Pulse condition feature analysis method, system and device and storable medium

Abstract

The invention discloses a pulse condition characteristic analysis method, system and device and a storable medium, a seventy-two-channel pressure sensor is arranged at the cun-guan-chi position of a wrist, pressure changes can be comprehensively and carefully captured, richer and more accurate pulse condition data can be obtained, three groups of different pressure values are applied to simulate floating, middle and sinking force, traditional Chinese medicine pulse beating operation is highly restored, and the pulse condition characteristic analysis accuracy is improved. Butterworth filtering is adopted for processing data of all channels, high-frequency noise and low-frequency interference are effectively removed, data quality is improved, smoothing processing is conducted on the filtered data, valley values are recognized, periods are determined, pulse period signals are accurately extracted, analysis of rhythm, frequency and other characteristics is facilitated, physiological parameters are extracted by means of a target detection model, and the accuracy of pulse detection is improved. The method is simple, efficient and accurate, manual subjective errors are avoided, finally, true values of physiological parameters are substituted into pulse type judgment standards, specific pulse types are obtained, pulse type analysis is scientific, accurate and repeatable, and the limitation existing when complex pulse conditions are processed is effectively solved.

Description

Technical Field

[0001] The present invention relates to the technical field of pulse characteristic analysis, and in particular to a pulse characteristic analysis method, system, equipment and storable medium. Background Art

[0002] In traditional Chinese medicine diagnosis, pulse characteristic analysis is a crucial link. Doctors touch the patient's pulse and sense the depth of the pulse, the speed of the pulse rate, the size of the pulse width, the length of the pulse, the strength of the pulse, the smoothness of the pulse, the tightness of the tension, and the regularity of the uniformity, etc., to determine the patient's pulse type, thereby providing an important basis for the subsequent selection of treatment drugs.

[0003] Modern medical research has gradually introduced sensing technology and algorithm analysis into pulse characteristic analysis. In terms of sensing technology, although a variety of sensors have been used for pulse collection, the detection and analysis results of the same pulse by a single sensor in different devices, different subjects or physiological states may still have deviations.

[0004] In addition, in terms of pulse feature analysis algorithms, current research mainly involves multiple dimensions such as signal processing, waveform recognition, rhythm analysis, force assessment, and slippery perception. These algorithms can be roughly divided into two categories: classical signal processing methods and artificial intelligence algorithms. Classical signal processing methods include using Fourier transform to analyze pulse frequency components and detecting local characteristics of pulse through wavelet transform. However, these methods often have limitations when dealing with complex pulse patterns. Summary of the Invention

[0005] In view of this, the present invention proposes a pulse characteristic analysis method, system, device and storable medium, which can effectively solve the defects of the existing technology that a single sensor leads to deviations in detection and analysis results, and the pulse characteristic analysis algorithm has limitations when processing complex pulse conditions.

[0006] The technical solution of the present invention is achieved as follows:

[0007] A pulse characteristic analysis method specifically comprises:

[0008] Data acquisition: A 72-channel pressure sensor was placed at the Cun, Guan, and Chi points on the wrist. Three different pressure values were applied to the back of the pressure sensor to simulate the floating, middle, and sinking forces used in traditional Chinese medicine pulse diagnosis. The data obtained by the pressure sensor was recorded.

[0009] Data filtering: Butterworth filtering is used to filter the data of each channel;

[0010] Period extraction: Smooth the filtered data, identify valleys, determine the period through the coordinates of adjacent valleys, and extract the pulse period signal;

[0011] Physiological parameter extraction: Convert the periodic signal data into an image, input the image into a target detection model that has completed feature training for the starting point, main wave, pre-dicrotic wave, dicrotic wave, and descending gorge, obtain the corresponding image position information, and convert the image position information to obtain the true value of each physiological parameter;

[0012] Pulse type analysis: Substitute the true values of various physiological parameters into the judgment criteria of each pulse type to obtain the specific pulse type corresponding to the pulse image.

[0013] As a further optional solution of the pulse characteristic analysis method, in the data acquisition step, the seventy-two channel pressure sensors are arranged in a 12x6 array distribution manner at the Cun, Guan and Chi positions of the wrist.

[0014] As a further optional solution of the pulse characteristic analysis method, in the data filtering step, the high-frequency cut-off frequency of the Butterworth filter is 18 Hz, the low-frequency cut-off frequency is 5 Hz, the sampling frequency is 100 Hz, and the order of the filter is 5.

