Floating pulse wave identification processing method

Abstract

The invention relates to the technical field of floating pulse waveform recognition, in particular to a floating pulse wave recognition processing method. The method comprises four core steps of data acquisition, period analysis, signal filtering and feature recognition. The method comprises the following steps: firstly, in a data acquisition stage, acquiring original pulse condition data of floating pulses by utilizing pulse diagnosis equipment; secondly, through the cycle analysis step, the cycle characteristic of the floating pulse wave is accurately determined; next, in a filtering processing link, processing the original pulse condition data by using a filtering algorithm so as to effectively eliminate irrelevant noise and highlight core feature information of floating pulses; finally, in the data recognition stage, the processed pulse condition signals are input into a preset mathematical model, key feature parameters are extracted from the pulse condition signals, feature matching is carried out, and therefore accurate recognition of the floating pulse condition is achieved. According to the method, the recognition efficiency can be remarkably improved, the diagnosis time is greatly shortened, and a feasible and efficient pulse diagnosis technical solution is provided for the field of medical diagnosis.

Description

Technical Field

[0001] The present invention relates to the field of signal processing, and in particular to a method for identifying and processing a floating pulse wave. Background Art

[0002] Floating pulse is a type of pulse in traditional Chinese medicine. It is characterized by being felt when lightly touched, slightly reduced but not empty when pressed hard, and being more than enough when lifted but less than enough when pressed. In the process of syndrome differentiation and treatment, doctors can identify the different characteristics of floating pulse and combine it with the other three diagnostic information to determine the location of the pathogenic factors (exterior syndrome) and the nature of the pathogenic factors (exterior syndrome or deficiency syndrome). For example, a floating and tight pulse often indicates that wind-cold binds the exterior, a floating and rapid pulse may mean that wind-heat invades the exterior, and a floating and slow pulse reflects different situations such as spleen deficiency and dampness trapping the spleen. Therefore, floating pulse is not only an important basis for diagnosing exogenous diseases, but also an important reference indicator for assessing the strength of physical constitution and the prognosis of the disease. However, in modern pulse diagnosis practice, the identification of floating pulse faces problems such as high subjectivity, low standardization, lagging technology updates, and many external interference factors. There are also challenges in the inheritance, education and training, scientific verification and international recognition of pulse diagnosis skills.

[0003] In recent years, the rapid progress of new materials and computer technology has greatly promoted the development of the medical field, including the research and application of traditional Chinese medicine pulse. New material technology has enhanced the accuracy and efficiency of pulse diagnosis equipment, and can accurately capture and record pulse characteristics such as floating pulse. Computer technology combined with big data analysis and machine learning algorithms can help achieve in-depth interpretation and intelligent diagnosis of pulse information. Through fine processing and pattern recognition of pulse data, the sensitivity to changes in pulse conditions such as floating pulse and the accuracy of clinical diagnosis have been greatly improved, promoting the modernization and intelligent reform of traditional Chinese medicine pulse diagnosis technology. In the article "Convolutional Network Based on Dimension Expansion and Pulse Recognition Application", Zhang Ning discussed the time scale variation of time series data, especially for the complexity of the field of traditional Chinese medicine pulse, and proposed an innovative preprocessing strategy to enhance feature extraction capabilities using CNN's dimension expansion technology. However, although dimension expansion theoretically improves feature richness, it may also be accompanied by the risk of overfitting, especially when samples are scarce, affecting the generalization ability of the model. In "Pulse Analysis and Identification Based on Recursive Graphs and Convolutional Neural Networks", Yan Jianjun demonstrated the potential of threshold-free recursive graphs in automatically extracting and classifying nonlinear features of pulse signals by combining nonlinear dynamics with VGG-16 networks, reducing the influence of human factors. However, the challenge of this method lies in parameter selection, which highlights the key role of optimized parameters in the effectiveness and reliability of research. Summary of the invention

[0004] The present invention provides a method for identifying and processing a floating pulse wave, the purpose of which is to fundamentally enhance the accuracy and standardization level of floating pulse diagnosis, and to pave a novel auxiliary diagnosis path for clinical practice of traditional Chinese medicine. The system, with its profound mathematical foundation and deep understanding of traditional Chinese medicine pulse diagnosis, has shown a wide range of application potentials in the field of medical technology. It can accurately identify the unique characteristics of floating pulses, realize rapid positioning of pulse types, explore an easy-to-operate path for the combination of traditional Chinese medicine pulse diagnosis and modern technology, and aims to promote the popularization and application of traditional Chinese medicine diagnostic technology.

