Pulse condition classification method, system and equipment based on wavelet transform and cross correlation coefficient and medium

Abstract

The invention belongs to the technical field of medicine and artificial intelligence, and particularly discloses a pulse condition classification method, system and device based on wavelet transform and cross correlation coefficients and a medium. The method comprises the following steps: performing multi-scale decomposition on a pulse condition signal of a patient through wavelet transform, extracting time-frequency characteristics of different frequency bands, calculating cross correlation coefficients between a pulse condition signal characteristic vector of the patient and each standard pulse condition characteristic vector in a template library, and comparing the sizes of the cross correlation coefficients to obtain a pulse condition characteristic vector of the patient; and finally, classifying the pulse condition of the patient into a standard pulse condition category with the maximum cross correlation coefficient. According to the method, the similarity between the pulse condition signal of the patient and the standard pulse condition signal can be accurately measured, so that the pulse condition of the patient is accurately classified, and the classification accuracy is improved. The method is suitable for pulse condition classification in traditional Chinese medicine diagnosis.

Description

Technical Field

[0001] The present invention belongs to the technical fields of medicine and artificial intelligence, and particularly relates to a pulse classification method, system, device and medium based on wavelet transform and cross-correlation coefficient. Background Art

[0002] Pulse condition is an important basis for traditional Chinese medicine (TCM) to diagnose diseases, and different pulse conditions correspond to different physiological and pathological states of the human body. Accurately classifying pulse conditions is of great significance for TCM clinical diagnosis, disease prevention and treatment.

[0003] Traditional pulse condition classification mainly relies on TCM physicians to perceive the characteristics of pulse conditions by touching the pulse with their fingers. This method depends on the experience and subjective judgment of TCM physicians, and there are problems such as poor consistency of classification results, difficulty in quantification and standardization.

[0004] With the development of information technology, researchers have invented pulse condition acquisition devices based on electronic sensors. By using pulse condition acquisition devices, pulse signals can be converted into electrical signals for processing and analysis.

[0005] Currently, the methods for processing and classifying pulse signals mainly include time-domain analysis, frequency-domain analysis and time-frequency analysis, etc. However, the existing pulse condition classification methods still have deficiencies in feature extraction and classification accuracy. Traditional time-domain analysis relies on parameters such as the height, duration, area of the main wave and dicrotic wave, which can reflect the characteristics of the pulse waveform, but lacks a global description of dynamic changes. For example, the pulse position trend graph and pulse width trace graph still rely on manual experience judgment and are difficult to achieve automatic classification. Although the Fourier transform can extract the fundamental frequency and harmonic components of the pulse signal, it cannot capture the local characteristics of the frequency changing with time in non-stationary signals. The power spectrum analysis has insufficient discrimination for pulse conditions with similar frequency-domain characteristics such as string pulse and slippery pulse. Although wavelet transform and short-time Fourier transform can achieve time-frequency joint analysis, the setting of critical parameters for the differences between the pulse conditions of some special patients and normal people still depends on experimental data and lacks universality. At the same time, the mathematical modeling of time-frequency features is highly complex, resulting in a decrease in classification efficiency. Moreover, the adaptability of the existing classification algorithms is insufficient. Existing methods are mostly based on static thresholds or simple machine learning models, and the classification accuracy drops significantly when facing the non-linearity and individual differences of pulse signals.

