Three-dimensional pulse signal quantitative analysis method and system

Abstract

The invention provides a three-dimensional pulse signal quantitative analysis method and system, and belongs to the technical field of signal processing. Through a binocular vision pulse image acquisition system, continuous contact membrane images of a traditional Chinese medicine pulse condition training instrument at a radial artery are acquired, a three-dimensional pulse signal is reconstructed, ROI region selection is performed, and a pulse condition is obtained. Obtaining a central region containing effective components of the three-dimensional pulse signal; segmenting the three-dimensional pulse signal into a plurality of two-dimensional slices by utilizing dynamic waveform slices to obtain a dynamic slice sequence; a composite sine cascade modeling model is used for fitting two-dimensional signals with sine similarity characteristics in a dynamic slice sequence, the pulse length, the pulse width and the pulse height of the pulse signals are determined, dynamic waveform slices are used, the three-dimensional pulse signals are subjected to dimensionality reduction to two-dimensional slices, the data size is reduced, and through composite sine cascade modeling, the accuracy of pulse signal processing is improved. Compared with a one-dimensional time sequence mode, the pulse analysis method has the advantage that the pulse analysis effect is effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of signal processing, and particularly to a three-dimensional pulse signal quantization analysis method and system. Background Art

[0002] The pulse signal is an important physiological signal, which reflects the activities of the cardiovascular system. The main source of the pulse signal is the pulsation of the heart, that is, the process of the heart pushing oxygen and blood to the whole body every time it beats. These pulsations are propagated by the arteries and finally generate a perceptible pulse signal at specific positions in the body. Research shows that the pulse signal carries rich physiological and pathological information, and the physiological information extracted from the pulse signal can be used to evaluate the heart health status. Although the existing pulse signal analysis methods have made certain progress, there are still some deficiencies. The traditional pulse signal processing methods mainly focus on one-dimensional time series.

[0003] However, the pulse signal processing method for one-dimensional time series cannot fully reflect the complexity and multi-dimensional characteristics of the pulse signal, resulting in poor pulse analysis effect. Summary of the Invention

[0004] The present invention provides a three-dimensional pulse signal quantization analysis method and system to solve the defect that the existing technology has poor effect in processing pulse signals relying on one-dimensional time series.

[0005] In a first aspect, the present invention provides a three-dimensional pulse signal quantization analysis method, including: Collecting continuous contact film images at the radial artery of a traditional Chinese medicine pulse training instrument through a binocular vision pulse image acquisition system, and reconstructing a three-dimensional pulse signal; Selecting an ROI region for the reconstructed three-dimensional pulse signal to obtain a central region containing the effective components of the three-dimensional pulse signal; Using dynamic waveform slicing to divide the three-dimensional pulse signal in the central region into multiple two-dimensional slices to obtain a dynamic slice sequence; Fitting the two-dimensional signals with sinusoidal similarity characteristics in the dynamic slice sequence using a composite sine cascade modeling model; Determining the pulse length, pulse width, and pulse height of the pulse signal according to the frequency, phase, and amplitude of the sine waveform after the fitting is completed.

[0006] According to the three-dimensional pulse signal quantization analysis method provided by the present invention, the selecting an ROI region for the reconstructed three-dimensional pulse signal to obtain a central region containing the effective components of the three-dimensional pulse signal includes: Representing the three-dimensional pulse signal using the pulse length grid length, pulse width grid length, and pulse height amplitude to obtain a three-dimensional pulse signal image; Based on the indicator function, filter the ROI region containing the effective components of the three-dimensional pulse signal in the three-dimensional pulse signal image as the central region.

[0007] According to a three-dimensional pulse signal quantization analysis method provided by the present invention, the step of using dynamic waveform slicing to divide the three-dimensional pulse signal in the central region into multiple two-dimensional slices to obtain a dynamic slice sequence includes: Select the dimension of the pulse width as the slicing direction according to the characteristics of the three-dimensional pulse signal in the central region; According to the dynamic change of the three-dimensional pulse signal in the central region in terms of pulse width and in combination with the grid distribution during the reconstruction of the three-dimensional pulse signal, adjust the position and range of the slices; Based on the position and range of the slices, extract two-dimensional slices from the three-dimensional pulse signal at each slice point; Arrange all the extracted two-dimensional slices in the order of the slicing direction to form a dynamic slice sequence.

[0008] According to a three-dimensional pulse signal quantization analysis method provided by the present invention, the step of using a composite sine cascade modeling model to fit the two-dimensional signals with sine similarity characteristics in the dynamic slice sequence includes: Construct a composite sine cascade modeling model based on the slice characteristics in the dynamic slice sequence; Combined with parameter boundary constraint conditions, solve for the parameters of the composite sine cascade modeling model; Use an optimization algorithm to determine the optimal characteristic parameters of each sine function among the solved parameters to minimize the fitting error and obtain the optimal model parameters.