[0005] A method for identifying and processing a floating pulse wave, wherein feature matching includes the pulse wave signal's pulsation start point, main wave peak point, main wave end point, and pulsation end point, which are set to (X1, Y1), (X2, Y2), (X3, Y3), and (X4, Y4) in sequence. The overall identification process is as follows: S1. Use the pulse diagnosis signal acquisition device to obtain the original pulse data under light, heavy and medium pressure; S2. Perform Butterworth low-pass filtering on the pulse data under three pressures to remove noise; S3. Get the three processed data, calculate the average main wave amplitude of all cycles for each data, and the pulse diagram appears pressure P 出 Smaller than normal pulse, P 出 ≤1.03N. If the pulse is recorded by taking the middle or deep pressure, the main wave amplitude will decrease. The pulse pressure should be less than 2.26N. The pulse pressure is correct. The average main wave amplitudes obtained under light, heavy and middle pressures are recorded as M1, M2 and M3 respectively. S4. The following further determines that if M1>M2 and M1>M3, it is determined to be a floating pulse type pulse wave; S5. Calculate the slope k of the ascending branch of the retained pulse wave signal, , remove the pulse wave signal with k<11.43; S6. Calculate the ejection period of the retained pulse wave signal, that is, the time t between the starting point of the pulse wave graph and the time value of the descending gorge, t = (X4-X1) × 0.008, and remove the pulse wave signal of t> 0.72 seconds; S7. Determine the height of the main wave peak Y2 and the height of the descending gorge Y3, according to the range set by the A / D circuit of the pulse diagnosis instrument used, and remove the pulse wave signal of Y3 / Y2>0.65; S8. After the above feature extraction and recognition, the pulse waveform that meets the conditions is extracted, which is the floating pulse wave.

[0006] Furthermore, in step S1, the pulse diagnosis device for collecting the original pulse signal is a pulse diagnosis instrument, and the pulse diagnosis instrument collects pulse wave data including: pulse beating data of the three parts of Cun, Guan and Chi under three pressures: light, heavy and medium.

[0007] Furthermore, the angle of the rising branch of the main wave is greater than 84°.

[0008] Furthermore, the ratio of the height of the descending gorge to the height of the main wave is less than 0.65.

[0009] Furthermore, in step S3, the main wave amplitudes M1, M2, and M3 are calculated as follows: S4.1 Add the main wave heights in each period to obtain a total; S4.2 Calculate the average main wave amplitude; S4.3 Calculate the average main wave amplitudes under light, heavy and medium pressures respectively, and record them as M1, M2 and M3.

[0010] In the present invention, the pulse diagnosis device can comprehensively collect the original pulse wave data of the three pulse positions of Cun, Guan and Chi to fully display the whole picture of the pulse beating. The present invention adopts the cycle extraction technology to divide the whole pulse wave original data into a series of representative independent cycle units.

[0011] On this basis, the present invention adopts a low-pass filtering processing strategy to optimize the subdivided periodic signal. This method aims to filter out the high-frequency noise components in the pulse wave data, so that the low-frequency signal reflecting the essential characteristics of the pulse is effectively highlighted, thereby enhancing the accuracy and reliability of subsequent floating pulse feature recognition.

[0012] Compared with the prior art, the present invention has the following beneficial effects: The present invention adopts a refined and effective recognition process. Compared with the complex deep learning model, the structure of this method is more concise and clear, and the execution efficiency is higher. Although the recognition process is simplified, the floating pulse recognition accuracy can still be maintained at a high level through the carefully designed algorithm and model.