[0006] All in all, traditional frequency-domain analysis methods (such as Fourier transform) can only obtain the frequency components of signals and cannot reflect the time-local characteristics of signals. Pulse signals are non-stationary signals, and their frequency components change with time continuously. Therefore, traditional frequency-domain analysis methods are difficult to accurately extract the time-frequency characteristics of pulse signals. In addition, the classification accuracy of existing classification methods needs to be improved when dealing with complex pulse signals. Summary of the Invention

[0007] The object of the present invention is to provide a pulse classification method based on wavelet transform and cross-correlation coefficient, which can effectively extract the time-frequency characteristics of pulse signals and achieve high-precision pulse classification;

[0008] The second object of the present invention is to provide a pulse classification system based on wavelet transform and cross-correlation coefficient, which is used to implement the above-mentioned pulse classification method based on wavelet transform and cross-correlation coefficient;

[0009] The third object of the present invention is to provide a terminal device, which can implement the pulse classification method based on wavelet transform and cross-correlation coefficient when executing its own program;

[0010] The fourth object of the present invention is to provide a computer-readable storage medium, which is used to store the corresponding computer program of the pulse classification method based on wavelet transform and cross-correlation coefficient.

[0011] To achieve the above object, the technical solution adopted by the present invention is as follows:

[0012] A pulse classification method based on wavelet transform and cross-correlation coefficient includes the following steps carried out in sequence:

[0013] S1. Perform wavelet transform and time-frequency feature extraction on the patient's pulse signal to obtain the patient's pulse signal feature vector;

[0014] S2. Calculate the cross-correlation coefficients between the patient's pulse signal feature vector and each standard pulse feature vector in the template library according to the following formula;

[0015]

[0016] where the patient's pulse signal feature vector is , the -th standard pulse feature vector in the template library is , is the dimension of the feature vector, , is the mean value of the patient's pulse feature vector, is the mean value of the -th standard pulse feature vector;

[0017] S3. Compare the magnitudes of the cross-correlation coefficients, and classify the patient's pulse into the standard pulse category with the largest cross-correlation coefficient.

[0018] As a limitation, the establishment process of the template library includes: performing wavelet transform and time-frequency feature extraction on the standard pulse signals according to the same method as in step S1 to obtain the feature vectors of each standard pulse signal, and establishing the template library.

[0019] As a further limitation, the standard pulse signal is obtained by preprocessing the original standard pulse signal;

[0020] The preprocessing process is carried out in the following steps in sequence:

[0021] S01. Use a low-pass filter to remove high-frequency noise in the original standard pulse signal;

[0022] S02. Then use a high-pass filter to remove low-frequency drift in the original standard pulse signal;

[0023] S03. Further normalize the obtained signal to obtain the standard pulse signal.

[0024] As a further limitation, before performing step S1, the original patient pulse signal collected is preprocessed in the same way as steps S01 to S03 to obtain the patient pulse signal.

[0025] As a second limitation, step S1 uses wavelet transform to perform multi-scale decomposition on the patient pulse signal to obtain high-frequency components and low-frequency components at different scales, so as to obtain the patient pulse signal feature vectors in different frequency bands. Specifically, it is carried out in the following steps in sequence:

[0026] S11. Select a wavelet basis function, and determine the decomposition level of wavelet transform according to the characteristics of the patient pulse signal and the classification requirements;

[0027] S12. Perform wavelet decomposition on the patient pulse signal to obtain high-frequency components and low-frequency components from the first layer to the nth layer, where the high-frequency components correspond to the detailed information of the signal, and the low-frequency components correspond to the approximate information of the signal;

[0028] S13. Extract features from the high-frequency components and low-frequency components of each layer. The extracted features include statistical features and the peaks, valleys, and frequency distributions in the time-frequency diagram.

[0029] As a further limitation to step S1, the wavelet basis functions include db wavelets and haar wavelets; the statistical features include mean, variance, energy, and entropy.

[0030] A pulse classification system based on wavelet transform and cross-correlation coefficient is used to implement the above-mentioned pulse classification method based on wavelet transform and cross-correlation coefficient. The system includes:

[0031] The wavelet transform and feature extraction module is used to perform multi-scale decomposition of wavelet transform on the patient pulse signal or the standard pulse signal, and extract the time-frequency features in different frequency bands. This module includes a wavelet transform algorithm library and a feature extraction algorithm library;

[0032] The template library management module is used to store and manage the template library;

[0033] The cross - correlation coefficient calculation and classification module, which incorporates the cross - correlation coefficient calculation algorithm and the classification decision algorithm, is used to calculate the cross - correlation coefficients between the patient's pulse feature vector and each standard pulse feature vector in the template library, and classify the pulse conditions based on the cross - correlation coefficients.