[0009] According to a three-dimensional pulse signal quantization analysis method provided by the present invention, the step of constructing a composite sine cascade modeling model based on the slice characteristics in the dynamic slice sequence includes: Find all the wave troughs of the dynamic slice sequence, perform threshold processing on the dynamic slice sequence, and extract the rising edge and the falling edge; Based on the rising edge and the falling edge, extract a two-dimensional slice array from the dynamic slice sequence from any current wave trough position to the next wave trough position; Based on the curve characteristics of all the two-dimensional slice arrays, respectively use a sine function model for fitting to construct a composite sine cascade modeling model.

[0010] According to a three-dimensional pulse signal quantization analysis method provided by the present invention, before combining parameter boundary constraint conditions to solve for the parameters of the composite sine cascade modeling model, it further includes: Initialize each parameter to obtain the corresponding initial value; Set boundary constraint conditions for each of the parameters based on the initial values.

[0011] According to a three-dimensional pulse signal quantization analysis method provided by the present invention, the use of an optimization algorithm to determine the optimal characteristic parameters of each sine function among the parameters to be solved, minimize the fitting error, and obtain the optimal model parameters includes: Determine the objective function to be minimized during the optimization solution process; Select the YS-PSO optimization algorithm to optimize the parameters to be solved, minimize the fitting error, obtain the optimal model parameters, and organize all the optimal model parameters into a feature vector set.

[0012] According to a three-dimensional pulse signal quantization analysis method provided by the present invention, the composite sine cascade modeling model is: ; Wherein, is the number of sine function models; is the composite sine cascade modeling model; is the total length of each two-dimensional slice pulse length divided by the number of grids; is the current sine function model, represents the th sine function model amplitude; represents the th sine function model frequency; represents the th sine function model phase shift; represents the th sine function model vertical offset.

[0013] According to a three-dimensional pulse signal quantization analysis method provided by the present invention, the reconstruction of the three-dimensional pulse signal includes: Filter, binarize, and extract feature points from the collected image pulse signal, and reconstruct the three-dimensional pulse signal in combination with the binocular vision imaging principle.

[0014] In a second aspect, the present invention also provides a three-dimensional pulse signal quantization analysis system, including: A reconstruction module for collecting continuous contact film images at the radial artery of a traditional Chinese medicine pulse training instrument through a binocular vision pulse image collection system and reconstructing a three-dimensional pulse signal; A selection module for selecting an ROI region from the reconstructed three-dimensional pulse signal to obtain a central region containing the effective components of the three-dimensional pulse signal; A slicing module for using dynamic waveform slicing to divide the three-dimensional pulse signal in the central region into multiple two-dimensional slices to obtain a dynamic slice sequence; A fitting module, configured to fit the two-dimensional signals with sinusoidal similarity features in the dynamic slice sequence by using a composite sine cascade modeling model; A determination module, configured to determine the pulse length, pulse width, and pulse height of the pulse signal according to the frequency, phase, and amplitude of the sine waveform after the fitting is completed.

[0015] In a third aspect, the present invention further provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, the three-dimensional pulse signal quantization analysis method described in any one of the above is implemented.

[0016] In a fourth aspect, the present invention further provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the three-dimensional pulse signal quantization analysis method described in any one of the above is implemented.

[0017] In a fifth aspect, the present invention further provides a computer program product, including a computer program. When the computer program is executed by a processor, the three-dimensional pulse signal quantization analysis method described in any one of the above is implemented.

[0018] A three-dimensional pulse signal quantization analysis method and system provided by the present invention collect continuous contact film images at the radial artery of a traditional Chinese medicine pulse training instrument through a binocular vision pulse image acquisition system, and reconstruct a three-dimensional pulse signal; perform ROI region selection on the reconstructed three-dimensional pulse signal to obtain a central region containing the effective components of the three-dimensional pulse signal; use dynamic waveform slicing to divide the three-dimensional pulse signal in the central region into multiple two-dimensional slices to obtain a dynamic slice sequence; use a composite sine cascade modeling model to fit the two-dimensional signals with sinusoidal similarity features in the dynamic slice sequence; determine the pulse length, pulse width, and pulse height of the pulse signal according to the frequency, phase, and amplitude of the sine waveform after the fitting is completed. By using dynamic waveform slicing, the three-dimensional pulse signal is reduced to two-dimensional slices, reducing the data volume, and through composite sine cascade modeling, the pulse signal characteristics are accurately fitted. Compared with the one-dimensional time series, the pulse analysis effect is effectively improved through the pulse length, pulse width, and pulse height. Description of the Drawings