[0015] As a further optional solution of the pulse characteristic analysis method, the data after the filtering process is smoothed, the valley value is identified, the cycle is determined by the coordinate values of adjacent valley values, and the periodic signal of the pulse is extracted, which specifically includes:

[0016] A low-pass filter is used to smooth the filtered data;

[0017] identifying a valley in the smoothed data, the valley being the lowest point in the pulse waveform;

[0018] Record the horizontal coordinate values of all identified valley values and calculate the coordinate difference between two adjacent valley values;

[0019] Analyze the distribution interval and mode of all adjacent valley coordinate differences to determine a pulse cycle;

[0020] Utilizing the determined pulse cycle, extracting it from the original filtered data to obtain multiple complete pulse cycle signals;

[0021] The obtained multiple complete pulse cycle signals are compared for similarity, and the group with the highest similarity is selected as the pulse cycle signal for subsequent pulse type analysis.

[0022] As a further optional solution of the pulse characteristic analysis method, the periodic signal data is converted into an image, the image is input into a target detection model that has completed feature training of the starting point, main wave, pre-dicrotic wave, dicrotic wave, and descending gorge, the corresponding image position information is obtained, and the image position information is converted to obtain the true value of each physiological parameter, specifically including:

[0023] converting the extracted pulse cycle signal data into two-dimensional image data;

[0024] The converted image is input into the target detection model that has completed feature training for the starting point, main wave, pre-dicrotic wave, dicrotic wave, and descending gorge. The target detection model outputs a series of bounding box predictions, each of which contains the target location and the probability of the target appearing at that location.

[0025] According to the target position information of the bounding box, the corresponding position is found in the original pulse cycle signal, and the true value of each physiological parameter is obtained by conversion.

[0026] As a further optional solution of the pulse characteristic analysis method, the pulse type analysis step includes pulse position classification, pulse rate classification, pulse width classification, pulse length classification, pulse strength classification, fluency classification and tension classification.

[0027] As a further optional scheme of the pulse characteristic analysis method, the pulse position classification is judged based on the data similarity of the three forces of floating, middle and sinking; the pulse rate classification is calculated based on the data size and sampling frequency of the pulse cycle; the pulse width classification and pulse length classification are judged based on the approximate number of data of the sensor channels expanded horizontally and vertically and the center point data respectively; the pulse force classification is judged based on the extracted physiological parameters; the fluency classification and tension classification are judged based on the extracted physiological parameters and their mutual relationship.

[0028] A pulse characteristic analysis system comprising:

[0029] The data acquisition module is used to use a pressure sensor placed at the Cun, Guan, and Chi points of the wrist, apply three sets of different pressure values to the back of the pressure sensor to simulate the three forces of floating, middle, and sinking when taking the pulse in traditional Chinese medicine, and record the data obtained by the pressure sensor;

[0030] The data filtering module is used to filter the data of each channel acquired by the data acquisition module using Butterworth filtering;

[0031] The cycle extraction module is used to smooth the data filtered by the data filtering module, identify the valley value, determine the cycle by the coordinate values of adjacent valley values, and extract the pulse period signal;

[0032] The physiological parameter extraction module is used to convert the periodic signal data extracted by the period extraction module into an image, input the image into the target detection model that has completed feature training for the starting point, main wave, pre-dicrotic wave, dicrotic wave, and descending gorge, obtain the corresponding image position information, and convert the image position information to obtain the true value of each physiological parameter;

[0033] The pulse type analysis module is used to substitute the true values of the various physiological parameters obtained by the physiological parameter extraction module into the judgment criteria of each pulse type to obtain the specific pulse type corresponding to the pulse image.

[0034] A computing device comprises a memory, a processor and a computer program stored in the memory and runnable on the processor. When the processor executes the computer program, the steps of any one of the above-mentioned pulse characteristic analysis methods are implemented.

[0035] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of any one of the above-mentioned pulse characteristic analysis methods.

