Pulse wave identification processing method

Abstract

The invention provides a pulse wave identification processing method. The pulse wave identification processing method comprises the steps of data acquisition, period extraction, filtering processing and data identification. Pulse condition data are acquired through data acquisition, and periodic extraction is carried out to determine periodic characteristics of pulse waves; then, processing the data by adopting a filtering processing method so as to remove interference signals and noise, and extracting a target pulse wave signal; and constructing a mathematical model by using the extracted features, and carrying out feature matching and screening so as to realize accurate identification of the pulse condition. According to the method, the identification efficiency can be remarkably improved, and a feasible solution is provided for the medical field.

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 pulse wave. Background Art

[0002] Pi pulse is a type of pulse in traditional Chinese medicine, which is characterized by being large and hollow, like pressing a scallion tube. In pulse diagnosis in traditional Chinese medicine, Pi pulse is believed to be closely related to the occurrence and development of certain diseases, so the accurate identification and analysis of Pi pulse is of great significance. Current pulse diagnosis technology has some challenges in the identification of Pi pulse. Traditional pulse diagnosis mainly relies on the subjective experience of doctors, lacks objective theoretical basis and standardized diagnostic methods, resulting in a certain degree of subjectivity and uncertainty in the identification results of Pi pulse.

[0003] In the past few years, with the rapid development of new material technology and electronic computer technology, these technologies have been widely used in the medical field. In terms of the acquisition, processing and visualization of pulse information, remarkable progress has been made, which provides a foundation for the establishment of the theoretical basis of pulse signals. The breakthroughs in these new material technologies make pulse acquisition equipment such as pulse diagnosis instruments more accurate and efficient. By collecting the pulse data of patients and combining advanced signal processing algorithms, key feature information in the pulse can be accurately extracted. At the same time, the continuous development of electronic computer technology has provided powerful computing power and algorithm support for big data analysis and machine learning. The patent with the authorization publication number CN116089797A uses convolutional neural networks to simulate the characteristics of human neuronal structure to transmit information layer by layer, realize automatic feature extraction, and finally realize pulse recognition. The patent with the authorization publication number CN113712516A discloses a method for establishing a traditional Chinese medicine pulse recognition model based on a temporal convolutional network. The pulse recognition method proposed in the invention has a high accuracy rate and can provide a reference for the diagnosis of the health status of the human cardiovascular system. In the paper "Research on Pulse Signal Recognition Based on Time Series and Time Series Convolutional Network", Zhu Guangyao et al. obtained a pulse signal time series with consistent length by preprocessing and sequence normalization of pulse signals, and used a deep learning network based on time series convolution to extract the features of sequence morphology and establish a pulse signal recognition model. In the paper "Research on Pulse Signal Recognition Based on Weighted Soft Voting Fusion Model", Liu Qichao et al. solved the data imbalance problem by using the boundary synthesis minority class sample oversampling technology, and constructed a weighted soft voting fusion model based on four machine learning methods: XGBoost, random forest, LightGBM, and gradient boosting decision tree. The final model will output a specific pulse category, which has lower operational complexity and higher accuracy than the commonly used pulse recognition methods, and the shorter training time also makes it more clinically useful in the recognition of multiple types of pulse signals. Summary of the invention

[0004] The present invention provides a method for identifying and processing the Pi pulse wave, which can effectively improve the diagnostic accuracy and standardization level of the Pi pulse, and provide a new technical means for clinical diagnosis and treatment in traditional Chinese medicine. The method has broad application prospects in the medical field, and can be used to assist doctors in diagnosing the Pi pulse and predicting diseases, providing patients with more accurate traditional Chinese medicine treatment plans, and promoting the popularization and development of traditional Chinese medicine pulse diagnosis technology.

[0005] A method for identifying and processing a pulse wave comprises the following steps: S1. Data collection: Use the pulse diagnosis device to collect the raw data of the pulse wave, which includes the vibration of the pulse wave over a period of time; S2. Period extraction: Process the collected raw pulse wave data and convert it into periodic signal data. This step divides the pulse wave morphology into multiple periods in order to better analyze and identify the periodic characteristics of the pulse wave; S3. Filtering: Filter the extracted periodic signal data to obtain a filtered signal. Filtering helps remove noise and unnecessary fluctuations, thereby improving the quality and accuracy of the signal; S4. Data recognition: Input the filtered signal into the mathematical model for recognition. Data recognition includes extracting effective features and feature matching.

[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 beat data of three parts: Cun, Guan and Chi.

[0007] Furthermore, in step S4, effective features include main wave width, main wave height, tidal wave height, dicrotic wave height, descending gorge height, main peak angle and ascending branch angle.

