CN108078554A - A kind of human pulse ripple signal noise suppressing method - Google Patents

Abstract

The invention discloses a human body pulse wave signal noise suppression method. In order to clearly display the waveform form, the invention firstly performs normalization processing on the collected original pulse wave signal amplitude. Then, in order to reduce the influence of the edge effect, the period extension is carried out on both ends of the signal. Finally, a double median filter is designed for noise suppression of the human pulse wave signal: the first median filter is used to eliminate the high-frequency noise in the pulse wave signal, and the second median filter is used to estimate the noise in the signal. low-frequency noise, and then remove the estimated low-frequency noise from the suppressed high-frequency noise signal to obtain a denoised signal. The invention can suppress high-frequency noise, baseline drift and part of motion noise in the pulse wave signal of the human body, and can avoid adding additional circuits or using complicated noise elimination algorithms, and is a simple real-time noise suppression method applicable to microprocessors.

Description

A Noise Suppression Method for Human Pulse Wave Signal

technical field

The invention relates to a real-time pulse wave signal noise suppression method based on double median filtering, which is especially suitable for a portable pulse oximeter.

Background technique

Human pulse wave signal detection has been widely used in the evaluation of cardiovascular system, respiratory system and blood circulation system. The human pulse wave signal is weak and will inevitably be interfered by human movement and heavy breathing, which will affect the quality of signal acquisition and the accuracy of later calculation of clinical physiological parameters.

At present, the noise suppression algorithms at home and abroad mainly include EMD denoising method, adaptive filter denoising method, wavelet multi-resolution analysis denoising method, SVD denoising method, ICA denoising method, high-order statistical denoising method and step-by-step denoising method. Periodic Fourier series analysis, etc. In addition, wavelet denoising method can also be integrated in DSP to suppress power frequency interference, use acceleration sensor to collect signals as motion reference to eliminate motion interference, in 32-bit ARM microcontroller A real-time method based on contour analysis to detect pulse waveform segmentation and artifact detection etc.

However, the market share of traditional pulse oximeters based on ordinary microprocessors is relatively high. Considering factors such as cost, volume, real-time and implementation methods, the above methods may need to add additional hardware circuits to meet the requirements of motion interference signals. Acquisition, or requires the use of high-end microcontrollers to implement complex noise reduction algorithms, or the need to implement pulse wave signal data transmission and use the host computer to run complex algorithms, which is not suitable for the use of portable pulse oximeters. Therefore, the present invention focuses on providing a pulse wave real-time noise suppression method suitable for common microprocessors.

Contents of the invention

The object of the present invention is to provide a kind of real-time denoising method based on double median filtering for the noise in the pulse wave signal, and this method comprises the following steps:

(1) normalize the original pulse wave signal PPG to obtain the normalized pulse wave signal PPG1;

(2) Perform period extension processing on the pulse wave signal PPG 1 after the normalization processing to obtain the pulse wave signal PPG 2 after the extension processing;

(3) Carry out the first heavy median filter to the pulse wave signal PPG 2 after the continuation process, the sliding window size is W1, is used for suppressing the high-frequency noise in the pulse wave signal and obtains the pulse wave signal PPG3 that eliminates the high-frequency noise;

(4) Carry out the second heavy median filter to the pulse wave signal PPG 3 that eliminates high-frequency noise, and the sliding window size is W2, is used for estimating the low-frequency noise signal PPG 4 in the pulse wave signal;

(5) Remove the estimated low-frequency noise signal PPG 4 from the high-frequency noise-eliminated pulse wave signal PPG 3 to obtain a noise-suppressed pulse wave signal PPG 5 .

The normalization processing formula of described step (1) is as follows:

Wherein, i=1, 2, 3..., L; L is the data length of PPG 1, and max is the function of getting the maximum value, and min is the function of getting the minimum value;

In the step (2), the value of period extension length Q is 1/2 of the second sliding window size W2, and the extension formula is as follows:

In the formula, L is the data length of PPG 1, n=1, 2, 3..., 2Q+L;

The sliding window of described step (3) is 78ms, and the value of PPG3(n) makes in sequence PPG2(i) The smallest value, where Q+1≤n≤Q+L,

The sliding window of described step (4) is 8ms, and the value of PPG4(n) makes in sequence PPG3(i) The smallest value, where Q+1≤n≤Q+L,

The step (5) is specifically to use the following formula to obtain the signal PPG5 after noise suppression:

PPG5(n)=PPG3(n)-PPG4(n), Q+1≤n≤Q+L.

Beneficial effects of the present invention:

The invention can suppress high-frequency noise, baseline drift and part of motion noise in the pulse wave signal of the human body, and can avoid adding additional circuits or using complicated noise elimination algorithms, and is a simple real-time noise suppression method applicable to microprocessors.