[0036] The beneficial effects of the present invention are as follows: a 72-channel pressure sensor is placed at the inch, guan, and chi positions of the wrist, which can comprehensively and meticulously capture pressure change information at different positions of the wrist. Compared with traditional single-channel or small-channel sensors, richer and more accurate pulse data can be obtained. Three groups of different pressure values are applied to simulate the three forces of floating, middle, and sinking, highly restoring the actual pulse-taking operation process of traditional Chinese medicine, so that the collected data is closer to the cognition and judgment basis of traditional Chinese medicine on pulse. Butterworth filtering is used to filter the data of each channel. The Butterworth filter has the characteristics of a frequency response curve that is maximally flat within the passband without fluctuations and gradually decreases to zero in the stopband. This characteristic can effectively remove high-frequency noise and low-frequency interference in the collected pulse data, improve the quality and signal-to-noise ratio of the data, smooth the filtered data, identify valley values, determine the period through the coordinate values of adjacent valley values, and extract the pulse period. Period signal. This method can accurately extract the pulse period information from complex pulse data. The pulse period is one of the key parameters in pulse analysis. Accurately obtaining the period signal helps to further analyze the rhythm, frequency and other characteristics of the pulse. The target detection model is used to extract physiological parameters, which is efficient and accurate. It effectively solves the limitations in processing complex pulse patterns. The model has been trained with a large amount of data and can quickly identify key feature points in the image, greatly improving the efficiency and accuracy of physiological parameter extraction. It avoids the subjectivity and errors of manual identification and analysis in traditional methods, and substitutes the true values of various physiological parameters into the judgment criteria of each pulse type to obtain the specific pulse type corresponding to the pulse. This method is based on objective physiological parameter data, avoids the differences that may be caused by subjective judgment of traditional Chinese medicine, makes pulse type analysis more scientific, accurate and repeatable, and can provide more reliable pulse analysis results for clinical diagnosis of traditional Chinese medicine. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0038] Figure 1 Be a schematic flow diagram of a pulse characteristic analysis method of the present invention;

[0039] Figure 2 A schematic diagram of a pulse characteristic analysis system of the present invention;

[0040] Figure 3 A schematic diagram of the composition of a computing device according to the present invention;

[0041] Figure 4 It is a pulse waveform diagram. DETAILED DESCRIPTION

[0042] The following is a clear and complete description of the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0043] refer to Figures 1 to 4 , a pulse characteristic analysis method, specifically comprising:

[0044] Data acquisition: A 72-channel pressure sensor was placed at the Cun, Guan, and Chi points on the wrist. Three different pressure values were applied to the back of the pressure sensor to simulate the floating, middle, and sinking forces used in traditional Chinese medicine pulse diagnosis. The data obtained by the pressure sensor was recorded.

[0045] Data filtering: Butterworth filtering is used to filter the data of each channel;

[0046] Period extraction: Smooth the filtered data, identify valleys, determine the period through the coordinates of adjacent valleys, and extract the pulse period signal;

[0047] Physiological parameter extraction: Convert the periodic signal data into an image, input the image into a target detection model that has completed feature training for the starting point, main wave, pre-dicrotic wave, dicrotic wave, and descending gorge, obtain the corresponding image position information, and convert the image position information to obtain the true value of each physiological parameter;

[0048] Pulse type analysis: Substitute the true values of various physiological parameters into the judgment criteria of each pulse type to obtain the specific pulse type corresponding to the pulse image.

[0049] In this embodiment, a 72-channel pressure sensor is placed at the inch, guan and chi positions of the wrist, which can comprehensively and meticulously capture the pressure change information at different positions of the wrist. Compared with traditional single or small number of channel sensors, richer and more accurate pulse data can be obtained. Three groups of different pressure values are applied to simulate the three forces of floating, middle and sinking, which highly restores the actual pulse-taking operation process of traditional Chinese medicine, making the collected data closer to the cognition and judgment basis of traditional Chinese medicine on pulse. Butterworth filtering is used to filter the data of each channel. The Butterworth filter has the characteristics of the frequency response curve in the passband being maximally flat without fluctuations, and gradually decreasing to zero in the stopband. This characteristic can effectively remove high-frequency noise and low-frequency interference in the collected pulse data, improve the quality and signal-to-noise ratio of the data, and smooth the filtered data to identify the valley value. The coordinate values of the adjacent valley values determine the period and extract the pulse's periodic signal. This method can accurately extract pulse period information from complex pulse data. The pulse period is one of the key parameters in pulse analysis. Accurately obtaining the periodic signal helps further analyze the pulse's rhythm, frequency, and other characteristics. Using the target detection model for physiological parameter extraction is efficient and accurate. After training with a large amount of data, the model can quickly identify key feature points in the image, greatly improving the efficiency and accuracy of physiological parameter extraction and avoiding the subjectivity and errors of manual identification and analysis in traditional methods. The true values of each physiological parameter are substituted into the judgment criteria for each pulse type to obtain the specific pulse type corresponding to that pulse. This method is based on objective physiological parameter data, avoiding the differences that may be caused by subjective judgment in Traditional Chinese Medicine (TCM), making pulse type analysis more scientific, accurate, and repeatable. It can provide more reliable pulse analysis results for TCM clinical diagnosis.

[0050] Preferably, in the data acquisition step, the seventy-two-channel pressure sensors are arranged in a 12x6 array distribution at the Cun, Guan, and Chi positions of the wrist.