[0008] Furthermore, feature matching also includes: The pulse wave signal's pulsation start point, main wave peak point, main wave end point, and pulsation end point are set to (X1, Y1), (X2, Y2), (X3, Y3), and (X4, Y4) respectively.

[0009] S4.1 calculates the slope k of the ascending branch of the retained pulse wave signal, , remove the pulse wave signal with k<11.43; S4.2 Determine the height Y2 of the main wave peak and the height Y3 of the descending gorge, and remove the pulse wave signal with Y3 / Y2 < 0.15 according to the range set by the A / D circuit of the pulse diagnosis instrument used; S4.3 calculate the contraction period of the retained pulse wave signal, that is, the ejection period time t, t=(XX)×0.008, and remove the pulse wave signal with t>0.72 seconds; S4.4 calculates the distance L1 from the end point of the main wave in the retained pulse wave signal to the ascending branch of the pulse wave, , the distance from the end point of the main wave to the peak point of the main wave is L2, , the main wave angle θ, .

[0010] Furthermore, the waveform is required to have floating pulse characteristics, the angle of the ascending branch of the main wave is greater than 84°, and there are no small waves on the ascending branch, the ratio of the height of the descending gorge to the height of the main wave is less than 0.15, and the waveform only has the main wave and the heavy wave.

[0011] Furthermore, the data points of the pulse wave raw data are composed of multiple data points, and the period extraction calibrates the pulse wave raw data into multiple periods, that is, the pulse wave raw data is divided into multiple repeated periods.

[0012] Furthermore, in the process of filtering, a Butterworth filter is used, which is configured as a low-pass filter. Through this filter, the input signal can be effectively filtered out of high-frequency noise and interference components, thereby highlighting and extracting the characteristics of the pulse wave.

[0013] Furthermore, the extraction of effective features refers to extracting relevant information that can best describe the characteristics of the pulse wave from the filtered signal. These features should have the ability to identify and accurately distinguish the pulse wave from other pulse wave forms. Feature matching refers to comparing and matching the extracted effective features with the pre-built mathematical model to determine the pulse wave form represented by the filtered signal.

[0014] The above technical solution of the present invention has the following advantages compared with the prior art: The pulse wave recognition method of the present invention adopts a simple recognition step, which is more simplified and efficient than the complex deep learning method. Although it adopts a simple recognition step, the method can still maintain a relatively high recognition accuracy. Compared with deep learning methods, this method is simpler to implement and has lower costs, making it more feasible and practical.

[0015] In summary, the pulse wave identification method proposed in this study has the advantages of simplicity, high accuracy and low cost, and is expected to be widely used in the field of pulse diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a schematic diagram of a single cycle of the Pi pulse according to the present invention; Figure 2 It is the overall pulse signal recognition flow chart; Figure 3 It is a schematic diagram of matching feature points for the pulse diagram; Figure 4 It is a schematic diagram of light pulse image for the Pi pulse identification output of the present invention; Figure 5 It is a schematic diagram of the pulse image of the middle pulse output for the Pi pulse identification of the present invention; Figure 6 It is a schematic diagram of retaking pulse condition for Pi pulse identification output of the present invention; Figure 7 Schematic diagram of the pulse condition of the non-pi pulse.

[0017] In the figure, h1-main wave height; h2-falling isthmus height; h3-dicrotic wave height; t-time value between the starting point of the pulse wave graph and the falling isthmus.

[0018] DETAILED DESCRIPTION OF THE EMBODIMENTS The embodiments of the present invention will be described in more detail below in conjunction with the accompanying drawings so that the present invention can be better understood and implemented by those skilled in the art.

[0019] Embodiment 1: The present invention provides a method for identifying the pulse condition of the cun, guan and chi pulses, which is collected by sensors designed based on the theory of traditional Chinese medicine, and then pre-processed by dividing the period and filtering the noise, and finally by mathematical model and feature matching, so as to realize the identification of the pulse type. Figure 1 As shown, the filtered signal should include the main wave and the dicrotic wave, and the tidal wave completely disappears in the pulse image of the iliac pulse provided in this embodiment. Please continue to refer to Figure 2 Shown is a flow chart for identifying an example pulse condition, as shown in the figure, comprising the following steps.

[0020] Step 1: Get the original pulse signal of the target. In order to identify the pulse, it is necessary to obtain the original pulse signal of the target person as the basic data. In order to ensure the reliability of the data, this study uses a highly sensitive acquisition device to collect pulse wave data of light, medium and heavy pressure on the three parts of the left hand of the target person, namely, Cun, Guan and Chi. The sampling period for each part is 6 seconds, and a total of 18 sets of data are collected. Step 2: Filtering. The collected periodic signal is low-pass filtered, and the pulse wave signal is denoised by a low-pass filter to remove the noise component in the original signal. Step 3: Extract the period of the filtered pulse. Since the sampling period of the pulse wave signal is 6 seconds and the human heart rate is generally in the range of 60 to 100 times / minute, the extracted pulse wave pulsation period signal is usually 4 to 6 periods. Step 4: Input the pulse signal with divided periods into the model for identification. Input the pulse signal after filtering and period extraction into the pulse recognition model for pulse classification and identification. Step 5: Output the recognition result. The input pulse signal is analyzed and judged according to the model, and the result of pulse recognition is output, that is, the conclusion of classification and recognition of the pulse wave of the target person.

