Method and system for synchronous determination of cf-pwv based on multi-modal data fusion

Abstract

The application relates to the technical field of cf-PWV measurement, in particular to a cf-PWV synchronous measurement method and system based on multi-modal data fusion, which comprises a multi-channel signal acquisition unit, a signal synchronization unit, a signal fusion unit and a result generation unit, signal synchronization is realized through multi-modal signal acquisition and timestamp calibration, signal fusion is realized through wavelet transformation and principal component analysis, and a cf-PWV measurement result is generated by combining Kalman filtering optimization, the application can improve the cf-PWV measurement accuracy and real-time performance, enhance the portability and adaptability of the system, meet the clinical and family health management requirements, and provide technical support for early diagnosis of cardiovascular diseases.

Description

Technical Field

[0001] The present invention belongs to the field of biomedical engineering and health monitoring technology, and specifically relates to a cf-PWV synchronous measurement method and system based on multimodal data fusion. Background Art

[0002] Multimodal data fusion technology, by integrating signals and information from multiple sources, provides a powerful analytical tool for the biomedical engineering field. In cardiovascular health monitoring, central to peripheral pulse wave velocity (cf-PWV), a key indicator for assessing arteriosclerosis, relies on the simultaneous acquisition and efficient fusion analysis of multiple physiological signals. However, existing cf-PWV measurement technologies have limitations in data synchronization, multi-source signal fusion capabilities, and system integration, hindering their widespread application in clinical and home health management.

[0003] The existing technology has the following deficiencies:

[0004] Currently, existing synchronous measurement technologies have deficiencies in feature extraction and time synchronization precision control of multimodal signals, resulting in increased signal processing delays and reduced accuracy and stability of cf-PWV measurement results. Therefore, the present invention aims to provide a cf-PWV synchronous measurement method and system based on multimodal data fusion.

[0005] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not form the prior art that is already known to a person of ordinary skill in the art. Summary of the Invention

[0006] To overcome the aforementioned shortcomings of the prior art, embodiments of the present invention provide a method and system for synchronous cf-PWV measurement based on multimodal data fusion. This method utilizes a multi-channel signal acquisition unit to synchronously capture multimodal physiological signals and incorporates a dynamic signal processing algorithm to efficiently fuse and analyze multi-source signals, thereby improving the accuracy and real-time performance of cf-PWV measurement. Furthermore, through the design of a modular integrated architecture, the system's portability and adaptability are enhanced, addressing the shortcomings of current technologies in data synchronization, multi-source signal fusion capabilities, and adaptability to application scenarios.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] The method for synchronously measuring cf-PWV based on multimodal data fusion includes the following steps:

[0009] Step S1: Acquire multiple modal physiological signals through a multi-channel signal acquisition unit, extract characteristic parameters of each modal signal, and perform preliminary classification and labeling of the multimodal signals according to the characteristic parameters;

[0010] Step S2: Time stamp calibration is performed on the collected multi-modal signals, a synchronization deviation value between the signals is calculated, and a synchronization compensation strategy is generated according to the synchronization deviation value;

[0011] Step S3: A dynamic signal processing algorithm is used to perform fusion analysis on the multi-modal signals, a fusion signal feature matrix is generated, and an initial estimation value of the cf-PWV is calculated based on the fusion signal feature matrix;

[0012] Step S4: The fusion signal feature matrix is optimized and adjusted according to the initial estimation value, and a final cf-PWV measurement result is generated in combination with a feedback mechanism.

[0013] In a preferred embodiment, in step S1, the multi-modal physiological signals include an optical pulse wave signal, an electrical electrocardio signal, and a mechanical vibration signal;

[0014] The feature parameters of the respective modal signals are an optical signal intensity variation amplitude, an R-wave interval duration of an electrocardio signal, and a main frequency value of a mechanical vibration signal;

[0015] The feature parameters are used to preliminarily classify and label the multi-modal signals through fuzzy logic rules.