[0019] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0020] Figure 1 It is a flowchart of the three-dimensional pulse signal quantization analysis method provided in this embodiment; Figure 2 Schematic diagram of the reconstructed three-dimensional pulse signal; Figure 3 Schematic diagram of the three-dimensional pulse signal distribution provided by this embodiment; Figure 4 Schematic diagram of the curve characteristics of the two-dimensional slice array provided by this embodiment; Figure 5 Dynamic slice sequence diagram provided by this embodiment; Figure 6 Fitting effect diagram of a certain slice provided by this embodiment; Figure 7 Fitting effect diagram of the entire dynamic slice sequence provided by this embodiment; Figure 8 Partial enlarged schematic diagram of the fitting effect diagram of the entire dynamic slice sequence provided by this embodiment; Figure 9 Fitting residual diagram of the entire dynamic slice sequence provided by this embodiment; Figure 10 Average model diagram of a certain pulse image provided by this embodiment; Figure 11 Schematic diagram of the structure of the three-dimensional pulse signal quantitative analysis system provided by this embodiment; Figure 12 Schematic diagram of the structure of the electronic device provided by this embodiment. Detailed implementation manners

[0021] To make the objectives, technical solutions and advantages of the present invention clearer, the technical solutions in the present invention will be clearly and completely described below with reference to the accompanying drawings in the present invention. Apparently, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present invention without making creative efforts shall fall within the protection scope of the present invention.

[0022] Figure 1 Schematic diagram of the flow of the three-dimensional pulse signal quantitative analysis method provided by this embodiment, Figure 2 Schematic diagram of the reconstructed three-dimensional pulse signal, Figure 3 Schematic diagram of the three-dimensional pulse signal distribution provided by this embodiment.

[0023] As Figure 1 shown, the three-dimensional pulse signal quantitative analysis method provided by the embodiment of the present invention mainly includes the following steps: 101. Through the binocular vision pulse image acquisition system, collect the continuous contact film images at the radial artery of the traditional Chinese medicine pulse training instrument, and reconstruct the three-dimensional pulse signal.

[0024] In a specific implementation process, the binocular vision pulse image acquisition system is an innovative system that combines binocular vision technology and biomedical detection, mainly used for accurately extracting the three-dimensional dynamic information of the human pulse. Through the binocular vision imaging principle, the spatial deformation information of the pulse film is calculated, and the SIFT feature matching algorithm is used to match the feature points of the left and right images, reconstructing the three-dimensional dynamic model of the pulse. The measurement accuracy of the binocular system is verified using a laser displacement sensor, and the error analysis results show that the system has high reliability. During reconstruction, specifically, preprocessing such as filtering, binarization, and feature point extraction is performed on the acquired image pulse signal, and combined with the binocular vision imaging principle, that is, depth information is obtained through images from two different perspectives, and the three-dimensional pulse signal is reconstructed. As Figure 2 shown, it is a schematic diagram of the reconstructed three-dimensional pulse signal.

[0025] The core is divided into two parts. One is camera calibration and synchronous acquisition, and the other is signal processing and accuracy optimization. Among them, camera calibration and synchronous acquisition include: using Zhang Zhengyou's calibration method for binocular camera calibration, simplifying the operation process and reducing the implementation difficulty, initializing the CMOS sensor and synchronously acquiring data based on the FPGA and I2C interface, optimizing the data processing efficiency through the AXI HP bus and DDR3 cache, and combining hardware external triggering and software control to ensure frame-level synchronization of the binocular camera (error < 1ms). And signal processing and accuracy optimization include: designing a Butterworth low-pass filter to separate the actual pulse waveform from the laser signal and reduce noise interference; configuring enhancement coefficients for the overlapping area and non-overlapping area respectively to improve the image detail processing ability.

[0026] 102. Select the ROI region for the reconstructed three-dimensional pulse signal to obtain the central region containing the effective components of the three-dimensional pulse signal.

[0027] Using the pulse length grid length, pulse width grid length, and pulse height amplitude, represent the three-dimensional pulse signal to obtain the three-dimensional pulse signal image, as Figure 3 shown. Use T ( x , y , z ) to represent the three-dimensional pulse signal, x represents the pulse length grid length, y represents the pulse width grid length, z represents the pulse height amplitude.

[0028] Observing the three-dimensional pulse signal as Figure 3 it is found that the effective components of the signal are mainly concentrated in the central region, and this region is defined as the ROI region. Based on the indicator function, screen the ROI region in the three-dimensional pulse signal image that contains the effective components of the three-dimensional pulse signal as the central region, as shown in formula (1): (1) In the formula, T ROI represents the three-dimensional pulse signal after retaining the ROI region, I ( z > θ ) represents the indicator function, which retains its value when z > θ , and is 0 otherwise, θ is the set threshold.