[0021] For clear instructions on feature matching and pulse recognition, refer to Figure 3As shown, the pulsation starting point, main wave peak point, main wave end point and pulsation end point of the pulse wave signal are respectively set to (X1, Y1), (X2, Y2), (X3, Y3), (X4, Y4).

[0022] First, calculate the slope k of the ascending branch of the retained pulse wave signal, , remove the pulse wave signal with k<11.43; judge the height Y2 of the main wave peak and the height Y3 of the descending isthmus, and remove the pulse wave signal with Y3 / Y2<0.15 according to the range set by the A / D circuit of the pulse diagnosis instrument used. Calculate the systolic (ejection) time t of the retained pulse wave signal, t=(XX)×0.008, and remove the pulse wave signal with t>0.72 seconds. Calculate the distance L1 from the end point of the main wave in the retained pulse wave signal to the ascending branch of the pulse wave, , the distance from the end point of the main wave to the peak point of the main wave is L2, , the main wave angle θ, .

[0023] After the above multiple identification and calculation, the overall performance of the obtained data is that the pulse position is floating, the angle of the main wave rising branch is greater than 84°, and there is no small wave in the rising branch. The ratio of the height of the descending gorge to the height of the main wave is less than 0.15. The waveform only has the main wave and the heavy wave, which is the pulse wave of the Pi. The whole data identification process is highly accurate and can accurately identify the pulse wave of the Pi. Figures 4 to 6 shown.

[0024] Embodiment 2: Next, an embodiment process for correctly identifying the pulse condition of the Pi pulse is provided. This process is mainly identified by the identification process in the claims. Next, the identification process is specifically carried out to obtain the pulse condition result output by the real model operation.

[0025] First, the pulse signal is acquired through the device of the acquisition module. The acquisition device and the output device are mainly on the wrist pulse pillow device. The pulse signal is low-pass filtered and the baseline drift is removed to obtain the pulse signal after preprocessing.

[0026] The obtained filtered signal is then input into the mathematical model for recognition, which includes extracting effective features and performing feature matching. The following feature point division method refers to Figure 3 .

[0027] Extraction of effective features. The model will first extract special points in the pulse signal sequence, such as the pulse wave signal's pulsation starting point, main wave peak point, main wave end point and pulsation end point. The specific data points obtained can then be used to calculate the corresponding features. Calculate the slope k of the ascending branch of the retained pulse wave signal, as well as the ratio between the height Y2 of the main wave peak and the height Y3 of the descending gorge, calculate the systolic period (ejection period) time t of the retained pulse wave signal, and finally calculate the main wave angle through the pulsation starting point, main wave peak point and main wave end point.

[0028] Matching of effective features. After obtaining the feature data of the above feature points, the next step is to filter the signal. Keep the pulse wave signal that satisfies the ascending branch slope k>11.43, the ratio between the height of the main wave peak Y2 and the height of the descending gorge Y3 Y3 / Y2>0.15, and the contraction period t of this pulse wave signal is <0.72 seconds, and obtain this pulse signal.

[0029] Finally, it is judged that the pulse signal meets the pulse position floating, the angle of the main wave rising branch is greater than 84°, and there is no small wave in the rising branch, the ratio of the height of the descending gorge to the height of the main wave is less than 0.15, and the waveform has only the main wave and the heavy wave, then it is a Pi pulse waveform. The specific identification of the three types of light, medium and heavy Pi pulse examples is shown in the figure Figures 4 to 6 shown.

[0030] Embodiment 3: This embodiment provides an embodiment process of identifying non-pulse data, which mainly determines whether it is non-pulse data through feature matching in embodiment 1.

[0031] First, the pulse signal is acquired through the device of the acquisition module. The acquisition device and the output device are mainly on the wrist pulse pillow device. The pulse signal is low-pass filtered and the baseline drift is removed to obtain the pulse signal after preprocessing.

[0032] Next, the filtered signal is input into the mathematical model for effective feature extraction and feature matching.