[0016] In a preferred embodiment, in step S1, the specific steps of the classification and labeling are as follows:

[0017] Input definition: The optical signal intensity variation amplitude, the R-wave interval duration of the electrocardio signal, and the frequency distribution range of the mechanical vibration signal are defined as input variables, and are divided into different quantization intervals;

[0018] Output definition: The classification and labeling of the multi-modal signals are defined as output variables;

[0019] Rule formulation: A set of mapping rules are formulated to describe the corresponding relationship between different input variables and output variables;

[0020] Label generation: The input variables are classified and labeled according to the mapping rules;

[0021] According to the classification and labeling results, modal signal samples that meet the preset quality indicators are selected.

[0022] In a preferred embodiment, in step S2, a unified time reference is generated by receiving an external GPS time signal;

[0023] The time stamps of the modal signal samples are obtained, and the time stamps of the respective modal signal samples are calibrated based on the unified time reference;

[0024] The timestamp of any modal signal is selected as the time reference modal signal, and the remaining modal signals are selected as the target modal signals;

[0025] Calculate the time difference between the time reference modal signal and the target modal signal timestamp, and take the average as the synchronization time deviation value.

[0026] In a preferred embodiment, in step S2, a synchronization compensation strategy is generated based on the synchronization time deviation value, and the synchronization compensation strategy eliminates the synchronization time deviation value by introducing a time offset in the modal signal sampling process;

[0027] If the synchronization time deviation value is greater than the sampling period, the time axis translation method is used to perform overall offset compensation on the signal data;

[0028] If the synchronization time deviation value is less than the sampling period, the interpolation reconstruction method is used to adjust the time axis of the target modal signal.

[0029] In a preferred embodiment, in step S3, the dynamic signal processing algorithm is a wavelet transform algorithm;

[0030] The wavelet transform algorithm is used to perform time-frequency decomposition on modal signal samples and extract local features of the signal. The specific steps are as follows:

[0031] Signal decomposition: The modal signal samples are decomposed into multiple scales using wavelet basis functions to obtain wavelet coefficients at different scales;

[0032] Feature extraction: Statistical analysis is performed on the wavelet coefficients at each scale, and the mean, variance, and energy distribution are extracted as features;

[0033] Feature integration: Integrate the features of all modal signals into a multidimensional feature vector;

[0034] The fusion signal feature matrix is ​​constructed based on the multi-dimensional feature vectors at each decomposition scale.

[0035] In a preferred embodiment, in step S3, the Kalman filter algorithm performs state prediction and observation update on the fusion signal feature matrix, gradually approaching the optimal estimate. The specific operation process is as follows:

[0036] Generate the predicted value of the state cf-PWV at the next moment according to the current state of the fusion signal feature matrix;

[0037] The Kalman gain is calculated by the deviation between the actual observation value and the state prediction value, and the state estimation value is updated;

[0038] The iterative optimization is repeated until the state estimate converges, thereby obtaining the initial estimate of cf-PWV.

[0039] In a preferred embodiment, in step S4, if the error between the characteristic parameters of each modal signal and the initial estimated value exceeds the preset error threshold, an adjusted weight vector is calculated comprehensively;

[0040] An optimized characteristic vector is calculated by linear weighting based on the adjusted weight vector and the multi-dimensional characteristic vector;

[0041] The Kalman filtering algorithm is repeated based on the optimized characteristic vector to generate a final cf-PWV measurement result.

[0042] The cf-PWV synchronous measurement system based on multi-modal data fusion is used to implement the cf-PWV synchronous measurement method based on multi-modal data fusion:

[0043] It comprises a multi-channel signal acquisition unit, a signal synchronization unit, a signal fusion unit, and a result generation unit.

[0044] The multi-channel signal acquisition unit is used to acquire physiological signals of multiple modalities and transmit them to the signal synchronization unit.

