A method and system for anti-motion interference and signal enhancement of PPG signals

Abstract

The present invention discloses a method for anti-motion interference and signal enhancement for PPG signals. The method collects accelerometer signals and photoplethysmography (PPG) signals, combines the accelerometer signals to determine the period of motion, and then performs noise reduction processing on the PPG signals to accurately extract the PPG signals without motion interference. The enhanced PPG signals can be obtained by superimposing the PPG signals, thereby maximizing the signal-to-noise ratio of the PPG signals. The time T is subtracted from the observed value of the current PPG signal. N The PPG signal observation value before the first time is obtained to obtain a PPG signal with motion interference removed, solving the interference removal problem in the case of strong motion. Moreover, when the motion cycle is not close to the pulse wave cycle, it does not affect the identification of the P peak. The enhanced PPG signal is obtained by directly adding the PPG signals of n time-delayed cycles, which effectively improves the signal-to-noise ratio of the PPG signal while also maintaining good robustness.

Description

Technical Field

[0001] The present invention relates to the field of biomedical signal processing technology, and in particular to a method and system for anti-motion interference and signal enhancement of PPG signals. Background Art

[0002] Photoplethysmography (PPE) is a noninvasive method for detecting blood volume changes in living tissue using optoelectronics. It incorporates multiple signals, including heart rate and respiration, and can be collected at fingertips, wrists, and other locations, facilitating integration into wearable devices. Ideally, pulse signals are considered equivalent to electrocardiogram (ECG) signals. However, in practice, pulse wave signals are often subject to significant motion interference due to the complexity of the acquisition environment and conditions. This makes accurate recovery of a pure pulse signal, and even extraction of the pulse rate, difficult. Therefore, removing motion interference has always been a key and challenging aspect of pulse research.

[0003] During intense exercise, the pulse wave signal is often completely submerged in various interfering signals. Without prior knowledge or auxiliary information, signal recovery is typically complex and difficult. Lin CH et al. converted the pulse wave and acceleration signals into the frequency domain, defining them as the first and second frequency domain information, respectively. Denoising parameters were determined based on the maximum spectral peak of the second frequency domain information to denoise the first frequency domain information. Heart rate was then calculated based on the maximum spectral peak of the denoised first frequency domain information. Huang Haicheng, without prior information, recovered PPG (photoplethysmography) signals through a complex signal processing process. A denoising process based on singular value decomposition (SVD) was proposed, suitable for both periodic and non-periodic motion. Combining wavelet transform and EMD, the EMD was used to isolate the pulse signal components from the wavelet coefficients containing the pulse signal. A variable-step-size least mean square adaptive filtering algorithm was employed, combining existing adaptive algorithms. ICA was performed on dual-channel PPG data, with IC selected as the denoised PPG signal based on kurtosis and acceleration frequency. Finally, SVD reconstruction was performed.

[0004] Some researchers have used clean ECG or pulse wave signals to aid model training, thereby recovering the pulse wave signal or extracting heart rate. However, most existing algorithms address denoising under static or daily activity conditions. Denoising or removing interference from pulse wave signals during intense exercise is complex, and initial model training requires synchronized, interference-free signals. These factors limit the practical application of pulse wave analysis. Summary of the Invention

[0005] In view of this, the present invention provides a method and system for removing motion interference and signal enhancement for PPG signals, which can effectively remove motion interference of PPG signals and improve the signal-to-noise ratio of PPG signals.

[0006] A method for anti-motion interference and signal enhancement for PPG signals, comprising the following steps:

[0007] Step 1: Get the current motion period T based on the accelerometer signal N , determine T N The observed value of the PPG signal time ago;

[0008] Step 2: Current observation signal of PPG signal and time T N The observed signal of the PPG signal before is subtracted point by point to obtain the signal y without motion interference denosie (n);

[0009] Step 3: Remove the motion interference signal y den o sie (n) Determining a reference value of the time delay period T0, and then enhancing the PPG signal; obtaining a pulse rate value based on the enhanced PPG signal.

[0010] Preferably, in step 3, the inner product spectrum method is used to determine the reference value of the delay period T0, specifically: according to the number of sampling points N of a pulse wave cycle, the signal y is intercepted den o sie For the 2N sampling points in (n), take the first half of the data y T1 , y T1 The inner product of each point delay is calculated in the entire data range, and the inner product of each point constitutes the sequence signal y cov , for the sequence signal y cov When performing frequency domain analysis, the frequency domain extreme value is the pulse rate approximation, and its reciprocal is the reference value of the delay period T0.