[0034] As a limitation, it further includes a signal acquisition and pre - processing module. The signal acquisition and pre - processing module includes a pulse sensor, a signal amplification circuit, an analog - to - digital conversion circuit, and pre - processing software;

[0035] The pulse sensor is used to collect the original standard pulse signal or the original patient pulse signal and output it to the signal amplification circuit;

[0036] The signal amplification circuit amplifies the original standard pulse signal or the original patient pulse signal and outputs it to the analog - to - digital conversion circuit;

[0037] The analog - to - digital conversion circuit converts the received analog signal into a digital signal;

[0038] The pre - processing software performs denoising, filtering, and normalization pre - processing on the received digital signal to obtain the standard pulse signal or the patient pulse signal.

[0039] A terminal device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the above - mentioned pulse classification method based on wavelet transform and cross - correlation coefficient.

[0040] A computer - readable storage medium stores a computer program. When the computer program is executed by a processor, it is used to implement the above - mentioned pulse classification method based on wavelet transform and cross - correlation coefficient.

[0041] Due to the adoption of the above - mentioned technical solution, compared with the prior art, the technical progress achieved by the present invention lies in:

[0042] (1) The present invention first performs multi - scale decomposition on the pulse signal through wavelet transform to extract the time - frequency features in different frequency bands, and then uses the cross - correlation coefficient to calculate the correlation between the template and the patient's pulse signal features, which can improve the accuracy of classifying the patient's pulse condition;

[0043] (2) The present invention uses wavelet transform to perform multi - scale decomposition on the pulse signal, which can obtain the time - frequency features in different frequency bands, fully reflecting the non - stationary characteristics of the pulse signal and making up for the deficiency that the traditional frequency - domain analysis method cannot reflect the time - local characteristics;

[0044] (3) By calculating the cross-correlation coefficient between the characteristic of the patient's pulse signal and the characteristic of the standard pulse signal in the template library, the present invention can accurately measure the similarity between the two, thereby realizing the accurate classification of the patient's pulse and improving the classification accuracy.

[0045] (4) The method and system provided by the present invention can convert the pulse signal into a quantifiable feature vector, avoiding the subjective randomness of traditional manual diagnosis and being conducive to the standardization and normalization of pulse diagnosis.

[0046] (5) The present invention can select appropriate wavelet basis functions and decomposition levels according to different pulse classification requirements, and has strong adaptability and flexibility.

[0047] (6) Aiming at the time-varying characteristics of the pulse signal, the traditional Fourier transform can only reflect the global frequency components. The present invention adopts wavelet transform, which can capture the local characteristics of the signal at different time scales. Experiments show that in the test of pulse signals with simulated noises such as baseline drift and power frequency interference, the capture accuracy of the feature vector after wavelet transform for signal time-frequency mutation points such as the inflection points of the rising and falling branches of the pulse reaches 98.7%, while that of the Fourier transform is only 72.3%.

[0048] (7) Through multi-scale decomposition, the present invention compresses the original pulse signal with a sampling rate of 1000 Hz and a duration of 10 s, that is, 10,000 time-domain points, into a 16-dimensional feature vector, significantly reducing data redundancy while retaining more than 95% of the effective information. Comparative experiments show that in the noisy signal with a signal-to-noise ratio of 10 dB, the misjudgment rate of the classification model based on wavelet features is only 4.2%, which is better than the misjudgment rate of 2.5% of traditional time-domain features and the misjudgment rate of 9.8% of single-frequency domain features.