[0051] In this embodiment, 72-channel pressure sensors are arranged in a 12x6 array at the Cun, Guan, and Chi positions of the wrist. Compared with sensors with a small number of channels or a single position, they can cover a larger area of the Cun, Guan, and Chi parts of the wrist. Different positions of the Cun, Guan, and Chi in traditional Chinese medicine pulse correspond to different internal organs and physiological systems of the human body. The wider coverage can comprehensively collect pulse information related to each internal organ and avoid missing important information. There may be subtle differences in pulses at different positions. This array distribution method can accurately capture these differences. For example, in the horizontal and vertical directions of the Cun, Guan, and Chi, different channels can sense local pressure changes, which helps to discover some potential and difficult-to-detect pulse conditions. Features: The 12x6 array distribution enables the sensor to form dense sampling points on the wrist, greatly improving the spatial resolution of pulse information and being able to more accurately locate the specific position of pulse characteristics on the wrist. For example, the performance of characteristics such as the main wave and the dicrotic pre-wave on different channels helps to more accurately analyze the morphology and changing patterns of the pulse; comprehensive information collection and high spatial resolution can reduce the loss of pulse information during the collection process. Traditional sensors with a small number of channels may cause some important pulse information to be ignored due to limited coverage or insufficient resolution. This technical solution can retain the original information of the pulse to the greatest extent, thereby improving the accuracy of subsequent pulse analysis.

[0052] Preferably, in the data filtering step, the high-frequency cutoff frequency of the Butterworth filter is 18 Hz, the low-frequency cutoff frequency is 5 Hz, the sampling frequency is 100 Hz, and the filter order is 5.

[0053] In the present embodiment, the high frequency cut-off frequency of Butterworth filter is 18Hz, which means that the signal component with a frequency higher than 18Hz will be significantly attenuated. In the pulse data acquisition process, various high frequency noises will inevitably be introduced, such as the electromagnetic interference in the electronic noise of the sensor itself, the surrounding environment, etc. These high frequency noises will interfere with the pulse signal, affecting the accuracy of subsequent analysis. By setting the high frequency cut-off frequency of 18Hz, these high frequency noises can be effectively filtered out, making the filtered pulse signal smoother, retaining useful low-frequency components in the pulse signal, such as the fluctuation characteristics of pulse, etc., thereby improving the quality of pulse data; the low frequency cut-off frequency is 5Hz, which shows that the signal component with a frequency lower than 5Hz will also be attenuated by the filter. When actually collecting pulse data, there may be some low-frequency interferences, such as the baseline drift caused by physiological activities such as the slow movement of human body, breathing, and these low-frequency interferences will cover the true characteristics of the pulse signal, resulting in analysis results. If deviation occurs, setting the low-frequency cutoff frequency to 5Hz can remove these low-frequency interferences, so that the pulse signal can more accurately reflect the actual changes of the pulse, avoid the distortion of the pulse signal caused by low-frequency interference, help improve the accuracy of pulse analysis, and make the extracted physiological parameters more truly reflect the pulse characteristics of the human body; the sampling frequency is 100Hz. In order to be able to restore the signal without distortion, the sampling frequency should be at least twice the highest frequency component in the signal. In this scheme, the high-frequency cutoff frequency is 18Hz, so the sampling frequency of 100Hz can fully collect the effective information in the pulse signal; the order of the filter is 5. The higher the order, the narrower the transition band of the filter and the steeper the attenuation characteristics of the signal. In this scheme, the 5th-order Butterworth filter can achieve a relatively ideal filtering effect near the high-frequency and low-frequency cutoff frequencies, which can effectively filter out noise and interference without causing excessive attenuation to the useful components of the pulse signal.

[0054] Preferably, the step of smoothing the filtered data, identifying valleys, determining the period by the coordinate values of adjacent valleys, and extracting the pulse period signal specifically includes:

[0055] A low-pass filter is used to smooth the filtered data;

[0056] identifying a valley in the smoothed data, the valley being the lowest point in the pulse waveform;

[0057] Record the horizontal coordinate values of all identified valley values and calculate the coordinate difference between two adjacent valley values;

[0058] Analyze the distribution interval and mode of all adjacent valley coordinate differences to determine a pulse cycle;

[0059] Utilizing the determined pulse cycle, extracting it from the original filtered data to obtain multiple complete pulse cycle signals;

[0060] The obtained multiple complete pulse cycle signals are compared for similarity, and the group with the highest similarity is selected as the pulse cycle signal for subsequent pulse type analysis.