Description of drawings

Fig. 1 is a program block diagram of the present invention.

Fig. 2 is the pulse wave signal collected and normalized under the condition of high-frequency environmental noise of the present invention.

Fig. 3 is the result after the pulse wave signal shown in Fig. 2 is suppressed by double median filter noise.

FIG. 4 is a spectrum of the pulse wave signal shown in FIG. 2 .

FIG. 5 is a spectrum of the pulse wave signal shown in FIG. 3 .

Fig. 6 is the pulse wave signal collected and normalized in the exercise state of the present invention.

Fig. 7 is the result of the pulse wave signal shown in Fig. 6 after double median filter noise suppression.

Detailed ways

In order to clearly display the waveform form, the invention normalizes the amplitude of the collected original pulse wave signal. Then, in order to reduce the influence of the edge effect, the period extension is carried out on both ends of the signal. Finally, a double median filter is designed for noise suppression of the human pulse wave signal: the first median filter is used to eliminate the high-frequency noise in the pulse wave signal, and the second median filter is used to estimate the noise in the signal. low-frequency noise, and then remove the estimated low-frequency noise from the suppressed high-frequency noise signal to obtain a denoised signal. Figure 1 is a program flow chart of the noise suppression method.

The original pulse wave signal used in the specific embodiment of the present invention is obtained through experiments. Specifically, the pulse oximeter developed by Tianjin Jingfan Technology Co., Ltd. is used to collect the pulse wave signal. The sampling frequency of the pulse oximeter is 100Hz . Using the "LongDate suction-type electromagnetic vibrating table" to collect human pulse wave signals under the condition of simulating high-frequency environmental noise, arbitrarily intercept a section of signal data with a length of 2000 sampling points, and the double median filter pulse signal noise suppression method includes the following steps :

(1) In order to clearly show the shape of the human pulse wave signal, the original pulse wave signal PPG is normalized to obtain the normalized pulse wave signal PPG 1, as shown in Figure 2, and the normalization formula is as follows:

Among them, i=1, 2, 3..., 2000; max is the function of taking the maximum value, and min is the function of taking the minimum value;

(2) In order to reduce the influence of the edge effect, period extension is performed on both ends of the normalized pulse wave signal PPG 1, and the extension length Q is selected as one-half of the second sliding window size W2, namely Q=W2/2=50, the processed signal is the pulse wave signal PPG 2 after the extension processing, then there are:

(3) select window time to be 78ms, calculate the first sliding window size W1=10 according to the instrument sampling rate of 100Hz, utilize the first heavy sliding window median filter to eliminate high-frequency noise, and the signal after processing is to eliminate high-frequency noise The value of the pulse wave signal PPG3, PPG3(n) is in the sequence PPG2(i) such that The smallest value, where 51≤n≤2050, n-5≤i≤n+4;

(4) Select the window time to be 8ms, calculate the second sliding window size W2=100, utilize the second sliding window median filter to estimate the low-frequency noise and obtain the low-frequency noise signal PPG4, and the value of PPG4(n) is the sequence PPG3 (i) such that The smallest value, where 51≤n≤2050, n-50≤i≤n+49;

(5) Remove the estimated low-frequency noise signal PPG4 from the pulse wave signal PPG3 of the high-frequency noise to obtain the signal after denoising, and obtain the pulse wave signal PPG5 after the noise suppression, as shown in Figure 3:

PPG5(n)=PPG3(n)-PPG4(n), 51≤n≤2050

In order to further quantitatively evaluate the algorithm, spectrum analysis was performed on the pulse wave signals shown in Figure 2 and Figure 3 respectively to obtain Figure 4 and Figure 5. For the convenience of observation, the coordinate axis of the frequency in the spectrum diagram was intercepted to 0-60Hz.

Since the energy of the pulse wave is usually concentrated at 0.5-10Hz, the suppression effect on low-frequency noise cannot be clearly reflected from the above spectrogram. Therefore, the human body pulse wave signal is collected while the tester is standing and walking, and a section of signal data with a length of 2000 sampling points is arbitrarily intercepted. After processing in step (1), the normalized pulse wave shown in Figure 6 is obtained. Wave signal, where the main interference is the baseline drift low-frequency noise caused by human motion, as marked by the middle curve in the figure. The pulse wave signal after step (2), step (3), step (4), and step (5) double median filter noise suppression is shown in Fig. 7 .

