of each falling pulse.

In addition, the motion artifact correction engine 140 resamples each rising pulse with the number of samples of the rising pulse of the corresponding pulse in the drift-removed PPG signal to obtain a rectified rising pulse, and obtains a rectified falling pulse Each falling pulse is resampled by the number of samples of the falling pulse of the corresponding pulse in the drift-removed PPG signal.

In addition, the motion artifact correction engine 140 connects each rectified rising pulse with a corresponding rectified falling pulse to obtain a rectified PPG signal.

The global signal pulse is determined by dividing each pulse of the drift-removed PPG signal into rising and falling pulses based on the trough and peak of the drift-removed PPG signal, and the number of samples in each rising pulse. Re-sample each rising pulse to the average rising length using a linear interpolation method to obtain an average rising length as an average, an average falling length as an average of the number of samples in each falling pulse, and a resampled rising pulse, To obtain a resampled falling pulse, each falling pulse is resampled to the average falling length using a linear interpolation method, an average rising pulse is obtained as an average of the resampled rising pulses, and an average falling as an average of the resampled falling pulses It is determined by acquiring a pulse and connecting the average rising pulse and the average falling pulse. At this time, the linear interpolation method is a process well known to those of ordinary skill in the PPG signal processing area.

In one embodiment, troughs and peaks represent sample indices corresponding to troughs and peaks of the smoothed PPG signal.

If S(i)=min(S(i-K):S(i+K)), the sample index i is trough. Where K is half the expected minimum pulse width. In the example where the sample rate is 200 Hz, K is set to 35.

If S(i)=max(S(i-K):S(i+K)), sample index i is a peak.

In addition, the motion artifact correction engine 140 adjusts peaks and troughs by adjusting the troughs and peaks so that each trough follows a peak or each peak follows a trough. Through this, peaks occurring before the first trough and after the last trough can be eliminated, and as a result, the signal may start and end at the trough point.

At this time, the local signal pulse is determined by separating a predefined number of local pulses around the pulse to be rectified, dividing each local pulse into a local rising pulse and a local falling pulse, and obtaining a resampled local rising pulse. In order to resample each local rising pulse to the average rising length using a linear interpolation method, each local falling pulse is resampled to an average falling length using a linear interpolation method to obtain a resampled local falling pulse, and the resampled local rising pulse. The local signal pulse is determined by acquiring an average local rising pulse as an average of, obtaining an average local falling pulse as an average of the resampled local falling pulses, and connecting the average local rising pulse and the average local falling pulse.

Also, the pulse to be replaced is determined by adjusting the peak and trough so that peaks occurring before the first trough and after the last trough are eliminated, so only complete pulses starting at the trough and ending at the trough are considered rectification.

After acquiring the rectified PPG signal, the vital signal prediction engine 150 extracts features from the rectified PPG signal and predicts the vital signal based on the extracted features. Vital signals indicate blood pressure levels, blood oxygen levels, glucose values, heart rate levels, etc., where the features are, for example, Kaizer Teager Energy (KTE) features, Spectral Entropy features and spectral energy logarithms (Spectral Energy Logarithmic) It may be at least one of the mean, variance and skewness of the feature.

The processor 180 executes instructions stored in the memory 160 and performs various processes. In addition, the communication unit 170 may communicate with internal hardware components or external devices through one or more networks.

The memory 160 stores instructions to be executed by the processor 180. Also, the memory 160 may include non-volatile storage elements. Examples of such non-volatile storage devices may include magnetic hard disks, optical disks, floppy disks, flash memories, electrical programmable memory (EPROM), or electrically electrically erasable and programmable memory (EEPROM). Also, in one embodiment, the memory 160 may be considered a non-transitory storage medium. The term "non-transitory" may indicate that the storage medium is not embodied as a carrier wave or a propagated signal. However, "non-transitory" should not be construed as memory 160 being immovable. In one embodiment, the memory 160 may correspond to a large capacity memory. In addition, non-transitory storage media can store data that can change over time. (For example, Random Access Memory (RAM) or Cache)

Although FIG. 1 shows various hardware components of the apparatus 100 for preprocessing a PPG signal, other embodiments are not limited thereto. In other embodiments, the device 100 may include more components, and the label or name of the component is not used to limit the scope of rights when used for illustrative purposes only. Further, one or more configurations may be combined together to perform the same or substantially similar function to process the PPG signal.