[0061] In the present embodiment, adopt low-pass filter that the data after the filtering process are carried out smoothing process, can effectively remove the small fluctuation and random noise in the data, in pulse signal, these small fluctuations may be due to factors such as the instability of sensor, the slight jitter of human body causes, they can disturb the accurate identification to pulse cycle, by low-pass filtering smoothing process, make pulse signal smoother, stable, be easier to follow-up feature extraction, low-pass filter can keep the main feature of pulse signal when removing noise, as key information such as the fluctuation trend of pulse and valley value, guaranteed in the process of removing noise, can not lose important pulse characteristics; Valley value is defined as the lowest point in the pulse waveform, and in the data after smoothing, accurately identify valley value, valley value is a key characteristic point in the pulse cycle, accurately identify valley value and can clearly determine the ups and downs of pulse, by accurately identifying valley value, can more accurately record the abscissa value of valley value, thereby calculate the coordinate difference between adjacent two valley values, these coordinate difference values have reflected the length of pulse cycle, and accurate valley value identification helps to improve the accuracy of pulse cycle calculation; Analyze all adjacent valley value coordinates The distribution interval and mode of the difference are used to determine a pulse cycle. This method takes into account the data of multiple pulse cycles. Through statistical analysis, it can avoid the influence of single abnormal data on the cycle determination, making the determined pulse cycle more representative and accurate. The pulse cycles of different people may be different. By analyzing the distribution interval and mode of the coordinate difference, the pulse cycle suitable for the individual can be adaptively determined, improving the applicability of this technical solution to different populations; the pulse cycle signal is extracted from the original filtered data, ensuring the integrity and originality of the data, avoiding excessive changes to the pulse characteristics during the data processing process, so that the extracted pulse cycle signal can more truly reflect the human body's pulse condition; the similarity of the multiple complete pulse cycle signals obtained is compared, and the group with high similarity is selected as the pulse cycle signal for subsequent pulse type analysis. Since the human body may be interfered by some physiological or external factors during the measurement process, resulting in abnormalities in some pulse cycle signals, by screening the pulse cycle signals with high similarity, these abnormal signals can be excluded, improving the signal quality used for pulse type analysis, and thus improving the accuracy of pulse type analysis.

[0062] Preferably, the periodic signal data is converted into an image, the image is input into a target detection model that has completed feature training of the starting point, main wave, pre-dicrotic wave, dicrotic wave, and descending isthmus, the corresponding image position information is obtained, and the image position information is converted to obtain the true value of each physiological parameter, specifically including:

[0063] converting the extracted pulse cycle signal data into two-dimensional image data;

[0064] The converted image is input into the target detection model that has completed feature training for the starting point, main wave, pre-dicrotic wave, dicrotic wave, and descending gorge. The target detection model outputs a series of bounding box predictions, each of which contains the target location and the probability of the target appearing at that location.

[0065] According to the target position information of the bounding box, the corresponding position is found in the original pulse cycle signal, and the true value of each physiological parameter is obtained by conversion.

[0066] In this embodiment, the target detection model requires a large amount of labeled data for learning during training. After converting the pulse cycle signal into an image, feature annotation can be performed more conveniently. Key features such as the starting point, main wave, pre-dicrotic wave, dicrotic wave, and descending mid-valley can be clearly marked in the image. This enables the model to more accurately learn the patterns and regularities of these features, improving the model's training effectiveness and generalization capabilities. The target detection model outputs a series of bounding box predictions, each of which contains the target location and the probability of the target appearing at that location. This output method can accurately locate the locations of key features in the pulse cycle image, such as the starting point, main wave, and pre-dicrotic wave. The position information of the bounding box can be used to determine the specific coordinates of these features in the image, and the probability information helps to evaluate the credibility of each bounding box prediction. In practical applications, there may be some noise or interference that causes errors in the model prediction. The probability information can be used to screen out more reliable prediction results and improve the accuracy of feature positioning. According to the target position information of the bounding box, the corresponding position is found in the original pulse cycle signal, and the true value of each physiological parameter is obtained through conversion. This conversion method can accurately map the feature position in the image back to the original signal, thereby obtaining accurate physiological parameters such as pulse amplitude, width, time interval, etc.

[0067] Preferably, the pulse type analysis step includes pulse position classification, pulse rate classification, pulse width classification, pulse length classification, pulse strength classification, fluency classification and tension classification.

[0068] Preferably, the pulse position classification is judged based on the data similarity of the three forces of floating, middle and sinking; the pulse rate classification is calculated based on the data size of the pulse cycle and the sampling frequency; the pulse width classification and pulse length classification are judged based on the approximate number of data of the sensor channels expanded horizontally and vertically and the center point data respectively; the pulse force classification is judged based on the extracted physiological parameters; the fluency classification and tension classification are judged based on the extracted physiological parameters and their mutual relationship.