[0033] Effective feature extraction: First, extract special points from the pulse signal sequence, such as the starting point, main wave crest point, main wave end point and pulsation end point of the pulse wave signal, and calculate the features related to these points. Specifically, calculate and retain the ascending branch slope k of the pulse wave signal, the ratio between the main wave crest height Y2 and the descending gorge height Y3, and the systolic period (ejection period) time t of the pulse wave signal. In addition, use the pulsation starting point, main wave crest point and main wave end point to calculate the main wave angle.

[0034] Feature matching: After obtaining the above feature data, feature matching is performed to determine whether it is non-Qi pulse data. Specifically, if the ascending branch slope k is less than 11.43, the ratio between the main wave peak height Y2 and the descending isthmus height Y3 is Y3 / Y2<0.15, and the systolic time t>0.72s pulse wave signal, if any of the above conditions is met, it is determined to be a non-Qi pulse signal.

[0035] According to the judgment of the recognition model, Figure 7 In a single cycle, there are double-peak waves and tidal waves, and the ratio between the wave peak height Y2 and the descending gorge height Y3 is Y3 / Y2<0.15. The subsequent conditions are directly skipped and the model output is a non-pulse wave, such as Figure 7 shown.

[0036] Obviously, the above examples are only for illustration and do not limit the implementation methods. Those skilled in the art can make other forms of changes or adjustments based on the above description. It is not necessary to list all the implementation methods in detail, and obvious changes derived therefrom still belong to the protection scope of the present invention.

Claims

1. A method for identifying and processing a pulse wave, characterized in that: The steps include: S1. Data collection: Use the pulse diagnosis device to collect the raw data of the pulse wave, which includes the vibration of the pulse wave over a period of time; S2. Period extraction: Process the collected raw pulse wave data and convert it into periodic signal data; this step divides the pulse wave morphology into multiple periods in order to better analyze and identify the periodic characteristics of the pulse wave; S3. Filtering: filtering the extracted periodic signal data to obtain a filtered signal; filtering helps to remove noise and unnecessary fluctuations, thereby improving the quality and accuracy of the signal; S4. Data recognition: Input the filtered signal into the mathematical model for recognition. Data recognition includes extracting effective features and feature matching.

2. The method for identifying and processing a 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. The pulse diagnosis instrument collects pulse wave data including pulse beat data of three parts: Cun, Guan and Chi.

3. The method for identifying and processing a pulse wave according to claim 1, characterized in that: In step S4, the effective features include the main wave width, main wave height, tidal wave height, dicrotic wave height, descending gorge height, main peak angle and ascending branch angle.

4. The method for identifying and processing a pulse wave according to claim 3, characterized in that: Feature matching also includes: The pulse wave signal's pulsation start point, main wave peak point, main wave end point, and pulsation end point are set to (X1, Y1), (X2, Y2), (X3, Y3), and (X4, Y4) in sequence; S4.1 calculates the slope k of the ascending branch of the retained pulse wave signal, , remove the pulse wave signal with k<11.43; S4.2 Determine the height Y2 of the main wave peak and the height Y3 of the descending gorge, and remove the pulse wave signal with Y3 / Y2 < 0.15 according to the range set by the A / D circuit of the pulse diagnosis instrument used; S4.3 calculate the contraction period of the retained pulse wave signal, that is, the ejection period time t, t=(XX)×0.008, and remove the pulse wave signal with t>0.72 seconds; S4.4 calculates the distance L1 from the end point of the main wave in the retained pulse wave signal to the ascending branch of the pulse wave, , the distance from the end point of the main wave to the peak point of the main wave is L2, , the main wave angle θ, .

5. The method for identifying and processing a pulse wave according to claim 3, characterized in that: The waveform is required to have floating pulse characteristics, the angle of the ascending branch of the main wave is greater than 84°, and there are no small waves on the ascending branch. The ratio of the height of the descending gorge to the height of the main wave is less than 0.15, and the waveform only has the main wave and the heavy wave.

6. The method for identifying and processing a pulse wave according to claim 1, characterized in that: The data points of the pulse wave raw data are composed of multiple data points, and the period extraction calibrates the pulse wave raw data into multiple periods, that is, the pulse wave raw data is divided into multiple repeated periods.

7. The method for identifying and processing a pulse wave according to claim 1, characterized in that: In the process of filtering, a Butterworth filter is used, which is configured as a low-pass filter; through this filter, the input signal can be effectively filtered out of high-frequency noise and interference components, thereby highlighting and extracting the characteristics of the pulse wave.

8. The method for identifying and processing a pulse wave according to claim 1, characterized in that: The extraction of effective features refers to extracting relevant information that can best describe the characteristics of the Pi pulse from the filtered signal. These features should have the ability to recognize and accurately distinguish the Pi pulse from other pulse wave forms; feature matching refers to comparing and matching the extracted effective features with the pre-constructed mathematical model to determine the pulse wave form represented by the filtered signal.

Images