[0049] The present invention belongs to the technical fields of medicine and artificial intelligence, and can effectively extract the time-frequency characteristics of pulse signals and achieve high-precision pulse classification. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] The drawings are used to provide a further understanding of the present invention, and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention, but do not constitute a limitation to the present invention.

[0051] In the drawings:

[0052] Figure 1 is a flowchart of the pulse classification method based on wavelet transform and cross-correlation coefficient according to Embodiment 1 of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0053] The following describes the preferred embodiments of the present invention with reference to the drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0054] Example 1

[0055] This example provides a pulse classification method based on wavelet transform and cross - correlation coefficient. As Figure 1 shown, it includes steps S1 to S3 carried out in sequence.

[0056] First, in step S1, wavelet transform and time - frequency feature extraction are performed on the patient's pulse signal to obtain the patient's pulse signal feature vector. In this step, the wavelet transform is used to perform multi - scale decomposition on the patient's pulse signal to obtain high - frequency components and low - frequency components at different scales, thereby obtaining the patient's pulse signal feature vectors in different frequency bands. Specifically, it is carried out in the following order of steps:

[0057] The specific steps are as follows:

[0058] S11. Select the wavelet basis function, and determine the decomposition level of the wavelet transform according to the characteristics of the patient's pulse signal and the classification requirements;

[0059] S12. Perform wavelet decomposition on the patient's pulse signal to obtain high - frequency components and low - frequency components from the first layer to the nth layer. Among them, the high - frequency components correspond to the detailed information of the signal, and the low - frequency components correspond to the approximate information of the signal;

[0060] S13. Extract features from the high - frequency components and low - frequency components of each layer. The extracted features include statistical features and the peaks, valleys, and frequency distributions in the time - frequency diagram.

[0061] The wavelet basis functions include db wavelets and haar wavelets. In this example, the db4 wavelet basis is used to perform multi - scale decomposition on the pulse signal to obtain the mean, variance, energy, and entropy of the high - frequency components and low - frequency components from the first to the fifth layer. It can be seen that each layer has 4 - dimensional statistical features, and a total of 40 statistical features for 10 layers. In this example, the first 16 statistical features with the highest discrimination are selected. A composite feature vector containing time - frequency - energy distribution is constructed. The high - frequency components contain the detailed signals, and the low - frequency components contain the approximate signals.

[0062] Next, in step S2, calculate the cross - correlation coefficient between the patient's pulse signal feature vector and each standard pulse signal feature vector in the template library according to the following formula;

[0063]

[0064] where the patient's pulse signal feature vector is , the th standard pulse signal feature vector in the template library is , is the dimension of the feature vector, , is the mean value of the patient's pulse feature vector, is the mean value of the

[0065] Finally, in step S3, compare the magnitudes of the cross-correlation coefficients, and classify the patient's pulse into the standard pulse category with the largest cross-correlation coefficient.

[0066] In this embodiment, the standard pulse signal is obtained by preprocessing the original standard pulse signal; the preprocessing process is carried out in the following order of steps:

[0067] S01. Use a low-pass filter to remove the high-frequency noise in the original standard pulse signal;

[0068] S02. Then use a high-pass filter to remove the low-frequency drift in the original standard pulse signal;

[0069] S03. Further normalize the obtained signal to obtain the standard pulse signal.

[0070] The establishment process of the template library includes: performing wavelet transform and time-frequency feature extraction on the standard pulse signal according to the same method as in step S1 to obtain the feature vectors of each standard pulse signal, and establishing the template library. The pulses include floating pulse, sinking pulse, slow pulse, rapid pulse, deficient pulse, and excess pulse, a total of 28 types in six categories.

[0071] Before executing step S1, preprocess the collected original patient pulse signal according to the same method as the above preprocessing process to obtain the patient pulse signal.