[0029] Thus, the ROI region containing the effective components is effectively extracted.

[0030] 103. Using dynamic waveform slicing, the three-dimensional pulse signal in the central region is segmented into multiple two-dimensional slices to obtain a dynamic slice sequence.

[0031] Dynamic slicing is a technique for signal segmentation and analysis based on the dynamic characteristics of the signal, aiming to simplify the signal processing process through dimensionality reduction, while retaining key feature information and improving the efficiency and accuracy of signal analysis.

[0032] Specifically, according to the characteristics of the three-dimensional pulse signal in the central region, a suitable pulse width dimension is selected as the slicing direction, and the slice position and range are set according to the dynamic changes of the three-dimensional pulse signal in the pulse width dimension combined with the reconstructed grid distribution. By extracting the two-dimensional slices of each slice point, a dynamic slice sequence is formed, which completely presents the dynamic changes of the signal, as Figure 4 is a schematic diagram of the curve characteristics of the two-dimensional slice array, as Figure 5 shown is a diagram of the dynamic slice sequence.

[0033] 104. Use the composite sine cascade modeling model to fit the two-dimensional signals with sine similarity characteristics in the dynamic slice sequence.

[0034] The composite sine cascade modeling model is a tool for fitting data with complex characteristics. By cascading multiple sine functions together, a composite model is formed, which can accurately fit complex signals with different frequencies, phases, and amplitudes. The core idea of this module is to combine multiple sine waveforms to achieve higher-precision fitting of complex waveforms.

[0035] Specifically, each two-dimensional signal in the dynamic slice sequence is fitted through the composite sine cascade modeling model, and the waveform that best fits the two-dimensional signal is selected from among numerous sine waveforms.

[0036] As Figure 6 shown, it is the fitting effect diagram of a certain slice, Figure 7 is the fitting effect diagram of the entire dynamic slice sequence, Figure 8 is a partial enlarged schematic diagram of the fitting effect diagram of the entire dynamic slice sequence,Figure 9 It is the fitting residual plot for the entire dynamic slice sequence.

[0037] Through Figures 6 - 9 The corresponding indicators obtained from the fitting effect diagram are shown in Table 1: Table 1

[0038] Among them, MSE is the mean square error, RMSE is the root mean square error, and MAE is the mean absolute error.

[0039] 105. Determine the pulse length, pulse width, and pulse height of the pulse signal according to the frequency, phase, and amplitude of the sine waveform after fitting.

[0040] After obtaining the waveform that best fits the two-dimensional slice through fitting, the frequency, phase, and amplitude of the fitted sine waveform can be determined. Taking the frequency as the pulse length, the phase as the pulse width, and the amplitude as the pulse height, the three-dimensional characteristics of the pulse signal are successfully obtained.

[0041] Due to the adoption of the dynamic waveform slicing technology in the present invention, the problem of the complex and changeable three-dimensional pulse signal is effectively solved. By reducing the three-dimensional pulse signal to a two-dimensional signal, the data volume is significantly reduced, and the calculation efficiency is improved. This technology has high flexibility and can flexibly select the time points and ranges of slicing according to the dynamic changes of the signal to adapt to the analysis requirements in different scenarios.

[0042] Combined with the composite sine cascade modeling method, it can effectively fit the two-dimensional slice pulse signal. By accurately modeling the changes of each signal, the detailed characteristics of the pulse signal can be captured, thereby improving the accuracy of signal quantization analysis. In addition, this model has good scalability and can increase or decrease the number of sine functions according to needs to adapt to pulse signals of different complexities.

[0043] Compared with the analysis of the one-dimensional time series pulse signal, it can analyze complex pulse patterns from multiple angles, effectively improving the pulse analysis effect and accuracy.

[0044] Furthermore, on the basis of the above embodiments, in this embodiment, the dynamic waveform slicing is used to divide the three-dimensional pulse signal in the central region into multiple two-dimensional slices to obtain a dynamic slice sequence, including: Select the dimension of the pulse width as the slicing direction according to the characteristics of the three-dimensional pulse signal in the central region.

[0045] According to the dynamic changes of the three-dimensional pulse signal in the central region in terms of pulse width and combined with the grid distribution during the reconstruction of the three-dimensional pulse signal, adjust the position and range of the slice. Specifically, according to the dynamic changes of the three-dimensional pulse signal in terms of pulse width and combined with the grid distribution during reconstruction, set the position and range of the slice to the size of the grid, that is, assuming the original grid distribution isa × b , where a represents the pulse length grid length and b represents the pulse width grid length. According to the pulse width grid length b, the signal is segmented into b slices, and the range of each slice is 1 / b.