[0045] The signal synchronization unit generates a synchronization compensation strategy according to the timestamp calibration and transmits the synchronized signals to the signal fusion unit.

[0046] The signal fusion unit generates a fusion signal feature matrix through a dynamic signal processing algorithm and transmits the matrix to the result generation unit.

[0047] The result generation unit is used to receive the fusion signal feature matrix and generate a final cf-PWV measurement result.

[0048] The technical effects and advantages of the present application are:

[0049] The cf-PWV synchronous measurement system and method based on multi-modal data fusion have the following technical effects and advantages: The multi-channel signal acquisition unit is used to synchronously capture multi-modal physiological signals, solving the shortcomings of the prior art in data synchronization. The dynamic signal processing algorithm is used to realize efficient fusion analysis of multi-source signals, improving the accuracy and real-time performance of cf-PWV measurement. The modular integrated architecture design enhances the portability and application scenario adaptability of the system. The combination of the above technical means enables the present application to meet the efficient, accurate, and convenient requirements of cf-PWV measurement in clinical and family health management, providing technical support for early diagnosis and health management of cardiovascular diseases. BRIEF DESCRIPTION OF DRAWINGS

[0050] Figure 1 The dynamic signal processing algorithm running flowchart of the signal fusion unit in the embodiments of the present application.

[0051] Figure 2Schematic diagram of the process of the cf-PWV synchronous measurement method based on multimodal data fusion in an embodiment of the present invention. DETAILED DESCRIPTION

[0052] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. 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.

[0053] The present invention provides a cf-PWV synchronous measurement method and system based on multimodal data fusion, Figures 1 to 2 The figure marks and their corresponding functional modules shown in the figure are combined with specific embodiments to explain the technical solution of this solution in detail.

[0054] Example 1, a method for synchronously measuring cf-PWV based on multimodal data fusion, such as Figure 1 As shown, the following steps are included:

[0055] Step S1: Acquire multiple modal physiological signals through a multi-channel signal acquisition unit, extract characteristic parameters of each modal signal, and perform preliminary classification and labeling of the multimodal signals according to the characteristic parameters;

[0056] Step S2: performing time stamp calibration on the collected multimodal signals, calculating the synchronization deviation between the signals, and generating a synchronization compensation strategy based on the synchronization deviation;

[0057] Step S3: using a dynamic signal processing algorithm to perform fusion analysis on the multimodal signals, generating a fusion signal feature matrix, and calculating an initial estimate of cf-PWV based on the fusion signal feature matrix;

[0058] Step S4: Optimize and adjust the fusion signal feature matrix based on the initial estimate, and generate the final cf-PWV measurement result in combination with the feedback mechanism.

[0059] The specific implementation is as follows:

[0060] In step S1, a collection cycle is preset, and multimodal signals are sampled multiple times at equal intervals, and a multi-channel signal collection unit is used to obtain physiological signals of multiple modalities, including optical pulse wave signals, electrical electrocardiogram signals, and mechanical vibration signals;

[0061] The characteristic parameters of each modal signal are the amplitude of light intensity variation, the duration of R-wave interval and the main frequency value of the mechanical vibration signal;

[0062] The amplitude of light intensity variation is the difference between the highest frequency and the lowest frequency of the optical pulse wave signal within the acquisition period;

[0063] The time of the R-wave peak in the electrical electrocardio signal is recorded, and the mean value of the time difference between adjacent R-wave peaks in the collection period is taken as the R-wave interval length;

[0064] The main frequency value of the mechanical vibration signal is the frequency component with the maximum power in the frequency spectrum of the mechanical vibration signal;

[0065] According to the characteristic parameters, the multi-modal signal is preliminarily classified and labeled by fuzzy logic rules, and the specific steps are as follows:

[0066] Input definition: the amplitude of the light intensity change of the optical signal, the R-wave interval length of the electrocardio signal, and the main frequency value of the mechanical vibration signal are defined as input variables, and are divided into different quantization intervals;

[0067] Output definition: the classification label of the multi-modal signal is defined as an output variable;

[0068] Rule making: a set of mapping rules are made to describe the corresponding relationship between different input variables and output variables;

[0069] Label generation: the input variables are classified and labeled according to the mapping rules.