[0011] Preferably, the PPG signal y with motion interference removed denosie (n) is convolved with the original PPG signal, and the obtained signal is used as the signal to remove motion interference.

[0012] Preferably, in step three, PPG signals of multiple time delay periods T0 are superimposed point by point to obtain an enhanced PPG signal.

[0013] Preferably, in step 3, the PPG signal y(t) is time-shifted by 0, T0 / 6, T0 / 3, T0 / 2, 2T0 / 3, and 5T0 / 6, respectively, to obtain six signals y(t-T0 / 6), y(t-T0 / 3), y(t-T0 / 2), y(t-2T0 / 3), and y(t-5T0 / 6), and the enhanced signal y(t)enhance is constructed as follows:

[0014] y(t) enhance =y(t)+y(t-T0 / 6)-y(t-T0 / 3)-y(t-T0 / 2)-y(t-2T0 / 3)+y(t-5T0 / 6).

[0015] Preferably, in step 3, the method for obtaining the pulse rate value according to the enhanced PPG signal is:

[0016] The pulse wave signal period can be obtained by directly searching for the P peak in the time domain of the enhanced signal, or by performing spectrum transformation and finding the maximum spectrum peak within a reasonable range in the frequency domain. The corresponding frequency value is a more accurate pulse rate value.

[0017] The present invention has the following beneficial effects:

[0018] (1) A method for anti-motion interference and signal enhancement for PPG signals, which collects accelerometer signals and photoplethysmography (PPG) signals, combines the accelerometer signals to determine the period of motion, and then performs noise reduction processing on the PPG signals to accurately extract the PPG signals that have received motion interference; by superimposing the PPG signals, an enhanced PPG signal can be obtained, thereby maximizing the signal-to-noise ratio of the PPG signal.

[0019] (2) Subtract the time T from the current PPG signal observation value N The PPG signal is obtained by removing the motion interference from the observation value of the previous PPG signal, solving the problem of removing interference in strong motion. In addition, when the motion cycle is not close to the pulse wave cycle, it does not affect the identification of the P peak.

[0020] (3) The PPG signals of n time-delayed periods are directly added together to obtain an enhanced PPG signal, which effectively improves the signal-to-noise ratio of the PPG signal while also having good robustness. By performing a 2π phase shift on the PPG signal of one time-delayed period, and decomposing the 2π phase shift into several equal parts, the PPG signal is time-shifted according to the decomposed phase shift to obtain an enhanced PPG signal. The shift superposition and cancellation can be completed within one period, which can maximize the improvement of the signal-to-noise ratio.

[0021] (4) Convolution of the PPG signal with the original PPG signal after motion interference removal allows the pulse signal to acquire some of the energy of the motion interference signal, increasing its amplitude and facilitating decision-making. Even if the main frequency of the motion signal is close to that of the PPG signal, it can be restored through convolution, with good results. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 The figure is a flow chart of a method for anti-motion interference and signal enhancement for PPG signals of the present invention.

[0023] Figure 2 This is a comparison diagram of the signals before and after removing motion interference, i.e., denoising, according to the present invention. DETAILED DESCRIPTION

[0024] A method and system for anti-motion interference and signal enhancement of PPG signals, which collects accelerometer signals and photoplethysmography (PPG) signals, and obtains the current motion period T according to the accelerometer signals. N ; According to the current motion period T N The current PPG signal is subjected to noise reduction processing to obtain a PPG signal with motion interference removed; the delay period of the PPG signal with motion interference removed is obtained, and the PPG signal is superimposed according to the delay period to obtain an enhanced PPG signal.

[0025] The present invention is described in detail below with reference to the accompanying drawings and embodiments.

[0026] like Figure 1 As shown, the present invention provides a method for anti-motion interference and signal enhancement for PPG signals, comprising:

[0027] Step 1: Synchronously collect photoplethysmography (PPG) signals, acceleration signals, and ECG signals; preprocess the collected PPG signals to remove high-frequency noise and baseline drift through bandpass filtering and baseline fitting.