2 is a block diagram of a motion artifact correction engine 140 according to one embodiment.

In one embodiment, the motion artifact correction engine 140 includes an intermediate rectified PPG signal acquisition unit 140a, a pulse resampling unit 140b, and a connection unit 140c.

The intermediate rectified PPG signal acquisition unit 140a acquires the intermediate rectified PPG signal by replacing each local signal pulse with a convex combination of the local signal pulse and the global signal pulse.

In addition, the intermediate rectified PPG signal acquisition unit 140a divides each pulse of the intermediate rectified PPG signal into rising and falling pulses, where the number of samples of the rising pulse is the average rising length and the number of samples of the falling pulse is the average falling Corresponds to the length.

The pulse resampling unit 140b resamples each rising pulse with the number of samples of the rising pulse of the corresponding pulse in the drift-removed PPG signal to obtain a rectified rising pulse, and the falling pulse of the corresponding pulse in the drift-removed PPG signal. Each falling pulse is resampled by the number of samples to obtain a rectified falling pulse.

In addition, the connection unit 140c obtains the rectified PPG signal by connecting the rectified rising pulse with a corresponding rectified falling pulse.

Although FIG. 2 shows various hardware configurations of the motion artifact correction engine 140, it should be understood that the present invention is not limited thereto. In other embodiments, the motion artifact correction engine 140 may include fewer or more numbers of configurations. Also, the label or name of the composition is used for illustrative purposes and does not limit the scope. Further, one or more configurations may be combined together to perform the same or substantially similar function to process the PPG signal.

3 is a flowchart of a method of preprocessing a PPG signal and extracting features for predicting a vital signal according to an embodiment.

First, a PPG signal including a pulse is received (S302). In this case, the PPG signal receiver 110 receives the PPG signal.

Next, the PPG signal is smoothed using the moving average filter 120 (S304).

Next, a drift-removed PPG signal is obtained by removing drift from the smoothed PPG signal using a fitted polynomial function (S306). In this case, the drift removal engine 130 acquires the PPG signal with the drift removed.

Next, to obtain the rectified PPG signal, each pulse is replaced with a combination of a global signal pulse and a local signal pulse to correct the motion artifact of the drift-removed PPG signal (S308). In this case, the motion artifact correction engine 140 is used.

Next, the feature is extracted from the rectified PPG signal (S310), and the vital signal prediction engine 150 extracts the feature from the rectified PPG signal.

Next, the vital signal is predicted based on the extracted features (S312 ), and the vital signal prediction engine 150 predicts the vital signal.

Flow charts for a method of preprocessing a PPG signal and extracting features for predicting a vital signal can be performed in the order presented, in a different order or simultaneously. In addition, the above procedure can be omitted, added, modified, and the like.

4 is a flow chart for removing drift from a smoothed PPG signal using a fitted polynomial function to obtain a drift-removed PPG signal according to an embodiment. These steps S402 and S404 are performed by the drift removal engine 130.

First, a fitted polynomial function is determined as a polynomial approximation of the predefined smoothed PPG signal (S402).

Next, a drift-removed PPG signal is obtained by subtracting the fitted polynomial function from the smoothed PPG signal (S404).

The flowcharts can be performed in the order presented, in a different order or simultaneously. In addition, the above procedure can be omitted, added, modified, and the like.

5 is a flowchart of correcting a motion artifact of a drift-removed PPG signal to obtain a rectified PPG signal according to an embodiment. These steps (S502 to S510) are performed by the motion artifact correction engine 140.

First, an intermediate rectified PPG signal is obtained by replacing each local signal pulse with a convex combination of the local signal pulse and the global signal pulse (S502).

Next, each pulse of the intermediate rectified PPG signal is divided into rising and falling pulses (S504). At this time, the number of samples of the rising pulse is the average rising length, and the number of samples of the falling pulse corresponds to the average falling length.