[0070] According to the classification label result, the modal signal samples meeting the preset quality indicators are screened, and the signal data with large noise, abnormality or invalidity are removed, so as to ensure the quality and stability of the fusion signal.

[0071] It should be noted that the multi-channel signal acquisition unit includes a plurality of sensor interfaces for connecting optical pulse wave sensors, electrocardio sensors and mechanical vibration sensor devices. The optical pulse wave sensor detects the absorption change of light by blood flow by emitting light of a specific wavelength to the skin surface layer. The electrocardio sensor is a sensing device for detecting and recording the electrical activity of the human heart surface, and can collect the voltage signal generated with the change of the heart cycle. The mechanical vibration sensor is a sensor for sensing and measuring the small mechanical vibration or acceleration change of an object, and detecting the mechanical micro-motion signal generated by heartbeat, pulse or movement.

[0072] In step S2, a unified time reference is obtained through the GPS time signal, and the time consistency of the multi-modal signal is ensured;

[0073] Based on the screened modal signal samples, the time stamps of the modal signal samples are obtained through high-precision time equipment, the time stamps of the modal signal samples are calibrated based on the unified time reference, and the modal signal samples have consistent time sequence alignment relationship under the same time axis;

[0074] Record the timestamps of the R-wave peak value in the electrical ECG signal, the corresponding maximum frequency value in the optical pulse wave signal, and the main frequency value of the mechanical vibration signal in the modal signal sample;

[0075] The timestamp of any one modal signal is selected as the time reference modal signal, and the other modal signals are selected as the target modal signals. The time difference between the time reference modal signal and the target modal signal is calculated, and the average is taken as the synchronization time deviation value.

[0076] A synchronization compensation strategy is generated based on the synchronization time deviation value. The synchronization compensation strategy eliminates the synchronization time deviation value by introducing a time offset in the modal signal sampling process.

[0077] If the synchronization time deviation value is greater than the sampling period, the time axis translation method is used to perform overall offset compensation on the signal data;

[0078] If the synchronization time deviation value is less than the sampling period, the interpolation reconstruction method is used to adjust the time axis of the target modal signal;

[0079] The time axis translation method converts the synchronization time deviation value into the number of sampling intervals, and shifts the target modal signal forward or backward in the time dimension by the corresponding number of sampling intervals, thereby achieving timing alignment with the reference modal signal;

[0080] The interpolation reconstruction method is to insert a virtual point between two sampling intervals to reconstruct the value of the target modal signal at the offset time point;

[0081] Interpolation calculation is performed between the current sampling point of the target modal signal and the adjacent sampling points based on the synchronization time deviation to construct the virtual value of the target modal signal in the time reference modal signal, ensuring that the characteristic peaks of the multimodal signals coincide on the same reference time axis;

[0082] The interpolation calculation formula is: ,in, for The characteristic parameters of the moment, for The characteristic parameters at the moment, t is the interpolation time point, Is a virtual value.

[0083] Timestamp calibration and synchronization compensation processing are performed under a unified time reference based on modal signal samples to ensure the consistency of each modal signal on the time axis, establishing a stable and reliable timing foundation for subsequent fusion analysis and cf-PWV estimation.

[0084] It should be noted that high-precision time equipment refers to a type of time synchronization hardware used to provide high-resolution time reference and precise timestamp recording capabilities, ensuring time consistency and signal alignment accuracy between multiple data acquisition channels.