[0046] Then the slice position is expressed as (2): (2) where, k is the index of the slice, is the k th slice at the y dimension position.

[0047] Based on the position and range of the slice, at each slice point , a two-dimensional slice is extracted from the three-dimensional signal, which is composed of the values of the three-dimensional signal at this position. The specific formula is expressed as (3): (3) Arrange all the extracted two-dimensional slices in the slice direction order to form a dynamic slice sequence, as shown in (4): (4) where, b is the number of slices.

[0048] Furthermore, on the basis of the above embodiments, in this embodiment, a composite sine cascade modeling model is used to fit the two-dimensional signals with sine similarity characteristics in the dynamic slice sequence, including: constructing a composite sine cascade modeling model based on the slice characteristics in the dynamic slice sequence; combining parameter boundary constraint conditions to solve the parameters of the composite sine cascade modeling model; using an optimization algorithm to determine the optimal characteristic parameters of each sine function in the solved parameters, minimizing the fitting error, and obtaining the best model parameters.

[0049] Specifically, constructing a composite sine cascade modeling model based on the slice characteristics in the dynamic slice sequence includes: finding all the wave valleys of the dynamic slice sequence, performing threshold processing on the dynamic slice sequence, and extracting the rising edge and the falling edge; for the convenience of understanding, the following uses data ( k ) to replace the dynamic slice sequence, as shown in formula (5): (5) And extracting the rising edge and the falling edge, the formulas are as shown in (6)(7): (6) (7) where, Data isdata(k) The entire dynamic slice sequence array, BG stores the positions of the rising and falling edges of the square wave, i.e., the positions of the wave troughs; j is the index variable traversing each element of the array dynamic slice sequence data ( k ); line is data ( k )'s length.

[0050] Based on the rising and falling edges, extract from the dynamic slice sequence data ( k ) a two-dimensional slice array s from any current wave trough position BG ( n ) to the next wave trough position BG (n + 1), as shown in formulas (8) and (9): (8) (9) Among them, BG ( n ) and BG ( n + 1) respectively represent BG the n th and n + 1th elements in the array; i is BG the specific value inside the Data ( i ) is the entire dynamic slice sequence from position BG ( n ) to position BG ( n + 1) of the array, s is the two-dimensional slice array of the dynamic slice sequence, Length is the length of this two-dimensional array.

[0051] Based on the curve characteristics of all two-dimensional slice arrays, respectively use the sine function model for fitting to construct a composite sine cascade modeling model. The curve characteristics of the two-dimensional slice array are as Figure 4 shown, so use the sine function model for fitting, and the specific formula is as (10): (10) Among them, s ( τ ) is the sine function model, A is the amplitude of the model; B is the model frequency; C is the model phase shift, D is the model vertical offset,τ is the ratio of the length of the subarray to the number of pulse length grids.

[0052] The entire dynamic slice sequence is fitted with a composite sinusoidal cascade model function. The specific formula is as follows (11): (11) in, is the number of sinusoidal function models; It is a composite sinusoidal cascade modeling model; It is the total length of each 2D slice vein divided by the number of grids; is the current sine function model, Indicates The amplitude of the sine function model, that is, the maximum deviation of the sine wave; Indicates The frequency of the sine function model controls the period of the sine wave. The period is , the greater the frequency, the faster the waveform repeats; Indicates The phase shift of a sine function model represents the translation of the waveform along the x-axis, with positive values ​​representing translation to the left and negative values ​​representing translation to the right; Indicates The vertical displacement of the sine function model indicates the translation of the waveform along the y-axis, changing the centerline position of the waveform.

[0053] Specifically, before solving the parameters of the composite sinusoidal cascade model, combined with the parameter boundary constraints, the following steps are also included: initializing each parameter. Before optimizing, each parameter needs to be initialized. , , , Choose a reasonable initial value.

[0054] for The formula for the initialization value is as follows (12): (12) Among them, max( s m ) is the m Subarray s The maximum value, min( s m ) is the m Subarray s The minimum value of , the sub-array is a two-dimensional slice array.

[0055] for The formula for the initialization value is as follows (13): (13) Among them, is the grid interval, that is, the ratio of the total length of the grid to the total number of grids.

[0056] For The formula for the initialized value is as in (14): (14) For The formula for the initialized value is as in (15): (15) Among them, represents the th two-dimensional slice array.

