CN108334868A - A kind of pulse analysis method based on PPG signals and image enhancement - Google Patents

Abstract

The present invention relates to a kind of pulse analysis method based on PPG (photoplethysmography) signal and image enhancement, comprising the following steps: collecting face video images; enhancing the video images to eliminate the influence of low light; Channel separation; use ICA algorithm to process; G channel correlation analysis to obtain PPG pulse wave signal; denoise the noisy PPG pulse wave signal; find each characteristic point of PPG pulse wave signal; input the characteristic value of pulse wave into KNN algorithm , output the classification result of pulse condition; compare the classification result with the standard database and judge the disease. A pulse condition analysis method based on PPG signal and image enhancement method provided by the present invention combines the pulse condition theory of traditional Chinese medicine with computer technology, which can effectively realize objective and quantitative pulse condition analysis, and avoids the low Effect of lighting environment on pulse condition analysis.

Description

A pulse condition analysis method based on PPG signal and image enhancement

technical field

The invention relates to the field of image processing and signal processing, in particular to a pulse condition analysis method based on PPG signal and image enhancement.

Background technique

Pulse condition is the situation and dynamics of the pulse, and it is the main basis for judging human diseases in ancient China. Today's medicine mainly has two ways to detect the pulse of the human body. One is to feel the change of the pulse by putting the finger on the radial artery of the human body, that is, the traditional Chinese medicine method of feeling the pulse, and the other is to use the sensor instead of the finger to sense the pulse change of the radial artery. The former requires professional Chinese medicine personnel to detect, while the latter usually puts appropriate sensors on the measured part, converts the pulsation of the pulse into an electrical signal, and then inputs it into an amplifier circuit, processes the weak signal with a computer, and then processes the pulse wave. Analysis and diagnosis. This method requires complex instruments and equipment and is expensive, and it is inconvenient to carry and use at home. This paper attempts to obtain the pulse characteristics of the human body by analyzing the face video information collected by ordinary computer cameras, and then realize the detection of human health status using ordinary computer systems.

At present, the focus of technical attention is more inclined to the improvement of pulse signal acquisition instruments and methods, or to conduct in-depth analysis and research on a certain link of pulse signal processing and analysis, lacking certain coherence and systematic pulse detection and analysis methods, or It is because some factors that will affect the pulse detection are not properly dealt with, such as light. Therefore, it is extremely necessary to propose a complete, systematic and rigorous pulse condition analysis method.

Contents of the invention

In view of the deficiencies in the above-mentioned prior art, the purpose of the present invention is to provide a pulse condition analysis method based on PPG signal and image enhancement, eliminate the influence of illumination factors on pulse condition detection, and realize stable and objective pulse condition diagnosis.

In order to achieve the above object, the pulse condition analysis method based on PPG signal and image enhancement proposed by the present invention may further comprise the steps:

The first step: collect face video images through ordinary computer cameras;

The second step: the video image is enhanced through an improved image enhancement algorithm;

The third step: separating the RGB channel of the enhanced video image;

The fourth step: ICA algorithm processing, and correlation analysis with the G channel signal to obtain the PPG pulse wave signal;

The fifth step: pulse wave signal denoising;

The sixth step: obtaining each characteristic point of the PPG pulse wave signal;

Step 7: Input the extracted pulse wave feature value into the KNN algorithm output result

The eighth step: compare the results of pulse signal classification with the standard database to judge the disease.

Further, in the first step, the described pulse condition analysis method based on PPG signal and image enhancement is characterized in that: in the first step, the camera of an ordinary computer is used to collect face video images, and the person who collects the signal needs to Sit quietly in front of the computer, with your forehead exposed, stay for a while, prohibit large body movements, and maintain a natural facial expression.

Further, in the second step, the specific flow of the video image enhancement method is as follows:

Step 1: Corrected automatic contrast stretching for each channel of the image;

Step 2: Mutual conversion of HSV-based color spaces;

Step 3: Perform piecewise linear transformation on the V component.