[0085] In step S3, the dynamic signal processing algorithm is a wavelet transform algorithm;

[0086] The wavelet transform algorithm is used to perform time-frequency decomposition on modal signal samples and extract local features of the signal. The specific steps are as follows:

[0087] Signal decomposition: The modal signal samples are processed by discrete wavelet transform to obtain approximation coefficients and detail coefficients, which are denoted as and , where j is the preset decomposition scale layer; the modal signal samples are decomposed into multiple scales through wavelet basis functions to obtain wavelet coefficients at different decomposition scales: ,in, is the preset scale parameter, b is the preset translation parameter, is the conjugate complex function of the mother wavelet, is the wavelet coefficient;

[0088] Feature extraction: Statistical analysis is performed on the wavelet coefficients of each decomposition scale layer to obtain the mean, variance, and energy as features; the energy calculation formula is: ,in, is the i-th wavelet coefficient of each decomposition scale layer, N is the total number of wavelet coefficients, is the energy of the j-th decomposition scale.

[0089] Feature integration: Merge the features into a multidimensional feature vector, integrate the feature vectors extracted from the modal signal samples at each decomposition scale, and construct a fusion signal feature matrix.

[0090] The Kalman filter algorithm performs state prediction and observation updates on the fusion signal feature matrix, gradually approaching the optimal estimate. The specific operation process is as follows:

[0091] Generate the predicted value of the state cf-PWV at the next moment based on the current state of the fusion signal feature matrix, calculate the Kalman gain by the deviation between the actual observation value and the state prediction value, and update the state estimate;

[0092] The iterative optimization is repeated until the state estimate converges, thereby obtaining the initial estimate of cf-PWV.

[0093] It should be noted that the wavelet transform algorithm is a time-frequency analysis method that can decompose and represent signals at multiple scales and resolutions. It is suitable for feature extraction and analysis of non-stationary signals. It decomposes signals into components with different frequencies and local time, thereby revealing detailed features and trends at different scales. Wavelet transform algorithms include continuous wavelet transform and discrete wavelet transform. The discrete wavelet transform is computationally efficient and suitable for digital signal processing. Through layer-by-layer filtering and downsampling, the discrete wavelet transform decomposes the signal into approximation coefficients and detail coefficients, completing multi-scale decomposition. The number of decomposition scales is usually a preset parameter determined by professionals based on specific application requirements and signal characteristics. For example, a greater number of decomposition levels results in finer signal decomposition, enabling analysis of lower-frequency components, but increases computational complexity, and some high-level decompositions may be sensitive to noise. Too few decomposition levels may not effectively reveal the multi-scale structure of the signal, resulting in insufficient feature extraction. The preset scale parameter adjusts the wavelet's frequency range to enable signal analysis at different decomposition scales. The preset shift parameter determines the wavelet's position on the time axis to enable local temporal analysis of the signal. The specific settings are determined by professionals.

[0094] In step S4, the fusion signal feature matrix is ​​optimized and adjusted according to the initial estimated value. The specific steps are as follows:

[0095] The initial estimated value of cf-PWV is used as feedback to evaluate the degree of fit between the characteristic parameters of each modal signal and to dynamically adjust the weight distribution of each characteristic parameter in the fusion feature vector accordingly.

[0096] Specifically, if the error between the characteristic parameter of each modal signal and the initial estimated value exceeds a preset error threshold, the weight of the modal feature is reduced: ,in, is the preset maximum error value, is the error between the characteristic parameter of the i-th modal signal and the initial estimated value, is the preset error threshold, is the initial weight value of the characteristic parameter of the i-th modal signal in the fusion feature vector, is the adjusted weight vector to weaken its influence on the final measurement result;

[0097] Calculate the optimized eigenvector using the linear weighted expression: ,in, is a multidimensional feature vector, is the optimized feature vector;

[0098] The final cf-PWV measurement results were generated based on the optimized eigenvector repeated Kalman filtering algorithm.

[0099] It should be noted that the preset error threshold is used to limit the acceptable error range. Modal signal features that exceed the error range are regarded as abnormal or noise signals, and their weights should be reduced accordingly. The specific setting is performed by professionals and will not be elaborated here.