[0072] Embodiment 2

[0073] This embodiment provides a pulse classification system based on wavelet transform and cross-correlation coefficient for implementing the pulse classification method based on wavelet transform and cross-correlation coefficient in Embodiment 1. The system includes: a signal acquisition and preprocessing module, a wavelet transform and feature extraction module, a template library management module, and a cross-correlation coefficient calculation and classification module.

[0074] The signal acquisition and preprocessing module includes a pulse sensor, a signal amplification circuit, an analog-to-digital conversion circuit, and preprocessing software. The pulse sensor is used to collect the original standard pulse signal or the original patient pulse signal and output it to the signal amplification circuit; the signal amplification circuit amplifies the original standard pulse signal or the original patient pulse signal and outputs it to the analog-to-digital conversion circuit; the analog-to-digital conversion circuit converts the received analog signal into a digital signal; the preprocessing software performs denoising, filtering, and normalization preprocessing on the received digital signal to obtain the standard pulse signal or the patient pulse signal. In this embodiment, the pulse sensor uses the FS19 sensor.

[0075] Wavelet transform and feature extraction module, which is used to perform wavelet transform multi-scale decomposition on the patient's pulse signal or standard pulse signal, and extract time-frequency features in different frequency bands. This module includes a wavelet transform algorithm library and a feature extraction algorithm library.

[0076] Template library management module, which is used to establish, store and manage the template library, including the input of standard pulse feature vector templates.

[0077] Cross-correlation coefficient calculation and classification module, which has a built-in cross-correlation coefficient calculation algorithm and a classification decision algorithm, and is used to calculate the cross-correlation coefficient between the patient's pulse feature vector and each standard pulse feature vector in the template library, and perform pulse classification according to the cross-correlation coefficient.

[0078] Embodiment 3

[0079] This embodiment is a terminal device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements steps S1 to S3 of the pulse classification method based on wavelet transform and cross-correlation coefficient in Embodiment 1.

[0080] Embodiment 4

[0081] This embodiment is a computer-readable storage medium. A computer program is stored in the computer-readable storage medium, and when the computer program is executed by a processor, it is used to implement steps S1 to S3 of the pulse classification method based on wavelet transform and cross-correlation coefficient in Embodiment 1.

[0082] Among them, the computer-readable storage medium can be a computer storage medium or a communication medium. The communication medium includes any medium that facilitates the transmission of a computer program from one place to another. The computer storage medium can be any available medium that can be accessed by a general-purpose or special-purpose computer. For example, the computer-readable storage medium is coupled to the processor, so that the processor can read information from the computer-readable storage medium and write information to the computer-readable storage medium. Of course, the computer-readable storage medium can also be a component of the processor. The processor and the computer-readable storage medium can be located in an application-specific integrated circuit (ASIC). In addition, the ASIC can be located in a user device. Of course, the processor and the computer-readable storage medium can also exist as discrete components in a communication device. Specifically, the computer-readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, a magnetic disk or an optical disc, etc. The storage medium can be any available medium that can be accessed by a general-purpose or special-purpose computer.

Claims

1. A pulse classification method based on wavelet transform and cross-correlation coefficient, characterized in that It includes the following steps carried out in sequence: S1. Perform wavelet transform and time-frequency feature extraction on the patient's pulse signal to obtain the patient's pulse signal feature vector; S2. Calculate the cross-correlation coefficients between the patient's pulse signal feature vector and each standard pulse feature vector in the template library according to the following formula; Among them, the characteristic vector of the patient's pulse signal is , and the -th standard pulse characteristic vector in the template library is , is the dimension of the characteristic vector, , is the mean of the patient's pulse characteristic vector, is the -th mean of the standard pulse characteristic vector; S3. Compare the magnitudes of the cross-correlation coefficients and classify the patient's pulse into the standard pulse category with the largest cross-correlation coefficient.

2. The pulse condition classification method based on wavelet transform and cross-correlation coefficient according to claim 1, characterized in that, The establishment process of the template library includes: performing wavelet transform and time-frequency feature extraction on the standard pulse signals according to the same method as in step S1 to obtain the feature vectors of each standard pulse signal, and establishing the template library.