Next, each rising pulse is resampled by the number of samples of the rising pulse of the corresponding pulse in the drift-removed PPG signal to obtain a rectified rising pulse (S506).

Next, each falling pulse is resampled by the number of samples of the falling pulse of the corresponding pulse in the drift-removed PPG signal to obtain a rectified falling pulse (S508).

Next, each rectified rising pulse is connected to a corresponding rectified falling pulse to obtain a rectified PPG signal (S510).

The flowcharts can be performed in the order presented, in a different order or simultaneously. In addition, the above procedure can be omitted, added, modified, and the like.

6 is a flow diagram for determining a global signal pulse according to one embodiment. These steps (S602 to S616) are performed by the motion artifact correction engine 140.

First, each pulse of the drift-removed PPG signal is divided into a rising pulse and a falling pulse based on the trough and peak of the drift-removed PPG signal (S602).

Next, an average rising length is obtained as an average of the number of samples in each rising pulse (S604).

Next, an average falling length is obtained as an average of the number of samples in each falling pulse (S606).

Next, each rising pulse is resampled to the average rising length using a linear interpolation method to obtain a resampled rising pulse (S608).

Next, each falling pulse is resampled to the average falling length using a linear interpolation method to obtain a resampled falling pulse (S610).

Next, an average rising pulse is obtained as an average of the resampled rising pulse (S612).

Next, an average falling pulse is obtained as an average of the resampled falling pulse (S614).

Next, a global signal pulse is determined by connecting the average rising pulse and the average falling pulse (S616).

The flowcharts can be performed in the order presented, in a different order or simultaneously. In addition, the above procedure can be omitted, added, modified, and the like.

7 is a flow diagram for determining a local signal pulse according to one embodiment. These steps (S702 to S712) are performed by the motion artifact correction engine 140.

First, a pre-defined number of local pulses of the drift-removed PPG signal around the pulse to be rectified is separated (S702).

Next, each local pulse is divided into a local rising pulse and a local falling pulse (S704).

Next, each local rising pulse is resampled to the average rising length using a linear interpolation method to obtain a resampled local rising pulse (S706).

Next, each local falling pulse is resampled to an average falling length using a linear interpolation method to obtain a resampled local falling pulse (S708).

Next, an average local rising pulse is obtained as an average of the resampled local rising pulses, and an average local falling pulse is obtained as an average of the resampled local falling pulses (S710 ).

Next, a local signal pulse is determined by connecting the average local rising pulse and the average local falling pulse (S712).

The flowcharts can be performed in the order presented, in a different order or simultaneously. In addition, the above procedure can be omitted, added, modified, and the like.

10 is a graph showing that the average rising pulse and the average falling pulse are calculated and connected to obtain global/local signal pulses.

라고 표시된 펄스의 상승 신호의 샘플의 평균 개수를 결정한다.

는 평균 상승 길이에 해당한다.Initially, the motion artifact correction engine 140

Determine the average number of samples of the rising signal of the pulse marked.

Is the average length of ascent.

라고 표시된 펄스의 하강 신호의 샘플의 평균 개수를 결정한다.

는 평균 하강 길이에 해당한다.In addition, the motion artifact correction engine 140

Determine the average number of samples of the falling signal of the pulse marked.

Is the average descent length.

샘플을 갖도록 각 상승 펄스를 리샘플링하고,

샘플을 갖도록 각 하강 펄스를 리샘플링한다.In addition, the motion artifact correction engine 140

Resample each rising pulse to have a sample,

Each falling pulse is resampled to have a sample.

Then, the k-th pulse is expressed by Equation 3 below.

[Equation 3]

If L pulses are present in the signal, the entire signal is expressed by Equation (4).

[Equation 4]

In addition, the average rising pulse and the average falling pulse are respectively expressed by Equation (5).

[Equation 5]

와

는 각각 평균 상승 펄스의 ith 샘플값과 평균 하강 펄스의 ith 샘플값을 나타낸다.here,

Wow

Denotes the ith sample value of the average rising pulse and the ith sample value of the average falling pulse, respectively.