[0069] In this embodiment, pulse position classification is performed based on the data similarity of the three strengths of floating, middle and sinking, which can accurately reflect the characteristics of the pulse at different depths of the human body surface. The depth of the pulse position is closely related to the prosperity and decline of the human body's qi and blood, and the functional state of the internal organs. Through this data similarity judgment method, it is possible to objectively determine whether the pulse is a floating pulse (felt by light pressing), a sinking pulse (felt by heavy pressing) or a middle pulse, etc., which provides important information about the pulse position for traditional Chinese medicine diagnosis and helps to judge the deficiency and excess, exterior and interior of the disease; pulse rate classification is performed based on the data size and sampling frequency of the pulse cycle, which can accurately obtain the number of pulse beats per minute. Pulse rate is one of the important indicators reflecting the human heart function and the state of qi and blood circulation. Accurate pulse rate calculation helps traditional Chinese medicine practitioners understand the patient's heart rate and judge whether there are abnormalities such as tachycardia and bradycardia; pulse width classification and pulse length classification are based on the horizontal and vertical transmission respectively. By judging the approximate number of sensor channel data and center point data, the width and length characteristics of the pulse can be quantified. Pulse width and length reflect the temporal and spatial distribution of the pulse and are closely related to the circulation of qi and blood and the function of the internal organs. Quantifying these characteristics helps to more accurately describe the pulse morphology. Pulse strength classification based on extracted physiological parameters can objectively assess the strength of the pulse, thereby reflecting the state of qi and blood in the human body. Pulse strength is one of the important indicators in traditional Chinese medicine pulse diagnosis. A strong pulse usually indicates abundant qi and blood, while a weak pulse may indicate qi and blood deficiency. Through pulse strength classification, traditional Chinese medicine practitioners can more accurately understand the patient's qi and blood status and provide a basis for formulating treatment plans. Fluency classification and tension classification are based on extracted physiological parameters and their interrelationships. They can reflect the fluency and tension of the pulse, and thus reflect the state of qi and blood circulation in the human body. A pulse with high fluency usually indicates smooth qi and blood circulation, while a pulse with high tension may indicate obstructed qi and blood circulation or internal organ dysfunction.

[0070] Specifically:

[0071] Pulse position classification:

[0072] The data of a complete cycle of floating, middle, and sinking forces, d', d", are obtained. The similarity between the three is determined. Assume that the similarity between floating and middle data is SIM1, the similarity between floating and re-extraction is SIM2, and the similarity between middle and re-extraction is SIM3.

[0073] When SIM1, SIM2, and SIM3 are all greater than 0.85, the pulse type analysis is normal;

[0074] When SIM1 is greater than 0.85 and much greater than SIM2 and SIM3, the pulse type analysis is floating pulse;

[0075] When SIM3 is greater than 0.85 and much greater than SIM1 and SIM2, the pulse position is classified as deep pulse;

[0076] When SIM3 is approximately 0.85 and much larger than SIM1 and SIM2, the pulse position is classified as a hidden pulse;

[0077] When SIM2 is greater than 0.85 and much greater than SIM1 and SIM3, the pulse position is classified as Pi pulse;

[0078] When SIM1, SIM2, and SIM3 are all much smaller than 0.85, the pulse position is classified as Pi pulse.

[0079] Pulse rate classification:

[0080] Based on the sampling frequency of 100Hz of the pressure sensor corresponding to this invention, the pulse rate calculation formula is: Where T is the data size of one pulse cycle;

[0081] According to the speed value obtained by the above formula, when speed is less than 60, the pulse type analysis is a slow pulse;

[0082] When 60≤speed≤70, the pulse rate is classified as brady;

[0083] When 70 < speed ≤ 90, the pulse rate is classified as normal;

[0084] When 90<speed≤120, the pulse rate is classified as rapid pulse;

[0085] When 120<speed, the pulse rate is classified as rapid pulse.

[0086] Pulse width classification:

[0087] Get the data of the sensor channels in the same row as the center point, and determine the number n of channels whose data sizes are close to the center point (a difference of 5 data values is considered close);

[0088] When n = 5, the pulse width is classified as large pulse;

[0089] When n = 6, the pulse width is classified as a flood pulse;

[0090] When n = 1 or 2, the pulse width is classified as thin pulse;

[0091] When n=3 or 4, the pulse width is classified as normal.

[0092] Pulse length classification:

[0093] Get the data of the sensor channels in the same column as the center point, and determine the number m of data sizes of these channels that are close to the center point S1 (a difference of 5 data amounts is considered close);

[0094] When m < 6, the pulse length is classified as short pulse;

[0095] When m>9, the pulse length is classified as long pulse;

[0096] Otherwise, pulse length was classified as normal.

