A traditional chinese medicine health assessment method and system based on physiological signal enhancement

Abstract

The application relates to the field of physiological signal processing, in particular to a traditional Chinese medicine health assessment method and system based on physiological signal enhancement, which comprises the following steps: collecting pulse physiological signals through a finger collection instrument to generate an ordered original signal data set; decomposing each original signal into three types of independent components; generating a pretreated effective signal set; establishing a signal consistency evaluation model; calling the pre-trained signal consistency evaluation model to calculate the consistency scores of the effective signals, eliminating abnormal segments, and superimposing and reconstructing the effective signal segments according to the scores to generate standard physiological signals; generating a traditional Chinese medicine health assessment result based on the standard physiological signals, comparing the fluctuation values of the output assessment result and a benchmark result; and outputting a traditional Chinese medicine health assessment report meeting the stability requirements. The application improves the repeatability and reliability of the assessment result, adapts to the professional needs of clinical efficacy tracking, and effectively reduces the generation of poor-quality data.

Description

Technical Field

[0001] This application belongs to the field of physiological signal processing, specifically relating to a traditional Chinese medicine health assessment method and system based on physiological signal enhancement. Background Technology

[0002] In the process of large-scale commercialization of TCM digital diagnosis and treatment hardware, although TCM health assessment technology is already quite mature, there are still some problems in the physiological signal acquisition and health assessment technology based on finger-based data acquisition devices, including:

[0003] Traditional single-filter denoising has fundamental flaws. It cannot accurately distinguish between stable physiological components, transient interference components, and low-frequency drift components in the original signal. A uniform filtering strategy may weaken the key effective features of TCM organ and constitution differentiation, or cause signal distortion due to incomplete denoising, which directly affects the basic accuracy of health assessment.

[0004] There are core obstacles to the implementation in non-professional scenarios. In scenarios such as home use and offline stores, non-standard operations such as changes in user contact pressure and instantaneous shaking can easily lead to abnormal segmentation of the collected signal. Traditional solutions either output distorted evaluation results or require users to repeat the collection, which seriously affects user experience and the large-scale promotion of the product.

[0005] The disconnect between signal processing and health assessment means that even if a user's physical condition remains unchanged for a short period, the test results can fluctuate significantly. This results in insufficient repeatability and reliability, failing to meet the needs of clinical efficacy tracking and generating a large amount of low-quality data. Summary of the Invention

[0006] The purpose of this application is to provide a traditional Chinese medicine health assessment method based on physiological signal enhancement, and at the same time, to provide a traditional Chinese medicine health assessment system based on physiological signal enhancement, so as to solve the problems mentioned in the background art.

[0007] To solve the above-mentioned technical problems, this application provides the following technical solution: A traditional Chinese medicine health assessment method based on enhanced physiological signals includes the following steps: The pulse physiological signal is collected in a single detection process using a finger acquisition device, generating an ordered raw signal dataset. Each original signal segment is decomposed into three independent components: stable physiological components, transient interference components, and low-frequency drift components. Transient interference components are removed, low-frequency drift components are corrected, and feature-protected smoothing is performed on the stable physiological components to generate a preprocessed effective signal set. Establish a signal consistency evaluation model; The pre-trained signal consistency evaluation model is invoked to calculate the consistency score of each effective signal segment. Abnormal segments with scores lower than the preset qualified threshold are removed. The remaining effective signal segments are reconstructed by weighted superposition according to the scores to generate standard physiological signals. The system generates TCM health assessment results based on standard physiological signals. It compares the fluctuation values ​​of the output assessment results with the baseline results. If the fluctuations exceed the preset threshold, it automatically adjusts the component processing parameters and repeats the signal decomposition, consistency score calculation, and generation of standard physiological signals and TCM health assessment results until the results meet the stability requirements. Finally, it outputs a TCM health assessment report that meets the stability requirements.

[0008] Furthermore, by using a finger acquisition device, pulse physiological signals are collected within a single detection process to generate an ordered raw signal dataset. This includes: collecting pulse pressure physiological signals from the fingertip of the human body using a finger acquisition device, performing multiple consecutive short-term physiological signal acquisitions, using a fixed sampling frequency during the acquisition process, controlling the interval between adjacent acquisition segments, and generating an ordered raw signal dataset according to the chronological order of the acquisitions after completing the entire acquisition process.

[0009] Furthermore, the process of decomposing each original signal segment into three independent components—stable physiological components, transient interference components, and low-frequency drift components—includes the following steps: For each original single-segment signal within the original signal dataset, time-domain amplitude normalization is first performed. The normalized single-segment original signal is then subjected to layer-by-layer adaptive decomposition to obtain several intrinsic mode function components and a residual component. Based on the inherent frequency band characteristics of the human fingertip pulse pressure physiological signal, frequency domain features are extracted and clustered on the decomposed components. These components are then superimposed and integrated into three independent components: stable physiological components, transient interference components, and low-frequency drift components. After the three-component division of a single-segment signal is completed, component orthogonality verification is performed. Once the verification passes, the component separation of that segment is completed. The same process is then used to complete the component decomposition of all original single-segment signals within the dataset.