[0013] In summary, the floating pulse identification method created in this study is expected to be adopted and implemented in traditional Chinese medicine pulse diagnosis due to its simple operability, excellent identification accuracy and economical cost advantages. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 It is a schematic diagram of a single cycle of the floating pulse according to the present invention; Figure 2 Be pulse signal identification flow chart of the present invention; Figure 3 Examples of floating pulse conditions identified by the present invention Figure 1 ; Figure 4 Examples of floating pulse conditions identified by the present invention Figure 2 ; Figure 5 This is an example diagram of a non-floating pulse identified by the present invention. DETAILED DESCRIPTION

[0015] The present invention will be described in detail below in conjunction with the accompanying drawings and specific embodiments so that the present invention can be better understood and implemented by those skilled in the art.

[0016] Embodiment 1: The present invention provides a method for identifying a floating pulse, comprising the following steps.

[0017] The initial step involves using a pulse signal capture device to collect raw pulse data (returned in a list through multiple data points) from the left wrist of the subject at different compression intensities (light, heavy, medium). After collection, filtering operations are applied to these signals to filter out high-frequency noise and stabilize the baseline, thereby obtaining a pre-processed pulse signal, such as Figure 3 shown.

[0018] The pre-processed pulse signal is imported into the mathematical analysis model to perform the feature recognition procedure, which includes two key steps: extraction of key features and feature comparison. In this process, the key feature points of the pulse wave are defined as (X1, Y1), (X2, Y2), (X3, Y3), (X4, Y4), which correspond to the pulse initial point, the main wave trough point, the secondary wave peak point and the pulse termination point respectively. The expression of a single cycle is as follows: Figure 1 shown.

[0019] Get the original pulse data under light, heavy and medium pressures. 出 Smaller than normal pulse, P 出 ≤1.03N If the middle or deep pressure is used to record the pulse, the main wave amplitude will decrease and the pressure should be less than 2.26N.

[0020] In the feature matching stage, according to the preset identification criteria, only the pulse signals that meet the following criteria are retained: Calculate the average main wave amplitude of all cycles for each piece of data in the Cunbu data. The average main wave amplitudes obtained under light, heavy and medium pressures are recorded as M1, M2 and M3. The three values ​​calculated from the data in Table 1 are 4348.2, 2616 and 1653.5 respectively. If M1>M2 and M1>M3 are found, it is judged as a floating pulse wave, and the next step is continued.

[0021] Calculate the slope k of the ascending branch of the retained pulse wave signal, The slopes under light, heavy and medium pressures are 164, 94.4 and 61.8 respectively, satisfying k≥11.43. If k<11.43, the signal is removed.

[0022] Table 1 Statistics of light, heavy and medium amplitudes in Example 1 Calculate the ejection period (the time value between the starting point of the pulse wave graph and the Jiangzhongxia) time t of the retained pulse wave signal, t=(X4-X1)×0.008 =0.71 seconds, and the pulse wave signal that satisfies t≤0.72 seconds, if t>0.72 seconds, remove the signal.

[0023] Determine the height of the main wave peak Y2 and the height of the descending gorge Y3. According to the range set by the A / D circuit of the pulse diagnosis instrument used, the ratios of light, heavy and medium pressures are 0.35, 0.51 and 0.38 respectively, satisfying the Y3 / Y 2＜ 0.65, if Y3 / Y2>0.65, remove the signal.

[0024] Finally, when the pulse waveform shows a descending angle higher than 84°, there is no significant fluctuation in the descending branch, and the ratio of the valley to the secondary peak height is maintained within a range of less than 0.65, and the waveform clearly shows the main wave, secondary wave and weak re-beat wave, it can be concluded that the pulse wave has typical characteristics of a deep pulse. The example pulse diagrams of floating pulses at three compression levels of shallow, deep and moderate that were successfully identified by this method are shown in Figure 3 Among.

[0025] Embodiment 2: First, use the pulse signal acquisition module (such as a wrist pulse pillow instrument) to obtain the original pulse signal of the left wrist of the subject under different pressures (light, heavy, medium) (returned in the form of a list through multiple data points), and satisfy the pulse diagram with pressure P 出 Smaller than normal pulse, P 出 ≤1.03N. If the pulse is recorded by taking the middle or sinking pressure, the main wave amplitude will decrease, and the pressure should be less than 2.26N. After the acquisition is completed, these pulse signals are processed by Butterworth low-pass filtering to eliminate the baseline drift phenomenon and obtain the pre-processed pulse signal, such as Figure 4 shown.