Therefore, the global signal pulse corresponds to Equation (6).

[Equation 6]

에 대해, 모션 아티팩트 정정 엔진(140)은 수학식7에 따라 다음의 로컬 신호 펄스로 지칭되는 M 펄스 윈도우 평균을 계산한다.Each k th pulse

With respect to, the motion artifact correction engine 140 calculates the M pulse window average referred to as the following local signal pulse according to Equation (7).

[Equation 7]

The local signal pulse corresponds to the average of the Kth pulse, (M/2) left and (M/2) right, and the surrounding M pulses. In addition, for the Kth pulse, a differential weighting scheme may be applied to give (but is not limited to) more weight to a pulse closer to the Kth pulse, instead of equally weighting the pulses in the window. .

The intermediate rectified K-th pulse corresponds to a weighted sum of the global signal pulse and the K-th local signal pulse according to Equation (8).

[Equation 8]

When each pulse is corrected, the motion artifact correction engine 140 resamples each pulse to the actual number of samples of the drift-drived signal. This process is performed in order not to lose vital cycle information for each pulse.

FIG. 11 is a graph showing the original signal rectified and the rectified PPG signal when α=0.3 and M=3. Right is a snapshot of the signal section on the left. Perturbation in the drawing has been corrected.

FIG. 12 is a graph showing the original signal rectified and the rectified PPG signal when α = 0.5 and M = 5. Right is a snapshot of the signal section on the left. The pulse is uniform compared to Fig. 10.

Here, α and M operate as hyperparameters. The larger the values of α and M, the more uniform the pulse. The optimal values of α and M can be hypertuned through cross validation, or can be set by integrating prior knowledge of the signal.

13 is a graph showing characteristics of a PPG signal for defining a quality evaluation metric.

To determine the performance of the proposed method, the quality of the processed signal must be evaluated. The quality of the pre-processing data can be evaluated using a) Pulse amplitude variance and b) Trough depth variance.

a) Pulse amplitude dispersion

Ⅰ. The pulse amplitude is defined as the height of the pulse peak of the previous trough.

Ⅱ. Pulse amplitude is calculated by identifying peaks and troughs in the PPG signal.

Ⅲ. The difference between each peak and the previous trough is calculated.

Ⅳ. The variance of each difference is considered the pulse amplitude variance.

b) Trough depth dispersion

Ⅰ. The trough depth is defined as the absolute difference between two successive troughs.

Ⅱ. The trough depth is calculated by identifying the trough between peaks from the signal.

Ⅲ. The absolute difference of each trough identified is calculated.

Ⅳ. The variance of each difference is considered the trough depth variance.

The change in blood volume is not expected to change rapidly in 1 minute, so the PPG waveform change will be similar. This means that pulse amplitude variance and trough depth variance should be reduced for preprocessed data.

The above embodiments can be implemented using at least one software program executed by at least one hardware device and performing a network management function to control configurations.

On the other hand, the present embodiments can be implemented in computer readable code on a computer readable recording medium. The computer-readable recording medium includes all kinds of recording devices in which data readable by a computer system is stored.

Examples of computer-readable recording media include ROM, RAM, CD-ROM, magnetic tapes, floppy disks, optical data storage devices, etc., and also implemented in the form of carrier waves (for example, transmission via the Internet). Includes. In addition, the computer-readable recording medium can be distributed over network coupled computer systems so that the computer-readable code is stored and executed in a distributed fashion. In addition, functional programs, codes, and code segments for implementing the embodiments can be easily inferred by programmers in the technical field to which the present invention pertains.

Those of ordinary skill in the art to which the present disclosure pertains will appreciate that it may be implemented in other specific forms without changing the disclosed technical spirit or essential features. Therefore, it should be understood that the embodiments described above are illustrative in all respects and not restrictive.