[0010] Furthermore, the process of eliminating transient interference components, correcting low-frequency drift components, and performing feature-protective smoothing on stable physiological components to generate a preprocessed effective signal set includes: first, zeroing out transient interference components from each segment of the original signal after orthogonality verification, without touching the effective frequency band data of stable physiological components throughout the process; then, performing baseline correction and trend elimination on low-frequency drift components, fully preserving the extremely low-frequency effective physiological features; performing feature-protective smoothing on stable physiological components, locking the core feature points of the pulse wave to not participate in the smoothing calculation; smoothing the transition intervals between feature points, filtering out residual noise while fully preserving the core morphological features of the pulse wave related to TCM syndrome differentiation; after completing the differentiated preprocessing of the three types of components, superimposing the processed stable physiological components with the effective features preserved by the corrected low-frequency drift components to generate a single segment of preprocessed effective signal, and arranging them according to the original acquisition sequence to generate an ordered preprocessed effective signal set.

[0011] Furthermore, the establishment of the signal consistency evaluation model includes: constructing a hierarchical dataset based on labeled clinical and socially collected pulse physiological signal samples, including populations of different health states across all age groups; constructing a consistency feature system for preprocessed effective signals, including basic signal consistency features and TCM syndrome differentiation correlation features; generating consistency feature vectors as model inputs through standardization; constructing a signal consistency evaluation model; optimizing model parameters through cross-validation; solidifying model parameters and judgment rules after training and clinical validation; embedding the model into the system signal processing link; and independently outputting the consistency score and classification label for the corresponding signal segment.

[0012] Furthermore, the step of calling the pre-trained signal consistency evaluation model to calculate the consistency score of each effective signal segment includes: extracting two types of features, namely basic signal consistency and TCM syndrome differentiation, from each signal segment in the pre-processed effective signal set according to the original acquisition time sequence; generating feature vectors that match the input dimension of the pre-trained signal consistency evaluation model after standardization; outputting the consistency score of the corresponding signal segment from the model; and binding and storing the score with the original acquisition time sequence.

[0013] Furthermore, the process of removing abnormal segments with scores below a preset acceptable threshold and reconstructing the remaining valid signal segments by weighted superposition based on their scores to generate a standard physiological signal includes: comparing the consistency score corresponding to each individual signal segment in the valid signal set with the preset acceptable threshold one by one, marking and permanently removing abnormal segments, retaining valid stable signal segments to form a valid signal subset; if the number of valid stable signal segments is less than the preset minimum number of valid segments, a retry prompt is triggered; if the number requirement is met, weighted superposition reconstruction is performed; the weighted superposition reconstruction uses the consistency score of each valid stable signal segment as the basis for weight allocation, completes the normalized weight calculation and pulse cycle benchmark alignment, and generates an initial reconstructed physiological signal by weighted superposition based on the weights; after passing the integrity and validity verification, a standard physiological signal is generated; if the verification fails, the cycle alignment and weighted superposition calculation are re-executed.

[0014] Furthermore, if the fluctuation values ​​of the comparison output evaluation results and the benchmark results exceed a preset threshold, the component processing parameters are automatically adjusted and the signal decomposition, consistency score calculation, standard physiological signal generation, and TCM health assessment result generation are repeated until the results meet the stability requirements. This includes: setting a comprehensive fluctuation threshold for the TCM health assessment results and a sub-item fluctuation threshold for the core evaluation indicators; reconstructing the benchmark physiological signal by superimposing the effective signal segments in the preprocessed effective signal set with equal weights; generating a dimension-matched benchmark result using a TCM health assessment model with fixed input parameters; calculating the relative deviation between the initial judgment result and the benchmark result; obtaining the comprehensive result fluctuation value according to the preset weight ratio; comparing it with the corresponding threshold; if it meets the constraint, the initial judgment result is marked as the result to be output; if it exceeds the threshold, closed-loop feedback optimization is triggered, iteratively adjusting the component processing parameters and preprocessing process, and re-executing the entire signal processing and evaluation process until the results meet the stability requirements. If the maximum number of iterations is still not met, the iteration is terminated and an abnormal acquisition prompt is triggered.

[0015] Furthermore, the output of the TCM health assessment report that meets stability requirements includes: performing a final review of the full-dimensional compliance and consistency of the output results that have completed stability constraint verification; solidifying the basic information of this test after the review is passed; fully outputting the core TCM health assessment indicators, signal quality and assessment stability quantification data of this test; generating corresponding traceable TCM syndrome interpretation content and matching personalized conditioning direction based on the syndrome differentiation results; simultaneously outputting traceable information of the entire process of this test; marking the confirmed report usage prompts; and generating a standardized TCM health assessment report after completing full content verification.

[0016] This application also discloses an electronic device, including: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform the aforementioned method for TCM health assessment based on enhanced physiological signals according to this application.