[0026] Next, the pulse signal after filtering is input into the mathematical model for feature recognition, which mainly includes two steps: extraction of effective features and feature matching. Among them, the effective feature points of the pulse wave are marked as (X1, Y1), (X2, Y2), (X3, Y3), (X4, Y4), representing the pulsation start point, the main wave peak point, the main wave end point and the pulsation end point respectively. In the effective feature extraction stage, the relationship between the feature points is calculated, such as the ascending branch slope , the average main wave amplitude (M1 under light pressure, M2 under heavy pressure, and M3 under medium pressure), and the ejection time t=(X4-X1)×0.008.

[0027] In the effective feature matching stage, according to the set screening rules, the pulse wave signals that meet the following conditions are retained: the average main wave amplitudes calculated from Table 2 are M1=2611.8, M2=1652.2, and M3=247.8, which exceed the medium and heavy pressure states, that is, M1>M2 and M1>M3; the slopes k under light, heavy, and medium pressures are 104.26, 54.2, and 16.8, respectively, all greater than 11.43; the ejection phase time is 0.705 seconds and does not exceed 0.72 seconds; the ratio of the main wave peak to the descending isthmus height Y3 / Y2 under light, heavy, and medium pressures is 0.61, 0.56, and 0.53, respectively, less than 0.65.

[0028] Table 2 Statistics of light, heavy and medium amplitudes in Example 2 Finally, if the pulse wave morphology meets the above characteristic conditions, that is, the angle of the main wave rising branch is greater than 84°, the ejection period is normal or slightly short, the descending branch is fast, the ratio of the descending isthmus height to the main wave height is less than 0.65, and the waveform contains clearly identifiable main waves, tidal waves and dicrotic waves, then it is determined to be a pulse wave with typical floating pulse characteristics. This data is successfully identified to meet the requirements of floating pulse data.

[0029] Embodiment 3: In the present embodiment, we will provide a kind of implementation process of identifying non-floating pulse data, and this process mainly judges whether pulse condition meets floating pulse feature according to the feature matching principle mentioned in the present invention. Below is the detailed identification process: First, the pulse signal is collected by the wrist pulse pillow device equipped with the acquisition module (returned in the form of a list through multiple data points), and the collected pulse signal is low-pass filtered and baseline drift corrected to obtain the pre-processed pulse signal. Figure 5 shown.

[0030] Subsequently, the preprocessed pulse signal is input into the mathematical model to start the effective feature extraction and matching steps. Then, the preprocessed pulse signal is input into a specially designed mathematical model for feature analysis. The model focuses on the key feature points of the pulse wave and calculates the corresponding feature parameters.

[0031] In the feature extraction step, the pulse waveform's beat start point (X1, Y1), main wave peak point (X2, Y2), main wave end point (X3, Y3), and beat end point (X4, Y4) are determined. Based on these points, several key features are calculated, such as the average main wave amplitude (M1, M2, and M3 are calculated at light, heavy, and medium pressures, respectively), and the ascending limb slope And ejection time t=(X4-X1)×0.008.

[0032] In the feature matching and non-floating pulse judgment stage, if the characteristics of the pulse waveform do not meet the characteristic combination of floating pulse, that is: the average main wave amplitude does not meet the value under light pressure that is significantly greater than the pressure value under medium pressure and heavy pressure, that is, it does not meet the conditions M1>M2 or M1>M3; the ascending branch slope k<11.43, or the ejection phase time t>0.72 seconds; the ratio between the main wave peak height Y2 and the descending middle isthmus height Y3 is greater than 0.65.