[0097] Pulse strength classification:

[0098] According to the physiological parameter extraction algorithm, obtain Figure 4 Corresponding parameters (for different pulse types, some parameter points may not be clearly identified or may not exist);

[0099] When the pulse position is classified as floating pulse, and h1, When both are less than a certain set value, or when h3 is less than a certain set value, the pulse strength is classified as weak pulse;

[0100] When the pulse position is classified as sinking pulse, and h1, When both are less than a certain set value, the pulse strength is classified as a weak pulse;

[0101] When the pulse position is classified as floating pulse, and h1, When both are less than a certain set value, or when h3 is less than a certain set value, the pulse strength is classified as a weak pulse (the set value of this pulse type is less than the set value of a weak pulse);

[0102] When h1, When h3 is greater than a certain set value, the pulse strength is classified as a solid pulse;

[0103] Otherwise, the pulse strength is classified as normal.

[0104] Fluency Classification:

[0105] based on Figure 4 Parameters, when the pulse waveform is double-peaked (that is, point c and point e coincide, point e does not exist), and When both are much greater than a certain set value, the fluency is classified as slippery pulse;

[0106] When the pulse waveform is double-peaked (points c and e coincide, and point e does not exist), and When both are slightly greater than a certain set value, the fluency is classified as arterial;

[0107] When h1, When all are less than a certain set value and points d, e, f, and g are not clearly identified, the fluency is classified as wiry pulse;

[0108] Otherwise, fluency was classified as normal.

[0109] Tension classification:

[0110] based on Figure 4 Parameters, when the vertical coordinates of points c and d are much higher than a certain set value, the tension is classified as string pulse;

[0111] When the ordinates of points c and d are slightly higher than a certain set value, the tension is classified as tight pulse;

[0112] When the pulse position is classified as floating pulse, and h1, And when the horizontal coordinate of point d is greater than a certain set value, the tension is classified as leather vein;

[0113] When the pulse position is classified as sinking pulse, and h1, And when the horizontal coordinate of point d is greater than a certain set value, the tension is classified as a prison pulse;

[0114] When the pulse position is classified as floating pulse and h1 is less than a certain set value, the tension is classified as moist pulse;

[0115] Otherwise, the tension was classified as normal.

[0116] Uniformity classification:

[0117] Based on the pulse rate calculated by the cycle extraction algorithm and the pulse rate classification, the parameters T' and speed are obtained, wherein T' is the number of pulse period differences exceeding 20%;

[0118] When speed < 60 and T' > 3, the uniformity is classified as knotted vein;

[0119] When 60≤speed≤100 and T'>3, the uniformity is classified as generation pulse;

[0120] When speed>100 and T'>3, the uniformity is classified as accelerated pulse;

[0121] Otherwise, the uniformity was classified as normal.

[0122] A pulse characteristic analysis system comprising:

[0123] The data acquisition module is used to use a pressure sensor placed at the Cun, Guan, and Chi points of the wrist, apply three sets of different pressure values to the back of the pressure sensor to simulate the three forces of floating, middle, and sinking when taking the pulse in traditional Chinese medicine, and record the data obtained by the pressure sensor;

[0124] The data filtering module is used to filter the data of each channel acquired by the data acquisition module using Butterworth filtering;

[0125] The cycle extraction module is used to smooth the data filtered by the data filtering module, identify the valley value, determine the cycle by the coordinate values of adjacent valley values, and extract the pulse period signal;

[0126] The physiological parameter extraction module is used to convert the periodic signal data extracted by the period extraction module into an image, input the image into the target detection model that has completed feature training for the starting point, main wave, pre-dicrotic wave, dicrotic wave, and descending gorge, obtain the corresponding image position information, and convert the image position information to obtain the true value of each physiological parameter;

[0127] The pulse type analysis module is used to substitute the true values of the various physiological parameters obtained by the physiological parameter extraction module into the judgment criteria of each pulse type to obtain the specific pulse type corresponding to the pulse image.

[0128] A computing device comprises a memory, a processor and a computer program stored in the memory and runnable on the processor. When the processor executes the computer program, the steps of any one of the above-mentioned pulse characteristic analysis methods are implemented.

[0129] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of any one of the above-mentioned pulse characteristic analysis methods.

[0130] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, 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 pulse characteristic analysis method, characterized in that, Specifically include: Data acquisition: A 72-channel pressure sensor was placed at the Cun, Guan, and Chi points on the wrist. Three different pressure values were applied to the back of the pressure sensor to simulate the floating, middle, and sinking forces used in traditional Chinese medicine pulse diagnosis. The data obtained by the pressure sensor was recorded. Data filtering: Butterworth filtering is used to filter the data of each channel; Period extraction: Smooth the filtered data, identify valleys, determine the period through the coordinates of adjacent valleys, and extract the pulse period signal; Physiological parameter extraction: Convert the periodic signal data into an image, input the image into a target detection model that has completed feature training for the starting point, main wave, pre-dicrotic wave, dicrotic wave, and descending gorge, obtain the corresponding image position information, and convert the image position information to obtain the true value of each physiological parameter; Pulse type analysis: Substitute the true values of various physiological parameters into the judgment criteria of each pulse type to obtain the specific pulse type corresponding to the pulse image.