[0057] To ensure the rationality and physical meaning of the parameters, it is necessary to set boundary constraint conditions for each parameter in the model based on the initial values, as follows: Amplitude constraint: A m ∈ [0, 2 ⋅ Am ; Frequency constraint: B m ∈ B min , B max , where B min = , B max = ; Phase constraint: C m ∈ [0, 2π] (uniqueness within the period); Offset constraint: D m ∈ min(s m ) , max( s m )].

[0058] Specifically, determine the optimal characteristic parameters of each sine function among the parameters to be solved, minimize the fitting error, and obtain the best model parameters, that is, use an optimization algorithm to find the optimal characteristic parameters, and solve the parameters of each sine function in the composite sine cascade modeling model through the optimization algorithm , , , , so that the model minimizes the fitting error, thereby obtaining the best model parameters, including: Define the objective function, which is the function to be minimized during the solution process, usually the mean squared error (MSE), that is, the average of the sum of the squares of the differences between the model output and the actual data. The formula is as shown in (16): (16) where l is an index variable that traverses the dynamic slice elements of the array.

[0059] Select the parameters to be optimized by the YS-PSO optimization algorithm, minimize the fitting error, obtain the optimal model parameters, and organize all the optimal model parameters into a feature vector set.

[0060] For example, the pulse length grid length and the pulse width grid length are 115 and 114 respectively. That is, the original three-dimensional pulse signal is a two-dimensional array with a data structure of 115×114 elements, and the memory occupancy is about 98KB. After using the three-dimensional pulse signal quantization analysis method based on dynamic waveform slicing and composite sine cascade modeling, the data is optimized to a two-dimensional array of 115×4 elements, and the memory occupancy is reduced to 5KB. According to the calculation formula of the data compression ratio as shown in (17): (17) It is calculated that the data compression ratio reaches 24.5, which means that the size of the compressed data is 1 / 24.5 of the original data size. Or rather, the data is compressed by about 94.89%. This dimensionality reduction process not only simplifies the data structure but also retains the core features of the pulse signal, providing a more refined data basis for pulse waveform analysis and pulse condition recognition.

[0061] Using the parameters estimated by the three-dimensional pulse signal quantization analysis method based on dynamic waveform slicing and composite sine cascade modeling of the present invention, obtain the average model and feature vector of a certain pulse condition. The average model is as Figure 10 shown.

[0062] Furthermore, in this embodiment, after obtaining the pulse signal, it also includes converting the processed pulse signal into a classifiable pulse condition label. Its core process includes feature extraction, feature selection, model training, and classification result output. Specifically: After passing the two-sample Kolmogorov-Smirnov (KS) test, organize the optimal model parameters obtained by fitting the composite sine cascade modeling model into a feature vector set. In this example, the size of the feature vector is 24840×4.

[0063] For the tested feature vectors, 80% of the feature vectors can be used as the training set, and 20% of the feature vectors can be used as the test set.

[0064] Select different training classification models: Probabilistic Neural Network (PNN), Decision Tree (DT), Random Forest (RF) classifier, and a Convolutional Neural Network - Bidirectional Long Short-Term Memory Network with Attention mechanism (CNN-BiLSTM-Attention) model.

[0065] Use the new variables formed by different combinations of feature parameters as inputs to train the classifier, perform pulse condition recognition on the collected data pulses, and select the accuracy rate ( Ac ), unweighted average precision ( Pr ), unweighted average recall rate ( Re ), and unweighted average F1-score ( F1 ), and kappa coefficient ( Kap ) as evaluation metrics for the classification performance of the classifier, as shown in Table 2: Table 2

[0066] Based on the results in Table 2, the optimal model can be determined for the conversion of pulse signals, ensuring the effect and accuracy of signal conversion.

[0067] Based on the same general inventive concept, the present invention also protects a three-dimensional pulse signal quantization analysis system. The three-dimensional pulse signal quantization analysis system described below can be mutually corresponded and referred to with the three-dimensional pulse signal quantization analysis method described above.

[0068] Figure 11 It is a schematic structural diagram of the three-dimensional pulse signal quantization analysis system provided in this embodiment.

[0069] As Figure 11 shown, a three-dimensional pulse signal quantization analysis system provided in this embodiment includes: A reconstruction module 1101, configured to collect continuous contact film images at the radial artery of the traditional Chinese medicine pulse condition training instrument through a binocular vision pulse image acquisition system, and reconstruct a three-dimensional pulse signal; A selection module 1102, configured to select the ROI region for the reconstructed three-dimensional pulse signal to obtain a central region containing the effective components of the three-dimensional pulse signal; A slicing module 1103, configured to use dynamic waveform slicing to divide the three-dimensional pulse signal in the central region into multiple two-dimensional slices to obtain a dynamic slice sequence; A fitting module 1104, configured to fit the two-dimensional signals with sinusoidal similarity features in the dynamic slice sequence using a composite sine cascade modeling model; A determination module 1105, configured to determine the pulse length, pulse width, and pulse height of the pulse signal according to the frequency, phase, and amplitude of the sine waveform after fitting is completed.