[0017] This application also discloses a traditional Chinese medicine health assessment system based on physiological signal enhancement, comprising: The signal acquisition module is used to receive the pulse physiological signal from the finger acquisition device and generate an ordered raw signal dataset; the signal decomposition module is used to decompose each raw signal into three independent components: stable physiological component, transient interference component, and low-frequency drift component. The signal preprocessing module is used to remove transient interference components, correct low-frequency drift components, perform feature-protected smoothing on stable physiological components, and generate a preprocessed effective signal set. The model building module is used to build signal consistency evaluation models; The signal reconstruction module is used to call the pre-trained signal consistency evaluation model, calculate the consistency score of each effective signal segment, remove abnormal segments with scores lower than the preset qualified threshold, and reconstruct the remaining effective signal segments by weighted superposition according to the scores to generate standard physiological signals. The assessment feedback module is used to generate TCM health assessment results based on standard physiological signals, compare the fluctuation values ​​of the output assessment results with the benchmark results, and if the fluctuations of the two exceed the preset threshold, automatically adjust the component processing parameters and repeat the signal decomposition, calculation of consistency score, generation of standard physiological signals and TCM health assessment results until the results meet the stability requirements. The report output module is used to output TCM health assessment reports that meet stability requirements.

[0018] Beneficial Effects: This application overcomes the fundamental shortcomings of traditional single-filtering through precise multi-component separation and differentiated preprocessing. It can accurately distinguish stable physiological components, transient interference components, and low-frequency drift components. While eliminating interference and correcting baseline drift, it fully preserves the core effective features related to TCM organ and constitution differentiation, avoiding feature weakening and signal distortion, and improving the basic accuracy of health assessment from the source. The signal consistency evaluation model enables precise identification and removal of abnormal segments, allowing for weighted superposition and reconstruction of effective signals without repeated user sampling. This solves the problem of signal anomalies caused by non-standard user operations in non-professional scenarios, significantly improving user experience and lowering the threshold for large-scale commercial application. Through a closed-loop feedback optimization mechanism between signal processing and health assessment, it achieves linked calibration of physiological signal enhancement and assessment result stability, solving the pain point of disconnect between signal processing and assessment. This significantly reduces the fluctuation range of test results for the same user within a short period, greatly improving the repeatability and reliability of assessment results. It not only meets the professional needs of clinical efficacy tracking but also effectively reduces the generation of low-quality data, comprehensively covering application needs in multiple scenarios such as clinical settings, offline stores, and home use. Attached Figure Description

[0019] Figure 1 This is a flowchart of a traditional Chinese medicine health assessment method based on physiological signal enhancement, as described in this application. Figure 2 This is a flowchart of the signal differentiation preprocessing step in this application; Figure 3 This is a flowchart illustrating the construction process of the signal consistency evaluation model in this application; Figure 4 This is a flowchart of the signal consistency score calculation in this application; Figure 5 This is a flowchart illustrating the abnormal segmentation removal and standard physiological signal reconstruction process in this application. Figure 6 A flowchart for generating the TCM health assessment results in this application; Figure 7 This is a flowchart illustrating the stability verification and closed-loop feedback optimization of the evaluation results in this application. Detailed Implementation

[0020] The technical solutions in the embodiments of this application will be clearly and completely described below with reference to the embodiments of this application. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0021] This application provides a traditional Chinese medicine health assessment method based on enhanced physiological signals, such as... Figure 1 As shown, the method includes the following steps: The pulse physiological signal is collected in a single detection process using a finger acquisition device, generating an ordered raw signal dataset. Each original signal segment is decomposed into three independent components: stable physiological components, transient interference components, and low-frequency drift components. Transient interference components are removed, low-frequency drift components are corrected, and feature-protected smoothing is performed on the stable physiological components to generate a preprocessed effective signal set. Establish a signal consistency evaluation model; The pre-trained signal consistency evaluation model is invoked to calculate the consistency score of each effective signal segment. Abnormal segments with scores lower than the preset qualified threshold are removed. The remaining effective signal segments are reconstructed by weighted superposition according to the scores to generate standard physiological signals. Based on standard physiological signals, a TCM health assessment result is generated. The fluctuation value of the output assessment result is compared with that of the baseline result. If the fluctuation of the two exceeds the preset threshold, the component processing parameters are automatically adjusted and the signal decomposition, consistency score calculation, standard physiological signal and TCM health assessment result are repeated until the result meets the stability requirements. Output a TCM health assessment report that meets stability requirements.

[0022] Using a finger sampling device, pulse physiological signals are collected within a single testing process, generating an ordered raw signal dataset. Specifically, this involves: using a medical-grade finger sampling device to collect pulse pressure physiological signals from the fingertip; performing a preset fixed positive integer N consecutive short-duration physiological signal acquisitions within a single TCM health testing process, where N ranges from 3 to 10; setting a fixed acquisition duration T, ranging from 2 to 5 seconds; ensuring the time interval between adjacent acquisitions does not exceed 0.2 seconds; using a fixed sampling frequency Fs, with Fs not less than 200Hz; and generating an ordered raw signal dataset S={S1,S2,...,S...} according to the chronological order of acquisition. N}, where S N This is the original single-segment signal obtained from the Nth short-time acquisition.