[0100] Example 2, cf-PWV synchronous measurement system based on multimodal data fusion, such as Figure 2 As shown, a method for realizing synchronous cf-PWV measurement based on multimodal data fusion includes a multi-channel signal acquisition unit, a signal synchronization unit, a signal fusion unit, and a result generation unit. The signal connections between the modules have the following functions:

[0101] The multi-channel signal acquisition unit is used to collect physiological signals of multiple modalities and transmit them to the signal synchronization unit;

[0102] The signal synchronization unit generates a synchronization compensation strategy based on the timestamp calibration and transmits the synchronized signal to the signal fusion unit;

[0103] The signal fusion unit generates a fusion signal feature matrix through a dynamic signal processing algorithm and transmits the matrix to the result generation unit;

[0104] The result generating unit is used to receive the fusion signal feature matrix and generate the final cf-PWV measurement result.

[0105] In actual application scenarios, the various modules of this system are closely connected through hardware circuits and communication protocols.

[0106] For example, the multi-channel signal acquisition unit and the signal synchronization unit are connected via a high-speed serial communication interface to ensure the real-time and reliability of signal transmission.

[0107] A double buffer design is used between the signal synchronization unit and the signal fusion unit to avoid data loss or delay.

[0108] Data exchange between the signal fusion unit and the result generation unit is achieved through shared memory, further improving data processing efficiency.

[0109] In addition, the entire system adopts a modular integrated architecture design, which facilitates flexible configuration according to the needs of different application scenarios.

[0110] For example, in a clinical environment, a multi-channel signal acquisition unit can be combined with a portable host to form a set of lightweight mobile measurement equipment; in a home health management scenario, the acquisition unit can be connected to a smart terminal through wireless communication technology to achieve remote monitoring and data analysis.

[0111] To verify the performance of this system, a number of experimental tests were conducted. The experimental results show that this system has significant advantages in data synchronization, multi-source signal fusion capabilities, and application scenario adaptability.

[0112] For example, in an experiment on healthy volunteers, this system successfully achieved the synchronous acquisition and fusion analysis of optical pulse wave signals, electrocardiogram signals, and mechanical vibration signals. The measurement results were more accurate and real-time than traditional methods.

[0113] In addition, through modular design, this system can adapt to different usage environments and meet the diverse needs of clinical diagnosis and family health management.

[0114] The above embodiments describe in detail the specific implementation process of the present solution, covering every step from signal acquisition to result generation, to ensure that those skilled in the art can fully reproduce the technical solution according to the contents of the specification.

[0115] In order to better enable relevant personnel in this technical field to fully understand and implement this solution, the specific implementation principle of this solution is supplemented below with reference to a specific application scenario.

[0116] In a clinical setting, the cf-PWV simultaneous measurement system based on multimodal data fusion is configured as a lightweight mobile measurement device for cardiovascular health monitoring.

[0117] In this scenario, patients need to complete the measurement of central artery to peripheral artery pulse wave velocity (cf-PWV) under the guidance of a doctor to assess the degree of arteriosclerosis.

[0118] The whole process unfolds as follows:

[0119] First, in the signal acquisition stage, the multi-channel signal acquisition unit is connected to the optical pulse wave sensor, electrocardiogram sensor and mechanical vibration sensor through multiple sensor interfaces.

[0120] These sensors are placed at specific locations on the patient's body: the optical pulse wave sensor is attached to the radial artery at the wrist to collect optical pulse wave signals;

[0121] The ECG sensor is connected to the chest via a lead wire to collect ECG signals; the mechanical vibration sensor is fixed in the carotid artery area to capture mechanical vibration signals.

[0122] The high-resolution analog-to-digital converter integrated inside the multi-channel signal acquisition unit digitizes the original analog signal, while the anti-aliasing filter circuit effectively suppresses high-frequency noise interference to ensure signal quality.

[0123] By receiving external GPS time signals or internal clock signals, a unified time reference is provided for all acquisition channels, thus ensuring the time consistency of multimodal signals.