100: PPG signal pre-processing device 110: PPG signal receiver
120: moving average filter 130: drift removal engine
140: motion artifact correction engine 150: vital signal prediction engine
160: memory 170: communication unit
180: processor
140a: intermediate rectified PPG signal acquisition unit
140b: pulse resampling unit
140c: connection

Claims (20)

Receiving a PPG (Photoplethysmography) signal including a pulse;
Smoothing the PPG signal using a moving average filter to obtain a smoothed PPG signal;
Removing drift from the smoothed PPG signal using a fitted polynomial function to obtain a drift-removed PPG signal; And
And correcting the motion artifact of the drift-removed PPG signal by replacing the pulse with a combination of a global signal pulse and a local signal pulse to obtain a rectified PPG signal. According to claim 1,
Extracting features from the rectified PPG signal; And
And predicting a vital signal value based on the extracted feature. According to claim 1,
Using the fitted polynomial function to remove drift from the smoothed PPG signal to obtain a drift-removed PPG signal,
Determining the fitted polynomial function as a polynomial approximation of the predefined smoothed PPG signal; And
And subtracting a fitted polynomial function from the smoothed PPG signal to obtain the drift-removed PPG signal. According to claim 1,
The global signal pulse,
Dividing each pulse of the drift-removed PPG signal into rising and falling pulses based on troughs and peaks of the drift-removed PPG signal;
Obtaining an average rise length as an average of the number of samples in each rise pulse;
Obtaining an average falling length as an average of the number of samples in each falling pulse;
Resampling each rising pulse to the average rising length using a linear interpolation method to obtain a resampled rising pulse;
Resampling each falling pulse to the average falling length using a linear interpolation method to obtain a resampled falling pulse;
Obtaining an average rising pulse as an average of the resampled rising pulses;
Obtaining an average falling pulse as an average of the resampled falling pulses; And
PPG signal pre-processing method characterized in that it is determined by the step of connecting the average rising pulse and the average falling pulse. According to claim 1,
The local signal pulse,
Separating a predefined number of local pulses of the drift-removed PPG signal around the pulse to be rectified;
Dividing each local pulse into a local rising pulse and a local falling pulse;
Resampling each local rise pulse to an average rise length using a linear interpolation method to obtain a resampled local rise pulse;
Resampling each local falling pulse to an average falling length using a linear interpolation method to obtain a resampled local falling pulse;
Obtaining an average local rising pulse as an average of the resampled local rising pulses and obtaining an average local falling pulse as an average of the resampled local falling pulses; And
PPG signal pre-processing method characterized in that it is determined by the step of connecting the average local rising pulse and the average local falling pulse. According to claim 1,
Correcting a motion artifact of the drift-removed PPG signal to obtain a rectified PPG signal includes:
Obtaining an intermediate rectified PPG signal by replacing each local signal pulse with a convex combination of the local signal pulse and the global signal pulse; And
And resampling each pulse of the intermediate rectified PPG signal with the number of samples in the corresponding pulse of the drift-removed PPG signal. The method of claim 6,
Resampling each pulse of the intermediate rectified PPG signal with the number of samples in the corresponding pulse of the drift-removed PPG signal includes:
Dividing each pulse of the intermediate rectified PPG signal into a rising pulse whose number of samples corresponds to an average rising length and a falling pulse whose number of samples corresponds to an average falling length;
Resampling each rising pulse with the number of samples of the rising pulse of the corresponding pulse in the drift-removed PPG signal to obtain a rectified rising pulse;
Resampling each falling pulse with the number of samples of the falling pulse of the corresponding pulse in the drift-removed PPG signal to obtain a rectified falling pulse; And
And a step of connecting each rectified rising pulse with a corresponding rectified falling pulse to obtain the rectified PPG signal. According to claim 4,
The trough and peak (peak) is a PPG signal pre-processing method indicating a sample index corresponding to the trough and peak of the smoothed PPG signal. According to claim 1,
The pulse to be replaced is determined by adjusting peaks and troughs such that peaks occurring before and after the first trough are removed, and as a result, only the pulses starting at the trough and ending at the trough are rectified. According to claim 2,