[0033] According to the pulse signals collected in Example 3 and the data in Table 3 obtained by analysis, we found that the ratio parameter M1 < M2, which is inconsistent with the floating pulse feature, so it is confirmed that the pulse wave image data does not belong to the floating pulse category. Using this feature matching and exclusion strategy, we can effectively distinguish and eliminate pulse data with non-floating pulse features, providing an important reference for clinicians to further identify other pulse types. Figure 5 The examples shown are pulse images that are clearly marked as non-floating pulses after this series of feature analysis processes, demonstrating the practicality of this method in refining diagnostic classification and focusing on other potential pulse characteristics. This systematic elimination process not only helps to narrow the scope of diagnosis, but also lays a solid foundation for accurately guiding other possible pulse types.

[0034] Table 3 Statistics of light, heavy and medium amplitudes in Example 3 The present invention aims to propose a new pulse recognition framework, which not only avoids the overfitting risk that may be caused by the expansion of CNN, but also reduces the high dependence on the recursive graph generation parameters through the optimization of the algorithm and mathematical model adjustment. Specifically, through the optimization of the appropriate mathematical model, the present invention not only simplifies the pre-processing steps, effectively reduces the computational burden and resource requirements, but also integrates the principles of traditional Chinese medicine pulse diagnosis with modern mathematical models, comprehensively analyzes multiple pulse signals and clinical manifestations such as pulsation intensity, waveform morphology, frequency components, etc., significantly improves the comprehensiveness and accuracy of recognition, and enhances the practical value and deployment feasibility of the model.

Claims

1. A method for identifying and processing a floating pulse wave, characterized in that: The feature matching includes the pulse wave signal's pulsation start point, main wave peak point, main wave end point, and pulsation end point, which are set to (X1, Y1), (X2, Y2), (X3, Y3), (X4, Y4) in sequence. The overall recognition process is as follows: S1. Use the pulse diagnosis signal acquisition device to obtain the original pulse data under light, heavy and medium pressure; S2. Perform Butterworth low-pass filtering on the pulse data under three pressures to remove noise; S3. Get the three processed data, calculate the average main wave amplitude of all cycles for each data, and the pulse diagram appears pressure P 出 Smaller than normal pulse, P 出 ≤1.03N. If the pulse is recorded by taking the middle or deep pressure, the main wave amplitude will decrease. The pulse pressure should be less than 2.26N. The pulse pressure is correct. The average main wave amplitudes obtained under light, heavy and middle pressures are recorded as M1, M2 and M3 respectively. S4. The following further determines that if M1>M2 and M1>M3, it is determined to be a floating pulse type pulse wave; S5. Calculate the slope k of the ascending branch of the retained pulse wave signal, , remove the pulse wave signal with k<11.43; S6. Calculate the ejection period of the retained pulse wave signal, that is, the time t between the starting point of the pulse wave graph and the time value of the descending gorge, t = (X4-X1) × 0.008, and remove the pulse wave signal of t> 0.72 seconds; S7. Determine the height of the main wave peak Y2 and the height of the descending gorge Y3, according to the range set by the A / D circuit of the pulse diagnosis instrument used, and remove the pulse wave signal of Y3 / Y2>0.65; S8. After the above feature extraction and recognition, the pulse waveform that meets the conditions is extracted, which is the floating pulse wave.

2. The method for identifying and processing a floating pulse wave according to claim 1, characterized in that: In step S1, the pulse diagnosis device for collecting the original pulse signal is a pulse diagnosis instrument, and the pulse diagnosis instrument collects pulse wave data including: the beating data of the pulses in the three parts of Cun, Guan and Chi under light, heavy and medium pressures.

3. The method for identifying and processing a floating pulse wave according to claim 1, characterized in that: The angle of the rising branch of the main wave is greater than 84°.

4. The method for identifying and processing a floating pulse wave according to claim 1, characterized in that: The ratio of the height of the descending gorge to the height of the main wave is less than 0.

65.

5. The method for identifying and processing a floating pulse wave according to claim 1, characterized in that: In step S3, the calculation method of the main wave amplitudes M1, M2, and M3 is as follows: S4.1 Add the heights of the main waves in each period to obtain the total; S4.2 Calculate the average main wave amplitude; S4.3 Calculate the average main wave amplitudes under light, heavy and medium pressures respectively and record them as M1, M2 and M3.

Images