[0124] This design solves the data asynchrony problem caused by time deviation in traditional methods and significantly improves the reliability of signal acquisition.

[0125] Then, the signal synchronization phase begins. The signal synchronization unit reads the timestamp records t1 and t2 of each modal signal and calculates the synchronization deviation between the signals according to the formula Δt = |t1-t2|.

[0126] The time synchronization algorithm built into the embedded processor adjusts the sampling order in real time and eliminates synchronization deviation by introducing a time offset Δt, so that the time axes of different modal signals are accurately aligned.

[0127] The signal synchronization unit and the time synchronization module are closely connected through the data bus to ensure the accurate transmission of the time reference signal.

[0128] The time-stamped signals are then passed to a signal fusion unit, which prepares them for subsequent processing. This step achieves millisecond-level synchronization of multimodal signals, meeting the high temporal precision required for cf-PWV measurement.

[0129] Next, the signal fusion unit dynamically processes the synchronized multimodal signals.

[0130] Each modal signal is decomposed into multiple scales to extract wavelet coefficients at different scales. The characteristic parameters such as mean, variance and energy distribution are generated through statistical analysis of the wavelet coefficients.

[0131] These feature parameters are integrated into a multi-dimensional feature vector and fused to generate a signal feature matrix.

[0132] Finally, the Kalman filter module performs state prediction and observation update on the fusion signal feature matrix, gradually approaching the optimal estimate.

[0133] The specific operation process of Kalman filtering includes three steps: state prediction, observation update and iterative optimization. The final generated fusion signal feature matrix contains highly concentrated multimodal signal information.

[0134] The technical principle of this process is to maximize the extraction of useful information from multi-source signals through multi-level signal processing algorithms, thereby improving the accuracy of cf-PWV measurement.

[0135] Finally, the result generation unit receives the fusion signal feature matrix generated by the signal fusion unit and optimizes and adjusts it through a built-in feedback mechanism.

[0136] The above embodiments may be implemented in whole or in part through software, hardware, firmware or any other combination. When implemented using software, the above embodiments may be implemented in whole or in part in the form of a computer program product.

[0137] Those skilled in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application of the technical solution and the invention constraints. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0138] In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0139] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0140] Finally: 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 method for synchronously measuring cf-PWV based on multimodal data fusion, characterized in that: The following steps are involved: Step S1: Acquire multiple modal physiological signals through a multi-channel signal acquisition unit, extract characteristic parameters of each modal signal, and perform preliminary classification and labeling of the multimodal signals according to the characteristic parameters; Step S2: performing time stamp calibration on the collected multimodal signals, calculating the synchronization deviation between the signals, and generating a synchronization compensation strategy based on the synchronization deviation; Step S3: using a dynamic signal processing algorithm to perform fusion analysis on the multimodal signals, generating a fusion signal feature matrix, and calculating an initial estimate of cf-PWV based on the fusion signal feature matrix; In step S3, the Kalman filter algorithm performs state prediction and observation update on the fusion signal feature matrix, gradually approaching the optimal estimate. The specific operation process is as follows: Generate the predicted value of the state cf-PWV at the next moment according to the current state of the fusion signal feature matrix; The Kalman gain is calculated by the deviation between the actual observation value and the state prediction value, and the state estimation value is updated; The iterative optimization is repeated until the state estimate converges, thereby obtaining the initial estimate of cf-PWV; Step S4: Optimize and adjust the fusion signal feature matrix based on the initial estimate, and generate the final cf-PWV measurement result in combination with the feedback mechanism.

2. The method for synchronously measuring cf-PWV based on multimodal data fusion according to claim 1, characterized in that: In step S1, the multimodal physiological signal includes an optical pulse wave signal, an electrical electrocardiogram signal, and a mechanical vibration signal; The characteristic parameters of each modal signal are the light intensity variation amplitude of the optical signal, the R wave interval duration of the electrocardiogram signal, and the main frequency value of the mechanical vibration signal. The characteristic parameters are used to preliminarily classify and label the multimodal signals through fuzzy logic rules.