[0023] The process of decomposing each segment of the original signal into three independent components—stable physiological components, transient interference components, and low-frequency drift components—specifically includes, in practice, decomposing each segment of the original single signal S within the ordered original signal dataset S. N First, time-domain amplitude normalization is performed to uniformly map the signal amplitude to a preset fixed value range of 0 to 1, eliminating the overall amplitude deviation caused by differences in user contact pressure in different acquisition segments, and providing a standardized input signal for subsequent component decomposition.

[0024] The adaptive complete set empirical mode decomposition method is used to normalize the single-segment original signal S. N By performing layer-by-layer adaptive decomposition, several intrinsic mode function components with different center frequencies and a residual component characterizing the overall trend of the signal are obtained.

[0025] Based on the inherent frequency band characteristics of the human fingertip pulse pressure physiological signal, the effective frequency band of the pulse physiological signal is preset to be 0.5Hz to 30Hz. Frequency domain features are extracted and clustered from all the decomposed intrinsic mode function components. Intrinsic mode function components whose center frequency is within the preset effective frequency band and whose modal amplitude fluctuations conform to the normal human pulse cycle variation are superimposed and integrated into a stable physiological component S. N_s The intrinsic mode function components with center frequencies higher than the preset effective frequency band upper limit and exhibiting instantaneous burst amplitude jump characteristics are superimposed and integrated into an instantaneous interference component S. N_n The intrinsic mode function components with center frequencies below the preset effective frequency band lower limit are superimposed and integrated with the residual components obtained from decomposition to generate a low-frequency drift component S. N_d .

[0026] After completing the division of a single signal into three types of components, a component orthogonality check is performed to verify that there is no signal coupling or feature overlap between the three types of components. After the check passes, the component separation operation of the original signal segment is completed. The same processing flow is used to complete the component decomposition processing of all original single-segment signals in the ordered original signal dataset S.

[0027] The process involves eliminating transient interference components, correcting low-frequency drift components, and performing feature-protective smoothing on stable physiological components to generate a preprocessed effective signal set, such as... Figure 2 As shown, the specific implementation includes: processing the instantaneous interference component S corresponding to each original single-segment signal that has completed the orthogonality check. N_n The system performs a full-band zeroing removal operation to completely eliminate high-frequency burst noise and instantaneous jump interference caused by contact jitter in the signal. It does not touch the effective frequency band data of stable physiological components throughout the process, thus avoiding the attenuation or loss of pulse characteristics related to TCM organ differentiation.

[0028] For the low-frequency drift component S that is separated synchronously N_d First, a third-order polynomial fitting algorithm is used to fit the baseline drift trend curve of the signal. Then, the low-frequency drift component is subtracted from the fitted baseline trend curve to complete the baseline correction and trend elimination. This eliminates the slow baseline shift caused by the user's grip on the device and changes in body position. During the correction process, the effective features of extremely low frequencies related to physiological signals are fully preserved, and the slow-cycle signal features related to constitution differentiation are avoided from being mistakenly eliminated.

[0029] For the stable physiological component S that has completed component separation N_sThe smoothing process employs a sliding window weighted moving average method, with a fixed sliding window length of 5 sampling points. The window weights are distributed according to a Gaussian distribution, with the highest weight at the center of the window and decreasing linearly towards both sides. Before smoothing, the positions and amplitudes of the three core feature points of the pulse wave—the main peak, tidal wave, and dicrotic wave—are pre-locked. The sampling intervals containing these feature points are protected by weight locking and do not participate in the smoothing calculation. Smoothing is only performed on the transition intervals between feature points. This process filters out residual minor noise while fully preserving the core morphological features of the pulse wave that are related to the deficiency and excess of the internal organs and the constitution classification in Traditional Chinese Medicine, thus avoiding the weakening of effective diagnostic information by smoothing.

[0030] After all the differential preprocessing of the three types of components of a single signal is completed, the processed stable physiological components and the ultra-low frequency effective physiological features retained in the corrected low-frequency drift components are superimposed to generate the preprocessed effective signal S corresponding to that signal segment. N_p According to the original acquisition time sequence, all preprocessed valid signals corresponding to single signal segments are arranged in an ordered manner to generate an ordered preprocessed valid signal set P={S 1_p ,S 2_p ,...,S N_p}

[0031] The establishment of a signal consistency evaluation model, such as Figure 3 As shown, the implementation specifically includes: constructing a dedicated dataset for model training and validation based on labeled clinical pulse physiological signal samples from hospitals and pulse physiological signal samples collected from the general population. The samples in the dataset are labeled to distinguish between stable and valid signal samples that conform to the TCM collection standards and abnormal signal samples caused by improper operation or environmental interference. At the same time, the dataset is stratified according to age, gender, constitution type, and the state of deficiency and excess of the internal organs to ensure that the dataset includes the characteristics of people of all ages and different health states, and to eliminate the impact of sample bias on the model's generalization ability.