Claims (6)

1. a human body pulse wave signal noise suppression method is characterized in that: the method may further comprise the steps: (1) normalize the original pulse wave signal PPG to obtain the normalized pulse wave signal PPG1; (2) Perform period extension processing on the pulse wave signal PPG1 after the normalization processing to obtain the pulse wave signal PPG2 after the extension processing; (3) Carry out the first heavy median filter to the pulse wave signal PPG2 after the continuation process, the sliding window size is W1, is used for suppressing the high-frequency noise in the pulse wave signal and obtains the pulse wave signal PPG3 that eliminates the high-frequency noise; (4) Carry out second heavy median filtering to the pulse wave signal PPG3 that eliminates high-frequency noise, and the sliding window size is W2, is used for estimating the low-frequency noise signal PPG4 in the pulse wave signal; (5) Remove the estimated low-frequency noise signal PPG4 from the high-frequency noise-eliminated pulse wave signal PPG3 to obtain a noise-suppressed pulse wave signal PPG5. 2. a kind of human pulse wave signal noise suppressing method according to claim 1, is characterized in that: the normalization processing formula of described step (1) is as follows: <mrow><mi>P</mi><mi>P</mi><mi>G</mi><mn>1</mn><mrow><mo>(</mo><mi>i</mi><mo>)</mo></mrow><mo>=</mo><mfrac><mrow><mi>P</mi><mi>P</mi><mi>G</mi><mrow><mo>(</mo><mi>i</mi><mo>)</mo></mrow><mo>-</mo><mi>m</mi><mi>i</mi><mi>n</mi><mrow><mo>(</mo><mi>P</mi><mi>P</mi><mi>G</mi><mo>(</mo><mi>i</mi><mo>)</mo><mo>)</mo></mrow></mrow><mrow><mi>m</mi><mi>a</mi><mi>x</mi><mrow><mo>(</mo><mi>P</mi><mi>P</mi><mi>G</mi><mo>(</mo><mi>i</mi><mo>)</mo><mo>)</mo></mrow><mo>-</mo><mi>m</mi><mi>i</mi><mi>n</mi><mrow><mo>(</mo><mi>P</mi><mi>P</mi><mi>G</mi><mo>(</mo><mi>i</mi><mo>)</mo><mo>)</mo></mrow></mrow></mfrac></mrow> Among them, i=1, 2, 3..., L; L is the data length of PPG1, max is the function of obtaining the maximum value, and min is the function of obtaining the minimum value. 3. a kind of human body pulse wave signal noise suppressing method according to claim 1 is characterized in that: in the described step (2), the period extension length Q takes a value of 1/2 of the second sliding window size W2, The continuation formula is as follows: <mrow><mi>P</mi><mi>P</mi><mi>G</mi><mn>2</mn><mrow><mo>(</mo><mi>n</mi><mo>)</mo></mrow><mo>=</mo><mfenced open = "{" close = ""><mtable><mtr><mtd><mrow><mi>P</mi><mi>P</mi><mi>G</mi><mn>1</mn><mrow><mo>(</mo><mn>1</mn><mo>)</mo></mrow><mo>,</mo></mrow></mtd><mtd><mrow><mn>1</mn><mo>&amp;le;</mo><mi>n</mi><mo>&amp;le;</mo><mi>Q</mi></mrow></mtd></mtr><mtr><mtd><mrow><mi>P</mi><mi>P</mi><mi>G</mi><mn>1</mn><mrow><mo>(</mo><mi>n</mi><mo>-</mo><mi>Q</mi><mo>)</mo></mrow><mo>,</mo></mrow></mtd><mtd><mrow><mi>Q</mi><mo>&lt;</mo><mi>n</mi><mo>&amp;le;</mo><mi>Q</mi><mo>+</mo><mi>L</mi></mrow></mtd></mtr><mtr><mtd><mrow><mi>P</mi><mi>P</mi><mi>G</mi><mn>1</mn><mrow><mo>(</mo><mi>L</mi><mo>)</mo></mrow><mo>,</mo></mrow></mtd><mtd><mrow><mi>Q</mi><mo>+</mo><mi>L</mi><mo>&lt;</mo><mi>n</mi><mo>&amp;le;</mo><mn>2</mn><mi>Q</mi><mo>+</mo><mi>L</mi></mrow></mtd></mtr></mtable></mfenced></mrow> In the formula, L is the data length of PPG1, n=1, 2, 3..., 2Q+L. 4. a kind of human pulse wave signal noise suppressing method according to claim 1 is characterized in that: the sliding window of described step (3) is 78ms, and the value of PPG3 (n) is in sequence PPG2 (i) make The smallest value, where Q+1≤n≤Q+L, 5. a kind of human pulse wave signal noise suppressing method according to claim 1, is characterized in that: The sliding window of described step (4) is 8ms, and the value of PPG4(n) makes in sequence PPG3(i) The smallest value, where Q+1≤n≤Q+L, 6. a kind of human pulse wave signal noise suppressing method according to claim 1, is characterized in that: described step (5) is specifically to adopt following formula to obtain the signal PPG5 after noise suppressing: PPG5(n)=PPG3(n)-PPG4(n), Q+1≤n≤Q+L.