[0028] Get the current motion period T based on the accelerometer signal N Using the ECG signal as the standard reference signal for heart rate, the PPG signal is preprocessed in combination with the accelerometer signal to remove motion interference, thereby accurately extracting the pulse rate signal. To reduce the amount of computation, the PPG signal is divided into two states: static and dynamic, based on the acceleration signal. The algorithm is only executed on the signal in the dynamic state. Specifically:

[0029] In motion, the pulse wave signal is interfered by the motion signal. There are interference signals with the same frequency as the motion frequency and its harmonics in the pulse wave signal. Assuming that the observed signal is y(n), the true pulse wave signal is x(n), and the noise is N(n), where the noise includes random noise w(n) and motion noise s(n), then the PPG signal model can be expressed as:

[0030] y(n)=x(n)+N(n)=x(n)+w(n)+s(n) (1)

[0031] Where n represents the number of points of the discrete signal. It should be noted that the motion noise here is not completely equivalent to the accelerometer signal, so the accelerometer signal cannot be directly calculated with the PPG signal or used as a reference signal for the adaptive filter to remove interference. However, in the case of regular exercise (such as running, skipping, etc.), the exercise cycle can be approximately considered unchanged in a short period of time. The main frequency range (or exercise cycle) of the current exercise can be obtained based on the accelerometer signal, so the main motion interference in the PPG signal must also be concentrated in this frequency range. Calculate the main frequency f of the current exercise based on the accelerometer signal N (or period T N ), then T N The observed signal before time is:

[0032] y(nT N )=x(nT N )+w(nT N )+s(nT N ) (2)

[0033] Step 2: According to the current motion period T N The current PPG signal is subjected to noise reduction processing to obtain a PPG signal with motion interference removed.

[0034] For the current observation signal and time T N The previous observation signal is subtracted point by point, which can eliminate the motion interference signal to a certain extent and obtain the preliminary motion interference-free signal y denosie (n):

[0035]

[0036] Assuming that the random noise w(n) is relatively weak, it can be further approximated as:

[0037] y denoise (n)≈x(n)-x(nT N )+s(n)-s(nT N )(4)

[0038] From the perspective of time domain, since the period of the motion signal is approximately TN ,so

[0039] s(n)-s(nT N )≈0(5)

[0040] y denoise ≈x(n)-x(nT N )(6)

[0041] Analyzing from the frequency domain, ignoring the influence of other noise (mainly w(n)), and performing Fourier transform on the differenced signal, we get:

[0042]

[0043] Among them, j is the basic unit of imaginary number, frequency ω=2πf=2πT, when the period is T N When ω=2πf N =2πT N , the frequency domain coefficient is 0, that is, the frequency (T N The signal components of the period) are removed.

[0044] Therefore, this method can effectively remove motion interference. Although the waveform of the pulse wave signal changes, it does not affect the identification of the P peak, that is, the peak value of the PPG signal, when the motion cycle is not close to the pulse wave cycle. Figure 2 FIG. 1 is a comparison diagram of signals before and after removing motion interference, i.e., denoising, according to the present invention.

[0045] Step 3: Obtain the delay period of the PPG signal after removing motion interference, and superimpose the PPG signal according to the delay period to obtain an enhanced PPG signal.

[0046] The inner spectral product method is used to obtain the approximate period of the PPG signal after removing motion interference, and the approximate period is used as the delay period.

[0047] In the above steps, we roughly assume that the noise consists solely of motion interference. However, other types of noise are also present. If the pulse wave component is partially weakened when removing motion interference, the interference from other types of signals will increase, affecting frequency domain decision making. Therefore, it is necessary to strengthen the pulse wave component to facilitate frequency domain threshold determination.

[0048] The enhanced PPG signal is obtained by: transforming the PPG signal y of a time delay period into denosie The enhanced PPG signal can be obtained by directly adding (n), where a is a positive integer, specifically:

[0049] 1) Use inner product spectrum method to determine the frequency range:

[0050] After removing motion interference, the observed signal contains a pure pulse wave signal and random noise. The pulse wave also has an approximate periodicity. By calculating the inner product of time delays, we can find the maximum correlation delay points, which can be further converted into the pulse wave period and pulse rate.

[0051] The human heart rate (pulse rate, bpm) ranges roughly from 60 to 180 bpm. Assuming the sampling rate is fs, the number of sampling points in one pulse wave cycle is:

[0052]

[0053] Try to intercept data y that is greater than two cycles (2N) T , take the first half of the data y T1 , y T1 Throughout the y T The inner product of each point delay is calculated within the range, and the inner product of each point constitutes the sequence signal y cov ,y cov An extreme value will be obtained near the actual period of the pulse wave delay. When performing frequency domain analysis, the frequency domain extreme value is the approximate value of the pulse rate.

[0054] In addition, since the heart rate value will not change suddenly, the historical heart rate value can also be used as a reference for the current frequency domain decision range, that is, the reciprocal of the pulse rate approximation is calculated, which is the reference value of the delay period T0.