[0032] To address the inherent characteristics of preprocessed effective signals, a fixed-dimensional consistency feature system is constructed. This system is divided into two categories: basic signal consistency features and TCM syndrome differentiation correlation features. The basic signal consistency features include the mean amplitude, peak standard deviation, period variation coefficient, and waveform correlation coefficient of a single signal segment, as well as the effective frequency band energy ratio, harmonic energy distribution dispersion, and baseline fluctuation amplitude in the frequency domain. The TCM syndrome differentiation correlation features include the amplitude of the main peak of the pulse wave, the relative amplitude ratio of the tidal wave and dicrotic wave, and the ratio of the ascending limb slope to the descending limb slope. All features are standardized to eliminate dimensional differences, generating a consistency feature vector corresponding to each sample segment, which serves as the standard input to the model.

[0033] A signal consistency evaluation model is constructed using a composite algorithm architecture that integrates Mahalanobis distance discriminant analysis and support vector machine (SVM) classification. First, Mahalanobis distance is used to calculate the distance between the feature vector of a single signal segment and the feature center of a standard stable signal within the dataset, completing the initial screening of anomalies in the feature space. Then, the screened feature vectors are input into a pre-trained SVM model to achieve accurate classification and quantitative scoring of signal stability. During model training, five-fold cross-validation is used to optimize model parameters. Clinical annotation results are used as the gold standard, and the classification threshold and weight parameters are iteratively adjusted to ensure that the model's accuracy in identifying anomalous signal segments is no less than 95%, and its recall rate for identifying stable and valid signal segments is no less than 98%. The model construction specifically includes: The m-dimensional original consistency feature vector X of the input single-segment signal i =[x i1 ,x i2 ,...,x im ] T Where m is the fixed feature dimension, i.e., the total number of basic signal consistency features and TCM syndrome differentiation correlation features, and i is the signal segment index. Z-score standardization based on standard stable positive samples in the training set is adopted, with the formula: z i =(X i -μ) σ; in, This is the Hadamard element-wise division operator; μ is the mean vector of features of only standard stable positive samples in the training set, μ=[μ1,μ2,...,μ m ] T σ is the standard deviation vector of each feature of only the standard stable positive samples in the training set, and σ is constrained. j >0, σ=[σ1,σ2,...,σ m ] T ; Output a dimensionless m-dimensional standardized feature vector zi, which serves as the sole input to the model.

[0034] The Mahalanobis distance anomaly index is used for initial screening to address matrix singularity and threshold logic issues. Extreme anomalies are initially screened based on the feature distribution of standard stable samples, reducing SVM classification noise and adapting to the coupling characteristics of traditional Chinese medicine pulse features. The formulas include: The regularized covariance matrix of the standard stable samples in the pre-stored training set: ; Calculate the Mahalanobis distance between the input vector and the standard feature centers: ; in, To ensure a stable total number of positive samples in the training set; The mean vector of the standardized feature vectors of the standard stable samples; The non-negative regularization coefficient is used, with the default value being [value missing]. Solve the problem of singular and non-invertible covariance matrices; for An identity matrix of order 1; It is the inverse of the regularized covariance matrix.

[0035] Based on the chi-square distribution setting, the squared Mahalanobis distance Obeying the degree of freedom The chi-square distribution, take The critical value corresponding to the confidence level This is the threshold. If... Segments directly marked as anomalous are output with a consistency score. ;like If the initial screening is passed, the sample will proceed to SVM for precise classification.

[0036] This paper utilizes an SVM (Simultaneous Vector Machine) module for accurate classification and quantitative scoring to address the issues of probability calibration and parameter ambiguity. A binary SVM with an RBF kernel, optimized through five-fold cross-validation, is employed. After Platt probability calibration, the SVM is mapped to the original technique's 0-100 score range. The formula includes: Pre-trained and fixed SVM decision function: ; Wherein, the RBF kernel function is: ; Platt probability calibration generates statistically rigorous positive sample probability values: ; Linearly mapped to a scoring range of 0-100 points, strictly tied to the passing score: ; In the formula, This represents the total number of support vectors. For the non-zero Lagrange multipliers corresponding to the support vectors, and for the non-support vectors... It is not included in the calculation; For sample labels, As a standard stable positive sample, These are abnormal negative samples; The standardized feature vectors corresponding to the support vectors; For the bias term of the SVM decision hyperplane; The RBF core bandwidth parameters were optimized and solidified through five-fold cross-validation. The square of the Euclidean distance between the two vectors; The Platt calibration parameters are solidified using maximum likelihood estimation of the training set to ensure... This is a strict statistical probability value.

[0037] The classification rules are as follows Marked as an abnormal segment, Marked as a valid stable signal segment, It also serves as the basis for weighting subsequent signal weighting and reconstruction.

[0038] After completing model training and parameter optimization, the model's calculation parameters and judgment rules are fixed. The preset consistency score range is 0 to 100 points, and a fixed pass threshold of 60 points is preset. The higher the consistency score output by the model, the higher the standardization of the corresponding signal segment acquisition, the stronger the stability of physiological characteristics, and the higher the matching degree with the TCM diagnostic standards. Signal segments with a score below 60 points will be marked as abnormal segments, and signal segments with a score of not less than 60 points will be marked as valid and stable signal segments. At the same time, the consistency score output by the model will be directly used as the basis for weight allocation in subsequent signal weighted superposition and reconstruction.