Further, in the second step, the modified automatic contrast stretching formula is:

in The two thresholds representing the pixel values of the low-illumination image respectively can be obtained by the following formula:

In the above formula: 0≤t _low , t _high ≤1, t _low +t _high ≤1.

Further, in the second step, the piecewise linear transformation of the V component in the HSV space is:

The pixel values of the V component are sorted in ascending order, and the sorted vector is denoted as V'=h _{M × N} (x), and the parameter for segmenting V' is set to n, and the segmented pixels The number is Set a minimum value of one pixel for each segment V' _i and the maximum The expression for linear transformation of each segment is as follows: After linear transformation, g(x)=[g ₀ (x), g ₁ (x), Λ g _n (x)] is obtained, and g(x) is reconstructed into an M×N dimensional matrix V”=|g(x, y)|, and finally convert the image from HSV space to RGB space to obtain an enhanced video image.

Further, in the fourth step, the ICA algorithm extracts the PPG pulse wave signal from the face image sequence, and separates the signal into a linear combination of statistically independent non-Gaussian signal sources, that is, from the linear mixed signal recover the basic original signal

Further, in the fifth step, the denoising of the pulse wave signal is a method combining wavelet reconstruction and empirical mode decomposition (EMD). After the wavelet transform is decomposed, the high-frequency signal outside the frequency band is filtered out, which can effectively filter out the high-frequency interference. Decomposition can highlight signal features with different characteristics. Then, after wavelet decomposition, the wavelet coefficients in the pulse wave signal frequency band are further decomposed by the empirical mode decomposition (EMD) method to obtain new wavelet coefficients.

Further, in the sixth step, the feature points of the PPG pulse wave signal include the dynamics of the pulse wave, the speed of the pulse wave rise, the speed of the pulse wave fall, the width of a cycle pulse wave, the dicrotic wave The strength, the number of pulse wave cycles in one minute.

Further, in the seventh step, when the KNN algorithm recognizes and classifies the pulse signal, each feature point of the PPG signal is used as input, and the KNN clustering algorithm of machine learning is used to predict and classify the pulse signal.

Further, in the eighth step, in the judgment of the disease, the predicted classification result of the KNN algorithm is compared with the standard database to judge the symptom represented by the pulse condition.

Compared with the existing pulse condition analysis method, the present invention has the following advantages:

1. A non-contact pulse condition analysis method is proposed, which is easy to operate and low in cost.

2. The video image enhancement method proposed eliminates the influence of uneven illumination on pulse analysis and ensures the stability of pulse information.

3. Combining wavelet and EMD to denoise the pulse wave signal, further improving the results of pulse condition analysis.

3. According to the feature points of PPG pulse wave, the KNN algorithm is used to identify and classify the pulse condition, which is easy to understand and easy to implement.

4. The proposed pulse condition analysis method combines traditional medical theory with modern computer science, and can effectively realize stable, rigorous and objective pulse condition detection and diagnosis.

Description of drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings that need to be used in the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only These are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can also be obtained according to these drawings on the premise of not paying creative labor:

Figure 1 is a characteristic diagram of the PPG pulse wave signal.

Fig. 2 is a flow chart of the method of the present invention.

Fig. 3 is a schematic diagram of implementation of the face-based video image enhancement method proposed by the present invention.

Detailed ways

The present invention will be further described in detail below in conjunction with the accompanying drawings, but the implementation and protection scope of the present invention are not limited thereto.