[0070] Figure 12 It is a schematic structural diagram of the electronic device provided in this embodiment.

[0071] As Figure 12 shown, the electronic device may include: a processor 1210, a communications interface 1220, a memory 1230, and a communication bus 1240. Among them, the processor 1210, the communications interface 1220, and the memory 1230 communicate with each other through the communication bus 1240. The processor 1210 may call the logical instructions in the memory 1230 to execute the three-dimensional pulse signal quantization analysis method. The method includes: collecting continuous contact film images at the radial artery of the traditional Chinese medicine pulse training instrument through a binocular vision pulse image acquisition system, and reconstructing a three-dimensional pulse signal; performing ROI region selection on the reconstructed three-dimensional pulse signal to obtain a central region containing the effective components of the three-dimensional pulse signal; using dynamic waveform slicing to divide the three-dimensional pulse signal in the central region into multiple two-dimensional slices to obtain a dynamic slice sequence; using a composite sine cascade modeling model to fit the two-dimensional signals with sine similarity characteristics in the dynamic slice sequence; determining the pulse length, pulse width, and pulse height of the pulse signal according to the frequency, phase, and amplitude of the sine waveform after the fitting is completed.

[0072] In addition, when the logical instructions in the above-mentioned memory 1230 are implemented in the form of software function units and sold or used as independent products, they may be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, may be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The foregoing storage medium includes: various media such as a USB flash drive, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk, or an optical disc that can store program codes.

[0073] On the other hand, the present invention also provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the three-dimensional pulse signal quantization analysis method provided by the above-mentioned various methods. The method includes: collecting continuous contact film images at the radial artery of a traditional Chinese medicine pulse training instrument through a binocular vision pulse image acquisition system, and reconstructing a three-dimensional pulse signal; performing ROI region selection on the reconstructed three-dimensional pulse signal to obtain a central region containing the effective components of the three-dimensional pulse signal; using dynamic waveform slicing to divide the three-dimensional pulse signal in the central region into multiple two-dimensional slices to obtain a dynamic slice sequence; using a composite sine cascade modeling model to fit the two-dimensional signals with sine similarity characteristics in the dynamic slice sequence; determining the pulse length, pulse width, and pulse height of the pulse signal according to the frequency, phase, and amplitude of the sine waveform after the fitting is completed.

[0074] In yet another aspect, the present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it realizes the three-dimensional pulse signal quantization analysis method provided by the above-mentioned various methods. The method includes: collecting continuous contact film images at the radial artery of a traditional Chinese medicine pulse training instrument through a binocular vision pulse image acquisition system, and reconstructing a three-dimensional pulse signal; performing ROI region selection on the reconstructed three-dimensional pulse signal to obtain a central region containing the effective components of the three-dimensional pulse signal; using dynamic waveform slicing to divide the three-dimensional pulse signal in the central region into multiple two-dimensional slices to obtain a dynamic slice sequence; using a composite sine cascade modeling model to fit the two-dimensional signals with sine similarity characteristics in the dynamic slice sequence; determining the pulse length, pulse width, and pulse height of the pulse signal according to the frequency, phase, and amplitude of the sine waveform after the fitting is completed.

[0075] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. A person of ordinary skill in the art can understand and implement it without creative labor.

[0076] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on such an understanding, the essence of the above technical solution, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0077] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A three-dimensional pulse signal quantitative analysis method, characterized in that: include: The continuous contact membrane image of the Chinese medicine pulse training instrument at the radial artery is collected through the binocular vision pulse image acquisition system to reconstruct the three-dimensional pulse signal; Performing ROI region selection on the reconstructed three-dimensional pulse signal to obtain a central region containing effective components of the three-dimensional pulse signal; Using dynamic waveform slicing, the three-dimensional pulse signal in the central area is divided into a plurality of two-dimensional slices to obtain a dynamic slice sequence; Using a composite sinusoidal cascade modeling model to fit the two-dimensional signal with sinusoidal similarity characteristics in the dynamic slice sequence; The pulse length, pulse width and pulse height of the pulse signal are determined according to the frequency, phase and amplitude of the sinusoidal waveform after the fitting.

2. The three-dimensional pulse signal quantitative analysis method according to claim 1, characterized in that: The step of performing ROI region selection on the reconstructed three-dimensional pulse signal to obtain a central region containing effective components of the three-dimensional pulse signal includes: The three-dimensional pulse signal is represented by using the pulse length grid length, the pulse width grid length and the pulse height amplitude to obtain a three-dimensional pulse signal image; Based on the indicator function, the ROI region in the three-dimensional pulse signal image that contains the effective component of the three-dimensional pulse signal is screened as the central region.