[0039] After the model completes the entire clinical validation process, all calculation parameters and judgment rules will be solidified and embedded into the signal processing link of the TCM health assessment system. It will only receive the consistency feature vector of a single signal segment within the preprocessed effective signal set as the sole input, and output the consistency score and classification label of the corresponding signal segment. The relevant parameters of the TCM health assessment model will not be modified throughout the process, and the subsequent extraction process of TCM syndrome differentiation features will not be interfered with, thus ensuring the independence of the model operation and the stability of the results.

[0040] The pre-trained signal consistency evaluation model is invoked to calculate the consistency score of each segment of the effective signal, such as... Figure 4 As shown, the specific implementation includes: for each segment of preprocessed valid signal S within the ordered preprocessed valid signal set P. N_p Processing is performed sequentially according to the original acquisition time. First, the single segment of valid signal S is preprocessed. N_p Complete the extraction of consistency features in fixed dimensions, extracting two types of feature data: basic signal consistency features and TCM syndrome differentiation correlation features. The basic signal consistency features include the mean amplitude, peak standard deviation, period variation coefficient, and waveform correlation coefficient of a single signal in the time domain, as well as the effective frequency band energy ratio, harmonic energy distribution dispersion, and baseline fluctuation amplitude in the frequency domain. The TCM syndrome differentiation correlation features include the amplitude of the main peak of the pulse wave, the relative amplitude ratio of the tidal wave and the diphthong wave, and the ratio of the ascending limb slope to the descending limb slope.

[0041] For all extracted feature data, Z-score standardization is performed in the same way as in the model training phase to eliminate the dimensional differences between different features and generate a single-segment signal consistency feature vector that perfectly matches the input dimension of the pre-trained signal consistency evaluation model, ensuring that the input data format is completely consistent with the model training requirements.

[0042] The standardized consistency feature vector is input into the pre-trained signal consistency evaluation model with fixed parameters. First, the Mahalanobis distance discriminant module built into the model is used to calculate the Mahalanobis distance between the feature vector and the standard stable signal feature center stored in the model, thus completing the initial screening and matching quantification of anomalies in the feature space. Then, the feature vector after initial screening is input into the pre-trained support vector machine classification module built into the model. Based on the fixed and optimized weight parameters and classification rules of the model, the accurate quantification calculation of signal stability is completed.

[0043] The model outputs a consistency score corresponding to the preprocessed valid signal within a fixed range of 0 to 100. The score results are bound one-to-one with the original signal acquisition sequence and temporarily stored. The output consistency score directly corresponds to the acquisition standardization, physiological characteristic stability, and matching degree with the TCM diagnostic criteria of the corresponding signal segment. This provides a unique quantitative basis for subsequent abnormal segment removal and signal weighted superposition reconstruction. The model keeps the parameters fixed throughout the calculation process, does not modify the relevant parameters of the TCM health assessment model, and does not interfere with the subsequent extraction process of TCM diagnostic features.

[0044] The process involves removing abnormal segments with scores below a preset threshold, then reconstructing the remaining valid signal segments by weighted summation based on their scores, to generate a standard physiological signal, such as... Figure 5 As shown, the specific implementation includes: comparing the consistency score of all single-segment signals within the ordered preprocessed valid signal set with a preset pass threshold of 60 points; marking signal segments with a consistency score below 60 points as abnormal segments; permanently removing abnormal segments and simultaneously removing the original acquisition timing number, feature data, and preprocessing results corresponding to the abnormal segments, excluding them from subsequent reconstruction calculations; and marking signal segments with a consistency score of not less than 60 points as valid stable signal segments, retaining their complete preprocessed valid signal data, original acquisition timing information, and corresponding consistency score to form a valid signal subset.

[0045] If, after removing abnormal segments, the number of valid stable signal segments is less than the preset minimum number of valid segments (2 segments), a retry prompt will be triggered, guiding the user to perform the signal acquisition process again after following the correct procedure. If the number of valid stable signal segments is greater than or equal to 2 segments, the subsequent weighted overlay reconstruction stage will begin.

[0046] For all valid stable signal segments within the valid signal subset, the consistency score corresponding to each signal segment is used as the sole basis for weight allocation. Normalized weight calculation is performed, and the weight value of a single signal segment is equal to the consistency score of that signal segment divided by the sum of the consistency scores of all valid stable signal segments within the valid signal subset. This ensures that the sum of the weight values ​​of all valid signal segments is 1, and the signal segment with the higher the consistency score has a larger weight ratio and a higher contribution to the final reconstructed signal.