[0055] 2) Periodic superposition and enhancement of pulse wave signal:

[0056] According to the approximate pulse rate value, it can be converted into a time delay period for pulse wave enhancement. Taking the superposition of three time delay periods as an example, the enhanced pulse wave signal y can be obtained. enhance As follows:

[0057] y enhance =x(n)+x(n-T0)+x(n-2T0)+w(n)+w(n-T0)+w(n-2T0) (9)

[0058] Following the derivation in the previous section, we can see that the pulse wave component is enhanced by approximately a factor of 3, while the other components are enhanced by a factor between 0 and 3, improving the signal-to-noise ratio. Experimental verification shows that fluctuations in the estimated period within a small range around the true period have little impact on the results, demonstrating the robustness of the algorithm.

[0059] 3) In addition to using periodic superposition to enhance the pulse wave signal, the present invention also designs another signal enhancement method, specifically: performing a 2π phase shift on the PPG signal of a time delay period, decomposing the 2π phase shift into several equal parts, and time-shifting the PPG signal according to the decomposed 2π phase shift to obtain an enhanced PPG signal.

[0060] In the above-mentioned time-shift process, if you want to achieve a better denoising effect, you generally need data from several consecutive cycles. However, PPG signals and motion signals are not completely periodic signals. Therefore, the longer the duration, the greater the deviation, and the higher the requirements for data quality.

[0061] Therefore, it is considered to use only the data within two cycles to perform time delay difference or summation to eliminate motion interference or enhance the pulse wave signal.

[0062] like According to the time delay property and Euler formula, we can get:

[0063]

[0064] When t0=T0 (i.e., the overall delay of the cycle), the signal with a frequency of ω0 (i.e., T=T0) has a coefficient of a k Therefore, theoretically, the signal components of a specified frequency can be enhanced or completely removed by superposition or subtraction operations. If you want to enhance it n times, you must have n cycles of data.

[0065] According to Euler's formula:

[0066]

[0067] Delaying one period T0 results in a 2π phase shift on the ω0 component in the frequency domain. By breaking down this 2π phase shift and canceling out the imaginary part, we can enhance or weaken specific frequency components. This is explained below using a 2π phase shift divided into six equal parts.

[0068] If you want to enhance the ω0 component, do the following decomposition and time-shift the original signal by 0, T0 / 6, T0 / 3, T0 / 2, 2T0 / 3, and 5T0 / 6 to obtain 6 signals.

[0069] Construct enhanced signal y(t) enhance As follows:

[0070] y(t) enhance =y(t)+y(t-T0 / 6)-y(t-T0 / 3)-y(t-T0 / 2)-y(t-2T0 / 3)+y(t-5T0 / 6)(12)

[0071] According to the time-shift property, we can convert it to the frequency domain and calculate it when ω=ω0:

[0072]

[0073] The ω0 frequency component of the original signal is enhanced by 4 times, while other frequency components are randomly enhanced or weakened slightly, achieving the effect of enhancing the target signal.

[0074] 4) After obtaining the enhanced signal using method 2) or 3), the pulse wave signal period can be obtained by directly searching for the P peak in the time domain, or performing a spectrum transformation to find the maximum spectrum peak within a reasonable range in the frequency domain. The corresponding frequency value is a more accurate pulse rate (heart rate) value.

[0075] It should be noted that the method used in step 2 to remove motion interference is suitable for situations where the motion cycle and pulse wave cycle are not close. However, when the motion and pulse rate are too close, removing the motion signal will also significantly weaken the main component of the pulse wave. However, in this case, the pulse wave is generally not weakened to zero. In this case, convolution of the PPG signal with the original PPG signal can make the pulse signal acquire some of the energy of the motion interference signal, increase the amplitude, and obtain a PPG signal with restored pulse rate components. If the motion cycle and pulse wave cycle completely overlap, this method can also approximately restore the pulse rate component.

[0076] According to the above-mentioned method for anti-motion interference and signal enhancement of PPG signals, the present invention also provides an anti-motion interference and signal enhancement system for PPG signals, including: a signal acquisition module, a noise reduction processing module and a signal enhancement module.

[0077] The signal acquisition module is used to collect accelerometer signals and photoplethysmography (PPG) signals, and obtain the current motion period T according to the accelerometer signals. N .

[0078] The noise reduction processing module is used to calculate the period T of the current motion. N The current PPG signal is subjected to noise reduction processing to obtain a PPG signal with motion interference removed and sent to the signal enhancement module.

[0079] The signal enhancement module is used to obtain the delay period of the PPG signal after removing motion interference. The enhanced PPG signal can be obtained by superimposing the PPG signal according to the delay period.

[0080] The noise reduction process in the noise reduction processing module is as follows: the observation value of the current PPG signal is subtracted from the time T N The observation value of the previous PPG signal is obtained to obtain the PPG signal with motion interference removed.