Photoplethysmography (PPG) is a non-invasive detection method that uses light absorption by living tissue to detect changes in blood volume. As shown in FIG. 1 , it is a characteristic map of the PPG pulse wave signal, the horizontal axis is time, and the vertical axis is the change of blood volume. Point U is the lowest point of the pulse wave signal. At this time, the heart enters the systolic stage, a large amount of blood is injected into the aorta, and the blood volume in the blood vessels of the human body increases. This is the lowest point of the waveform, that is, the beginning of blood ejection. Point Z is the apex of the signal, at which point the blood volume in the blood vessel reaches the maximum value. Point W is the local apex in the descending branch of the signal, corresponding to the maximum ejection volume after the regurgitated blood encounters the reflection of the closed aortic valve (the inflection point of some individuals is not obvious or absent). Point V, the end of systole, represents the end of ejection.

As shown in Figure 2, it is a flow chart of the method of the present invention. The face video image is collected by an ordinary computer camera, and the face video image is collected by an ordinary computer camera. The person collecting the signal needs to sit quietly in front of the computer, expose the forehead, and stay for about Minutes, large body movements are prohibited, and natural facial expressions are maintained; video images are enhanced through an improved image enhancement algorithm to prevent low-light problems from affecting pulse detection and analysis; and then the RGB channels of the enhanced video images are separated ; Subsequently, use the ICA algorithm to process, and conduct correlation analysis between the processing results and the G channel signal to obtain the PPG pulse wave signal; then combine wavelet and EMD to denoise the obtained PPG pulse wave signal; then calculate the PPG pulse wave signal Each feature point; input the extracted pulse wave feature value into the KNN algorithm output result; finally, compare the pulse signal classification result with the standard database to judge the disease.

As shown in Figure 3, the video image enhancement method includes automatic contrast stretching for correcting each channel of the image; mutual conversion of color spaces based on HSV; segmented linear transformation for V components.

The formula for automatic contrast stretching that corrects each channel of a video image is:

in The two thresholds representing the pixel values of the low-illumination image respectively can be obtained by the following formula:

The pixel value can be less than or equal to the formula greater or equal to The values of are mapped to x _min and x _max respectively, and other values are mapped to between x _min and x _max .

The piecewise linear transformation of the V component in HSV space is:

The pixel values of the V component are sorted in ascending order, and the sorted vector is denoted as V'=h _{M × N} (x), and the parameter for segmenting V' is set to n, and the segmented pixels The number is Set a minimum value of one pixel for each segment V' _i and the maximum The expression for linear transformation of each segment is After linear transformation, g(x)=[g ₀ (x), g ₁ (x), L g _n (x)], reconstruct g(x) into M×N dimensional matrix V”=|g(x, y)|, and finally convert the image from HSV space to RGB space to obtain an enhanced video image;

The ICA algorithm is a linear combination of signal sources that separates the signal into statistically independent non-Gaussian signals, that is, recovers the basic original signal from the linear mixed signal. The filter preprocessing of PPG pulse wave signal is to select the appropriate threshold H and set special rules to filter out the interference points. In order to eliminate the influence of relatively large interference, the sample points are segmented, and the maximum value of each segment is calculated, and then the average value of these maximum values is calculated, and finally a certain ratio is set as the threshold value; each of the PPG pulse wave signal The characteristic points include the strength of the pulse wave, the speed of the pulse wave rising, the speed of the pulse wave falling, the width of a cycle pulse wave, the strength of the pulse wave, the number of cycles of the pulse wave in one minute, etc. The clustering algorithm of machine learning is used to predict and classify the pulse signal; the predicted classification results are compared with the standard database, and the symptoms represented by the pulse are judged.

The above-mentioned embodiment is a preferred embodiment of the present invention, but the embodiment of the present invention is not limited by the above-mentioned embodiment, and any other changes, modifications, substitutions, combinations, Simplifications should be equivalent replacement methods, and all are included in the protection scope of the present invention.