3. The method for synchronously measuring cf-PWV based on multimodal data fusion according to claim 2, characterized in that: In step S1, the specific steps of classification marking are as follows: Input definition: The light intensity variation amplitude of the optical signal, the R-wave interval duration of the ECG signal, and the frequency distribution range of the mechanical vibration signal are defined as input variables and divided into different quantization intervals; Output definition: Define the classification label of the multimodal signal as the output variable; Rule formulation: formulate a set of mapping rules to describe the correspondence between different input variables and output variables; Label generation: classify and label the input variables according to the mapping rules; According to the classification labeling results, modal signal samples that meet the preset quality indicators are screened.

4. The method for synchronously measuring cf-PWV based on multimodal data fusion according to claim 3, characterized in that: In step S2, a unified time reference is generated by receiving an external GPS time signal; Obtain the timestamps of the modal signal samples and calibrate the timestamps of the modal signal samples based on a unified time reference; The timestamp of any modal signal is selected as the time reference modal signal, and the remaining modal signals are selected as the target modal signals; Calculate the time difference between the time reference modal signal and the target modal signal timestamp, and take the average as the synchronization time deviation value.

5. The method for synchronously measuring cf-PWV based on multimodal data fusion according to claim 4, characterized in that: In step S2, a synchronization compensation strategy is generated based on the synchronization time deviation value. The synchronization compensation strategy eliminates the synchronization time deviation value by introducing a time offset in the modal signal sampling process. If the synchronization time deviation value is greater than the sampling period, the time axis translation method is used to perform overall offset compensation on the signal data; If the synchronization time deviation value is less than the sampling period, the interpolation reconstruction method is used to adjust the time axis of the target modal signal.

6. The method for synchronously measuring cf-PWV based on multimodal data fusion according to claim 2, characterized in that: In step S3, the dynamic signal processing algorithm is a wavelet transform algorithm; The wavelet transform algorithm is used to perform time-frequency decomposition on modal signal samples and extract local features of the signal. The specific steps are as follows: Signal decomposition: The modal signal samples are decomposed into multiple scales using wavelet basis functions to obtain wavelet coefficients at different scales; Feature extraction: Statistical analysis is performed on the wavelet coefficients at each scale, and the mean, variance, and energy distribution are extracted as features; feature Integration: Integrate the features of all modal signals into a multidimensional feature vector; The fusion signal feature matrix is ​​constructed based on the multi-dimensional feature vectors at each decomposition scale.

7. The method for synchronously measuring cf-PWV based on multimodal data fusion according to claim 1, characterized in that: In step S4, if the error between the characteristic parameter of each modal signal and the initial estimated value exceeds a preset error threshold, an adjusted weight vector is obtained by comprehensive calculation; The optimized feature vector is calculated based on the adjusted weight vector and the multi-dimensional feature vector through linear weighting; The final cf-PWV measurement results were generated based on the optimized eigenvector repeated Kalman filtering algorithm.

8. A cf-PWV synchronous measurement system based on multimodal data fusion, and a cf-PWV synchronous measurement method based on multimodal data fusion according to any one of claims 1 to 7, characterized in that: It includes a multi-channel signal acquisition unit, a signal synchronization unit, a signal fusion unit and a result generation unit; The multi-channel signal acquisition unit is used to collect physiological signals of multiple modalities and transmit them to the signal synchronization unit; The signal synchronization unit generates a synchronization compensation strategy based on the timestamp calibration and transmits the synchronized signal to the signal fusion unit; The signal fusion unit generates a fusion signal feature matrix through a dynamic signal processing algorithm and transmits the matrix to the result generation unit; The result generating unit is used to receive the fusion signal feature matrix and generate the final cf-PWV measurement result.

Images