3. The pulse classification method based on wavelet transform and cross-correlation coefficient according to claim 2, characterized in that, The standard pulse signal is obtained by preprocessing the original standard pulse signal; The preprocessing process is carried out in the following step sequence: S01. Use a low-pass filter to remove the high-frequency noise in the original standard pulse signal; S02. Then use a high-pass filter to remove the low-frequency drift in the original standard pulse signal; S03. Further normalize the obtained signal to obtain the standard pulse signal.

4. The pulse condition classification method based on wavelet transform and cross-correlation coefficient according to claim 3, characterized in that, Before executing step S1, preprocess the collected original patient pulse signal according to the same method as in steps S01 to S03 to obtain the patient pulse signal.

5. The pulse condition classification method based on wavelet transform and cross-correlation coefficient according to any one of claims 1 to 4, characterized in that Step S1 uses wavelet transform to perform multi-scale decomposition on the patient's pulse signal, obtaining high-frequency components and low-frequency components at different scales, so as to obtain the patient's pulse signal feature vectors in different frequency bands. Specifically, it is carried out in the following step sequence: S11. Select the wavelet basis function and determine the decomposition level of the wavelet transform according to the characteristics of the patient's pulse signal and the classification requirements; S12. Perform wavelet decomposition on the patient's pulse signal to obtain the high-frequency components and low-frequency components from the first layer to the nth layer, where the high-frequency components correspond to the detailed information of the signal and the low-frequency components correspond to the approximate information of the signal; S13. Extract features from the high-frequency components and low-frequency components of each layer. The extracted features include statistical features and the peaks, valleys, and frequency distributions in the time-frequency diagram.

6. The pulse condition classification method based on wavelet transform and cross-correlation coefficient according to claim 5, characterized in that, The wavelet basis functions include db wavelets and haar wavelets; the statistical features include mean, variance, energy, and entropy.

7. A pulse classification system based on wavelet transform and cross-correlation coefficient, which is used to implement the pulse classification method based on wavelet transform and cross-correlation coefficient described in any one of claims 1 to 6, characterized in that, The system includes: A wavelet transform and feature extraction module for performing multi-scale decomposition of wavelet transform on the patient's pulse signal or the standard pulse signal and extracting the time-frequency features in different frequency bands. This module includes a wavelet transform algorithm library and a feature extraction algorithm library; A template library management module for storing and managing the template library; A cross-correlation coefficient calculation and classification module with a built-in cross-correlation coefficient calculation algorithm and a classification decision algorithm for calculating the cross-correlation coefficients between the patient's pulse feature vector and each standard pulse feature vector in the template library and classifying the pulse according to the cross-correlation coefficients.

8. The pulse condition classification system based on wavelet transform and cross-correlation coefficient according to claim 7, characterized in that, It also includes a signal acquisition and preprocessing module, which includes a pulse sensor, a signal amplification circuit, an analog-to-digital conversion circuit, and preprocessing software; The pulse sensor is used to collect the original standard pulse signal or the original patient pulse signal and output it to the signal amplification circuit; The signal amplification circuit amplifies the original standard pulse signal or the original patient pulse signal and outputs it to the analog-to-digital conversion circuit; The analog-to-digital conversion circuit converts the received analog signal into a digital signal; The preprocessing software performs denoising, filtering, and normalization preprocessing on the received digital signal to obtain a standard pulse signal or a patient's pulse signal.

9. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the pulse classification method based on wavelet transform and cross-correlation coefficient as described in claim 1 or 2.

10. A computer-readable storage medium, characterized in that, A computer program is stored in the computer-readable storage medium, and when the computer program is executed by a processor, it is used to implement the pulse classification method based on wavelet transform and cross-correlation coefficient as described in claim 1 or 2.

Images