3. The three-dimensional pulse signal quantitative analysis method according to claim 1, characterized in that: The method of using dynamic waveform slicing to divide the three-dimensional pulse signal in the central area into a plurality of two-dimensional slices to obtain a dynamic slice sequence includes: Selecting the dimension of pulse width as the slicing direction according to the characteristics of the three-dimensional pulse signal in the central area; According to the dynamic change of the pulse width of the three-dimensional pulse signal in the central area, combined with the grid distribution when reconstructing the three-dimensional pulse signal, the position and range of the slice are adjusted; Based on the position and range of the slice, at each slice point, extracting a two-dimensional slice from the three-dimensional pulse signal; All the extracted two-dimensional slices are arranged in order according to the slice direction to form a dynamic slice sequence.

4. The three-dimensional pulse signal quantitative analysis method according to claim 1, characterized in that: The method of fitting the two-dimensional signal having a sinusoidal similarity feature in the dynamic slice sequence by using a composite sinusoidal cascade modeling model comprises: constructing a composite sinusoidal cascade modeling model based on the slice characteristics in the dynamic slice sequence; In combination with parameter boundary constraints, solving parameters of the composite sinusoidal cascade modeling model; An optimization algorithm is used to determine the optimal characteristic parameters of each sinusoidal function in the solved parameters, minimize the fitting error, and obtain the optimal model parameters.

5. The three-dimensional pulse signal quantitative analysis method according to claim 4, characterized in that: The step of constructing a composite sinusoidal cascade model based on the slice characteristics in the dynamic slice sequence includes: Find all the troughs of the dynamic slicing sequence, perform threshold processing on the dynamic slicing sequence, and extract rising edges and falling edges; Based on the rising edge and the falling edge, extracting a two-dimensional slice array from any current trough position to the next trough position from the dynamic slice sequence; Based on the curve characteristics of all the two-dimensional slice arrays, sinusoidal function models are respectively used for fitting to construct a composite sinusoidal cascade modeling model.

6. The three-dimensional pulse signal quantitative analysis method according to claim 4, characterized in that: Before solving the parameters of the composite sinusoidal cascade modeling model in combination with the parameter boundary constraints, the method further includes: Initialize each parameter to obtain the corresponding initial value; Based on the initial value, a boundary constraint condition is set for each of the parameters.

7. The three-dimensional pulse signal quantitative analysis method according to claim 4, characterized in that: The optimization algorithm is used to determine the optimal characteristic parameters of each sine function in the solved parameters, minimize the fitting error, and obtain the optimal model parameters, including: Determine the objective function to be minimized in the optimization solution process; The YS-PSO optimization algorithm is selected to optimize the solved parameters, minimize the fitting error, obtain the optimal model parameters, and organize all the optimal model parameters into a feature vector set.

8. The three-dimensional pulse signal quantitative analysis method according to any one of claims 1 to 7, characterized in that: The composite sinusoidal cascade modeling model is: ； in, is the number of sinusoidal function models; It is a composite sinusoidal cascade modeling model; It is the total length of each 2D slice vein divided by the number of grids; is the current sine function model, Indicates A sine function models the amplitude; Indicates Sine function model frequency; Indicates A sine function models phase shift; Indicates A sinusoidal function models vertical offset.

9. The three-dimensional pulse signal quantitative analysis method according to any one of claims 1 to 7, characterized in that: The reconstructing of the three-dimensional pulse signal comprises: The collected image pulse signal is filtered, binarized and feature points are extracted, and combined with the binocular vision imaging principle, the three-dimensional pulse signal is reconstructed.

10. A three-dimensional pulse signal quantitative analysis system, characterized in that: include: A reconstruction module is used to collect the continuous contact membrane image of the traditional Chinese medicine pulse training instrument at the radial artery through a binocular vision pulse image acquisition system to reconstruct a three-dimensional pulse signal; A selection module, used for performing ROI area selection on the reconstructed three-dimensional pulse signal to obtain a central area containing effective components of the three-dimensional pulse signal; A slicing module, used to divide the three-dimensional pulse signal in the central area into a plurality of two-dimensional slices by using dynamic waveform slicing to obtain a dynamic slice sequence; A fitting module, used for fitting the two-dimensional signal with sinusoidal similarity characteristics in the dynamic slice sequence by using a composite sinusoidal cascade modeling model; The determination module is used to determine the pulse length, pulse width and pulse height of the pulse signal according to the frequency, phase and amplitude of the sinusoidal waveform after the fitting.

Images