[0047] Before weighted superposition, pulse cycle reference alignment processing is performed on all effective stable signal segments. The effective stable signal segment with the earliest time sequence in the effective signal subset is used as the reference segment. The position of the main wave peak of the pulse wave in the reference segment is locked as the period alignment reference point. The main wave peak positions of the remaining effective stable signal segments are sequentially time-domain shifted and aligned with the reference point. During the alignment process, the core morphological characteristics of the pulse wave of each signal segment are completely preserved, the amplitude and time sequence relative relationship of the signal are not changed, and the period phase deviation between adjacent acquisition segments is eliminated.

[0048] After completing the period alignment, the signal amplitude at each sampling time is weighted and superimposed according to the normalized weight value corresponding to each effective stable signal segment. The amplitude of all signal segments at the same sampling time is multiplied by the corresponding weight value and then summed to obtain the reconstructed signal amplitude at that sampling time. The weighted superposition calculation of the entire sampling interval is completed in sequence according to the sampling time order to generate the initial reconstructed physiological signal.

[0049] The initial reconstructed physiological signal undergoes integrity and validity verification. First, it is verified that the time domain length and sampling frequency of the reconstructed signal are completely consistent with the preset parameters of the original single-segment acquired signal. Then, it is verified that the position and amplitude of the three core feature points of the pulse wave main peak, tidal wave, and dicrotic wave are completely preserved without feature loss or distortion. At the same time, it is verified that the energy ratio of the effective frequency band of the signal meets the preset requirements of the effective frequency band of pulse physiological signal from 0.5Hz to 30Hz. If the verification passes, the initial reconstructed physiological signal is marked as the calibrated standard physiological signal. If the verification fails, the period alignment and weighted superposition calculation of the effective signal segment are re-executed until a standard physiological signal that meets the verification requirements is generated.

[0050] The method of generating TCM health assessment results based on standard physiological signals, such as Figure 6 As shown, the specific implementation includes: first, performing a full-dimensional compliance review on the calibrated standard physiological signal after verification to confirm that the signal sampling frequency is completely consistent with the original preset fixed sampling frequency Fs, the signal time domain length is completely matched with the preset duration T of a single segment acquisition, the energy proportion in the effective frequency band from 0.5Hz to 30Hz meets the preset threshold, and the position and amplitude of the three core feature points of the pulse wave main peak, tidal wave, and dicrotic wave are complete and without distortion. After the review is passed, the feature extraction process begins.

[0051] For the standard physiological signals that have passed the review, a fixed-dimensional set of physiological features bound to the TCM syndrome differentiation system is extracted. The feature set is divided into three categories: time-domain amplitude features, frequency-domain harmonic features, and time-frequency-domain energy distribution features. The time-domain amplitude features include the amplitude of the pulse wave main peak, the relative amplitude of the tidal wave, the relative amplitude of the dicrotic wave, the slope of the ascending limb, the slope of the descending limb, the duration of the pulse cycle, the width of the main peak, and the coefficient of variation of amplitude between adjacent cycles. The frequency-domain harmonic features include the proportion of pulse wave fundamental frequency energy, the proportion of energy distribution of the 2nd to 10th harmonics, the amplitude ratio of the fundamental frequency to each harmonic, and the dispersion of harmonic energy within the effective frequency band. The time-frequency-domain energy distribution features include the proportion of energy during systole and diastole within a single pulse cycle and the consistency coefficient of time-frequency energy distribution between different cardiac cycles. All features correspond one-to-one with the input feature dimensions of the pre-trained model, with no additional features introduced or core features missing.

[0052] For the extracted full-dimensional feature data, Z-score standardization processing is performed in complete consistency with the training phase of the TCM health assessment model to eliminate the dimensional differences between different features and generate standardized feature vectors that perfectly match the input dimensions of the pre-trained TCM health assessment model, ensuring that the input data format is fully compatible with the model training requirements.

[0053] The standardized feature vectors are input into a pre-trained TCM health assessment model with fixed parameters. Based on the input feature vectors, the model performs full-dimensional calculations and outputs the preliminary TCM health assessment results. These results include three core data categories: First, quantitative scores for organ deficiency / excess, including quantitative values ​​for heart deficiency, spleen deficiency, stomach deficiency, lung deficiency, kidney deficiency, liver deficiency, large intestine deficiency, small intestine deficiency, bladder deficiency, and gallbladder deficiency. Each value is matched with a preset normal reference range, and values ​​exceeding the reference range are marked as abnormal. Second, constitution classification results, including classifications for balanced constitution, yang deficiency constitution, yin deficiency constitution, qi deficiency, qi stagnation, blood deficiency, blood stasis, body fluid deficiency, body fluid stagnation, cold, and heat. Each classification result provides a confidence level, and biased constitutions exceeding warning thresholds are marked as risky. Third, quantitative indicators of disease tendency, corresponding to the risk level classification of each syndrome differentiation dimension, simultaneously generate an overall health index value.

[0054] After the initial assessment results are generated, the results are marked as TCM health assessment results R1. Simultaneously, the correlation between each quantitative indicator and the corresponding physiological characteristic in the results is traced and marked to ensure that each assessment result can be mapped to a specific characteristic source in the standard physiological signal, providing traceable benchmark data for subsequent stability verification and closed-loop feedback optimization.