Claims (10)

1. a pulse condition analysis method based on PPG signal and image enhancement, is characterized in that, comprises the following steps: The first step: collect face video images through ordinary computer cameras; The second step: the video image is enhanced through an improved image enhancement algorithm; The third step: separating the RGB channel of the enhanced video image; The fourth step: ICA algorithm processing, and correlation analysis with the G channel signal to obtain the PPG pulse wave signal; The fifth step: pulse wave signal denoising; The sixth step: seeking each feature point of the PPG pulse wave signal; Step 7: Input the extracted pulse wave feature value into the KNN algorithm output result The eighth step: compare the results of pulse signal classification with the standard database to judge the disease. 2. the pulse analysis method based on PPG signal and image enhancement according to claim 1, is characterized in that: in the first step, the camera of common computer is used to gather face video images, and the person who collects signals needs to sit quietly before the computer , Expose the forehead, stay for a while, prohibit large body movements, and maintain a natural facial expression. 3. the pulse condition analysis method based on PPG signal and image enhancement according to claim 1, is characterized in that: the concrete flow process of described second step is as follows: Step 1: Corrected automatic contrast stretching for each channel of the image; Step 2: Mutual conversion of HSV-based color spaces; Step 3: Perform piecewise linear transformation on the V component. 4. video image enhancement method according to claim 3, is characterized in that: the described automatic contrast stretching formula that each channel of video image is corrected is: in The two thresholds representing the pixel values of the low-illumination image respectively can be obtained by the following formula: In the above formula: 0≤t _low , t _high ≤1, t _low +t _high ≤1. 5. video image enhancement method according to claim 3, is characterized in that: described V component subsection linear transformation in HSV space is: The pixel values of the V component are sorted in ascending order, and the sorted vector is denoted as V'=h _{M × N} (x), and the parameter for segmenting V' is set to n, and the segmented pixels The number is Set a minimum value of one pixel for each segment V' _i and the maximum The expression for linear transformation of each segment is as follows: After linear transformation, g(x)=[g ₀ (x), g ₁ (x), Λ g _n (x)] is obtained, and g(x) is reconstructed into an M×N dimensional matrix V”=|g(x, y)|, and finally convert the image from HSV space to RGB space to obtain an enhanced video image. 6. a kind of pulse condition analysis method based on PPG signal and image enhancement according to claim 1, is characterized in that: in the described 4th step, described ICA algorithm is to separate signal into the signal of statistically independent non-Gaussian signal The linear combination of sources, that is, the recovery of the basic original signal from the linear mixed signal. 7. the pulse condition analysis method based on PPG signal and image enhancement according to claim 1, is characterized in that: in the described 5th step, described pulse wave signal denoising is to utilize wavelet reconstruction and empirical mode decomposition (EMD ) combination method, after the wavelet transform is decomposed, the high-frequency signal outside the frequency band is filtered out, which can effectively filter out high-frequency interference. In the process of signal processing using wavelet transform, the choice of wavelet basis function is very important , using different wavelet basis functions to decompose the signal can highlight the signal characteristics of different characteristics. Here, the Sym8 wavelet function is selected as the wavelet basis to decompose the threshold function, and then the wavelet coefficients in the frequency band of the pulse wave signal after wavelet decomposition are further processed. Decomposition by the Empirical Mode Decomposition (EMD) method to obtain new wavelet coefficients. 8. the pulse condition analysis method based on PPG signal and image enhancement according to claim 1, is characterized in that: in the described 6th step, ask each characteristic point of PPG pulse wave signal, comprise the dynamics of pulse wave, pulse wave rises The speed of the pulse wave, the speed of the pulse wave drop, the width of the pulse wave in one cycle, the strength of the pulse wave, and the number of cycles of the pulse wave in one minute. 9. a kind of pulse condition analysis method based on PPG signal and image enhancement according to claim 1, is characterized in that: in described 7th step, described KNN algorithm is to utilize the clustering algorithm of machine learning to predict pulse condition signal Classification. 10. A kind of pulse condition analysis method based on PPG signal and image enhancement according to claim 9, characterized in that: in the eighth step, the prediction classification result is compared with the standard database, and the symptom represented by the pulse condition is judged.