2. A pulse characteristic analysis method according to claim 1, characterized in that, In the data collection step, the seventy-two-channel pressure sensors are arranged in a 12x6 array at the Cun, Guan, and Chi positions of the wrist.

3. A pulse characteristic analysis method according to claim 2, characterized in that, In the data filtering step, the high-frequency cutoff frequency of the Butterworth filter is 18 Hz, the low-frequency cutoff frequency is 5 Hz, the sampling frequency is 100 Hz, and the filter order is 5.

4. A pulse characteristic analysis method according to claim 3, characterized in that, The method of smoothing the filtered data, identifying valleys, determining the period by the coordinate values of adjacent valleys, and extracting the pulse period signal specifically includes: A low-pass filter is used to smooth the filtered data; identifying a valley in the smoothed data, the valley being the lowest point in the pulse waveform; Record the horizontal coordinate values of all identified valley values and calculate the coordinate difference between two adjacent valley values; Analyze the distribution interval and mode of all adjacent valley coordinate differences to determine a pulse cycle; Utilizing the determined pulse cycle, extracting it from the original filtered data to obtain multiple complete pulse cycle signals; The obtained multiple complete pulse cycle signals are compared for similarity, and the group with the highest similarity is selected as the pulse cycle signal for subsequent pulse type analysis.

5. A pulse characteristic analysis method according to claim 4, characterized in that, The periodic signal data is converted into an image, and the image is input into a target detection model that has completed feature training of the starting point, main wave, pre-dicrotic wave, dicrotic wave, and descending gorge, to obtain the corresponding image position information, and convert the image position information to obtain the true value of each physiological parameter, specifically including: converting the extracted pulse cycle signal data into two-dimensional image data; The converted image is input into the target detection model that has completed feature training for the starting point, main wave, pre-dicrotic wave, dicrotic wave, and descending gorge. The target detection model outputs a series of bounding box predictions, each of which contains the target location and the probability of the target appearing at that location. According to the target position information of the bounding box, the corresponding position is found in the original pulse cycle signal, and the true value of each physiological parameter is obtained by conversion.

6. A pulse characteristic analysis method according to claim 5, characterized in that, The pulse type analysis step includes pulse position classification, pulse rate classification, pulse width classification, pulse length classification, pulse strength classification, fluency classification and tension classification.

7. A pulse characteristic analysis method according to claim 6, characterized in that, The pulse position classification is based on the data similarity of the three forces of floating, middle and sinking. The pulse rate classification is calculated based on the data size and sampling frequency of the pulse cycle. The pulse width classification and pulse length classification are respectively based on the approximate number of data from the horizontally and vertically expanded sensor channels and the center point data. The pulse force classification is based on the extracted physiological parameters. The fluency classification and tension classification are based on the extracted physiological parameters and their mutual relationship.

8. A pulse characteristic analysis system, characterized in that: include: The data acquisition module is used to use a pressure sensor placed at the Cun, Guan, and Chi points of the wrist, apply three sets of different pressure values to the back of the pressure sensor to simulate the three forces of floating, middle, and sinking when taking the pulse in traditional Chinese medicine, and record the data obtained by the pressure sensor; The data filtering module is used to filter the data of each channel acquired by the data acquisition module using Butterworth filtering; The cycle extraction module is used to smooth the data filtered by the data filtering module, identify the valley value, determine the cycle by the coordinate values of adjacent valley values, and extract the pulse period signal; The physiological parameter extraction module is used to convert the periodic signal data extracted by the period extraction module into an image, input the image into the target detection model that has completed feature training for the starting point, main wave, pre-dicrotic wave, dicrotic wave, and descending gorge, obtain the corresponding image position information, and convert the image position information to obtain the true value of each physiological parameter; The pulse type analysis module is used to substitute the true values of the various physiological parameters obtained by the physiological parameter extraction module into the judgment criteria of each pulse type to obtain the specific pulse type corresponding to the pulse image.

9. A computing device, characterized in that The method comprises a memory, a processor and a computer program stored in the memory and operable on the processor, wherein the processor implements the steps of the pulse characteristic analysis method according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium, characterized in that The storage medium stores a computer program, which, when executed by a processor, implements the steps of the pulse characteristic analysis method according to any one of claims 1 to 7.

Images