[0081] The method for obtaining the time delay period of the PPG signal with motion interference removed in the signal enhancement module is: using the inner spectral product method to obtain the approximate period of the PPG signal with motion interference removed, and using the approximate period as the time delay period.

[0082] In the signal enhancement module, the enhanced PPG signal is obtained by directly adding the PPG signals of n time-delayed periods, where n is a positive integer.

[0083] Alternatively, a 2π phase shift is performed on a PPG signal of a time delay period, and the 2π phase shift is decomposed into several equal parts, and the PPG signal is time-shifted according to the decomposed 2π phase shift to obtain an enhanced PPG signal.

[0084] The system of the present invention also includes a convolution recovery module; the convolution recovery module is used to perform a convolution operation on the PPG signal from which motion interference has been removed and the original PPG signal to obtain a PPG signal with restored pulse rate components.

[0085] This paper combines the principles of frequency-domain analysis to propose a method for time-delay superposition denoising and signal enhancement. Based on accelerometer motion information, the dominant frequency of the current motion is determined. Motion interference is removed by periodic time-delay subtraction. The inner product spectrum method is used to estimate the pulse wave period. Finally, the pulse wave is enhanced by periodic time-delay addition, improving the signal-to-noise ratio and enabling effective P-peak identification. This method is simple, effective, computationally efficient, and largely unconstrained by specific hardware.

[0086] By introducing triaxial acceleration signals and adding prior knowledge of motion during PPG preprocessing, a lightweight denoising algorithm is proposed to address the interference removal problem in the presence of strong motion. Although the periods of strong motion and pulse wave signals are not constant, they can be assumed to have an approximate period over short periods of time. Based on this short-term approximate invariance of their periods, time shifting and addition / subtraction are used to enhance the pulse wave signal and attenuate the motion signal.

[0087] The method of the present invention is lightweight, computationally efficient, and incorporates prior knowledge of the accelerometer. It is highly effective in removing artifacts during certain intense yet regular movements. It can also effectively remove random noise through superposition. It can be used as a standalone denoising method or in combination with other denoising methods.

[0088] In summary, the above are only preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for anti-motion interference and signal enhancement of PPG signals, characterized in that: The steps include: Step 1: Get the current motion period T based on the accelerometer signal N , determine T N The observed value of the PPG signal time ago; Step 2: Current observation signal of PPG signal and time T N The observed signal of the PPG signal before is subtracted point by point to obtain the signal y without motion interference denosie (n); Step 3: Remove the motion interference signal y denosie (n) determining a reference value of the time delay period T0, and then enhancing the PPG signal; obtaining a pulse rate value based on the enhanced PPG signal; In step 3, the inner product spectrum method is used to determine the reference value of the delay period T0, specifically: according to the number of sampling points N of a pulse wave cycle, the signal y is intercepted. denosie For the 2N sampling points in (n), take the first half of the data y T1 , y T1 The inner product of each point delay is calculated in the entire data range, and the inner product of each point constitutes the sequence signal y cov , for the sequence signal y cov Perform frequency domain analysis. The frequency domain extreme value is the pulse rate approximation. Its reciprocal is the reference value of the delay period T0. In step three, the method for obtaining the pulse rate value based on the enhanced PPG signal is: directly searching for the P peak in the time domain of the enhanced signal to obtain the pulse wave signal period, or performing a spectrum transformation to find the maximum spectrum peak within a reasonable range in the frequency domain, and the corresponding frequency value is a more accurate pulse rate value.

2. The method for anti-motion interference and signal enhancement for PPG signals according to claim 1, wherein: The PPG signal y denosie (n) is convolved with the original PPG signal, and the obtained signal is used as the signal to remove motion interference.

3. The method for anti-motion interference and signal enhancement for PPG signals according to claim 1, wherein: In the step 3, the PPG signal y of multiple delay periods T0 is used. denosie (n) Perform point-by-point superposition to obtain an enhanced PPG signal.

4. The method for anti-motion interference and signal enhancement for PPG signals according to claim 1, wherein: In the step 3, the PPG signal y denosie (n) Time shift 0, T0 / 6, T0 / 3, T0 / 2, 2T0 / 3, 5T0 / 6 respectively, and obtain 6 signals y(t), y(t-T0 / 6), y(t-T0 / 3), y(t-T0 / 2), y(t-2T0 / 3) and y(t-5T0 / 6) to construct the enhanced signal y(t) enhance As follows: y(t) enhance =y(t)+y(t-T0 / 6)-y(t-T0 / 3)-y(t-T0 / 2)-y(t-2T0 / 3)+y(t-5T0 / 6)。

Images