The feature is a PPG signal including at least one of Kaiser Teager Energy (KTE) features, Spectral Entropy features and Spectral Energy Logarithmic features. Pretreatment method. A PPG signal receiver for receiving a PPG signal including pulses;
A moving average filter that smoothes the PPG signal to obtain a smoothed PPG signal;
A drift removal engine that removes drift from the smoothed PPG signal using a fitted polynomial function to obtain a drift-removed PPG signal;
And a motion artifact correction engine that corrects the motion artifact of the drift-removed PPG signal by replacing the pulse with a combination of a global signal pulse and a local signal pulse to obtain a rectified PPG signal. The method of claim 11,
PPG signal pre-processing apparatus further comprising a vital signal prediction engine that extracts features from the rectified PPG signal and predicts a vital signal based on the extracted features. The method of claim 11,
The drift removal engine
A PPG signal pre-processing device for determining the fitted polynomial function as a polynomial approximation of a predefined smoothed PPG signal and calculating a fitted polynomial function from the smoothed PPG signal to obtain the drift-removed PPG signal. The method of claim 11,
The global signal pulse
Each pulse of the drift-removed PPG signal is divided into rising and falling pulses based on the trough and peak of the drift-removed PPG signal, and the average rising length as the average of the number of samples in each rising pulse To obtain the average falling length as an average of the number of samples in each falling pulse, and to obtain a resampled rising pulse, each rising pulse is resampled to the average rising length using a linear interpolation method, and the resampled falling pulse is obtained. To obtain, each falling pulse is resampled to the average falling length using a linear interpolation method, an average rising pulse is obtained as an average of the resampled rising pulses, and an average falling pulse is obtained as an average of the resampled falling pulses, PPG signal pre-processing device determined by connecting the average rising pulse and the average falling pulse. The method of claim 11,
The local signal pulse
Linear interpolation is used to separate a predefined number of local pulses of the drift-removed PPG signal around the pulse to be rectified, divide each local pulse into a local rising pulse and a local falling pulse, and obtain a resampled local rising pulse. Each local rising pulse is resampled to an average rising length, and each local falling pulse is resampled to an average falling length using a linear interpolation method to obtain a resampled local falling pulse, and averaged as an average of the resampled local rising pulses. PPG signal pre-processing apparatus characterized by determining by acquiring a local rising pulse, obtaining an average local falling pulse as an average of the resampled local falling pulses, and connecting the average local rising pulse and the average local falling pulse. . The method of claim 11,
The motion artifact correction engine
Intermediate rectified PPG signal is obtained by replacing each local signal pulse with a convex combination of the local signal pulse and the global signal pulse,
PPG signal pre-processing device characterized in that for re-sampling each pulse of the intermediate rectified PPG signal with the number of samples in each pulse of the drift-removed PPG signal. The method of claim 16,
The motion artifact correction engine resampling each pulse of the intermediate rectified PPG signal with the number of samples in each pulse of the drift-removed PPG signal,
Each pulse of the intermediate rectified PPG signal is divided into a rising pulse whose number of samples corresponds to an average rising length and a falling pulse whose number of samples corresponds to an average falling length, and the drift-removed PPG is obtained to obtain a rectified rising pulse. Each rising pulse is resampled by the number of samples of the rising pulse of the corresponding pulse in the signal, and each falling pulse is taken as the number of samples of the falling pulse of the corresponding pulse in the drift-removed PPG signal to obtain a rectified falling pulse. A PPG signal pre-processing apparatus characterized by resampling and connecting each rectified rising pulse with a corresponding rectified falling pulse to obtain the rectified PPG signal. The method of claim 14,
The trough (trough) and peak (peak) PPG signal pre-processing apparatus characterized in that it represents a sample index corresponding to the trough and peak of the smoothed PPG signal. The method of claim 11,
The pulse to be replaced is determined by adjusting the peak and trough so that peaks occurring before the first trough and after the last trough are removed, and as a result, only the pulses starting at the trough and ending at the trough are rectified. The method of claim 12,
The feature is a PPG signal including at least one of Kaiser Teager Energy (KTE) features, Spectral Entropy features and Spectral Energy Logarithmic features. Pretreatment device.

Images