[0055] The fluctuation values ​​of the comparison output assessment results and the benchmark results are compared. If the fluctuation exceeds a preset threshold, the component processing parameters are automatically adjusted, and the signal decomposition, consistency score calculation, and generation of standard physiological signals and traditional Chinese medicine health assessment results are repeated until the results meet the stability requirements. Figure 7 As shown, the implementation specifically includes: setting a fixed fluctuation threshold Th for the TCM health assessment results in advance. For example, the fluctuation threshold Th is fixed at 10%. At the same time, corresponding sub-item fluctuation thresholds are preset for the three core assessment indicators: the quantitative score of the deficiency and excess of the internal organs, the confidence level of the constitution classification, and the overall health index. The sub-item thresholds do not exceed 15%.

[0056] First, all valid signal segments in the preprocessed valid signal set P are reconstructed by superimposing equal weights to obtain a baseline physiological signal that has not undergone consistency weighting and closed-loop optimization. This baseline physiological signal is then input into a TCM health assessment model with fixed parameters. The model outputs the corresponding baseline result R0. The baseline result R0 includes three core data categories that are completely corresponding to the initial judgment result R1: the quantitative score of visceral deficiency and excess, the constitution classification result, and the quantitative index of disease tendency. This ensures that the two indicators are completely matched in terms of dimensions and are comparable.

[0057] For each corresponding core evaluation indicator, the relative deviation between the preliminary result R1 and the benchmark result R0 is calculated. The relative deviation is calculated as follows: |R1 corresponding indicator value - R0 corresponding indicator value| / R0 corresponding indicator value × 100%; Based on the relative deviations of all core indicators, the comprehensive result fluctuation value ΔR is calculated according to the preset fixed weight ratio. Among them, the weight of the quantitative score of visceral deficiency and excess accounts for 60%, the weight of the constitution classification related indicators accounts for 25%, the weight of the disease tendency quantitative indicators accounts for 15%, and the total weight is 100%.

[0058] The calculated comprehensive result fluctuation value ΔR is compared with the preset fixed fluctuation threshold Th. If ΔR≤Th and the relative deviation values ​​of all sub-indicators do not exceed the corresponding sub-item fluctuation threshold, then the preliminary judgment result R1 is determined to meet the stability constraint and is marked as the result to be output. If ΔR > Th, or if the relative deviation of any sub-item exceeds the corresponding sub-item fluctuation threshold, a closed-loop feedback optimization mechanism is triggered. This mechanism automatically performs iterative adjustments to the component processing parameters and preprocessing process. Priority is given to adjusting the modal component clustering weight parameters in the component decomposition stage. The clustering weights of the stable physiological components corresponding to the effective frequency bands from 0.5Hz to 30Hz are gradually increased, and the frequency band division boundaries of the transient interference components and low-frequency drift components are tightened to ensure that more effective features related to TCM syndrome differentiation are fully incorporated into the stable physiological components. If the fluctuation value recalculated after a single parameter adjustment still exceeds the threshold, the strategy parameters of the preprocessing stage are iteratively adjusted sequentially. This includes adjusting the polynomial fitting order of the low-frequency drift component baseline correction, adjusting the sliding window length and Gaussian weight allocation coefficient of the stable physiological component feature protection smoothing process, and always locking the protection rules of the three core feature points of the pulse wave main peak, tidal wave, and dicrotic wave during the adjustment process. The core logic of feature point locking protection must not be modified, and the fixed calculation parameters of the TCM health assessment model and the signal consistency evaluation model must not be changed throughout the entire process.

[0059] After each adjustment of component processing parameters and preprocessing process, the entire process of multi-component separation and differential preprocessing of the original signal, signal consistency feature extraction and scoring calculation, abnormal segment removal and weighted superposition reconstruction of standard physiological signals, and generation of preliminary judgment results of TCM health assessment are re-executed according to the established procedure. In each iteration, the comprehensive result fluctuation value ΔR and the relative deviation value of the sub-indicators between the newly generated preliminary judgment result and the benchmark result R0 are recalculated.

[0060] During the iteration process, if the fluctuation value still fails to fall below the threshold after three consecutive parameter adjustments, the system automatically switches to a conservative parameter adjustment mode. This lowers the pass threshold for the consistency score of the effective signal segment to 50 points, incorporates more highly consistent effective signal segments into the weighted reconstruction, and simultaneously tightens the frequency band division rules for component decomposition until the combined fluctuation value ΔR of the generated evaluation result and the benchmark result are less than or equal to Th. Furthermore, the relative deviation values ​​of all sub-indicators meet the threshold requirements. This completes the feature enhancement optimization and generates a result that meets the stability requirements. Simultaneously, a maximum number of iterations is preset. If the stability constraint is still not met after reaching the maximum number of iterations, the iteration process is terminated, triggering an acquisition anomaly prompt. This guides the user to eliminate environmental interference, standardize operations, and then re-execute the signal acquisition process to avoid invalid iterations.

[0061] The process of generating a TCM health assessment report that meets stability requirements includes the following steps: First, a full-dimensional compliance and consistency final review is performed on the output results that have completed stability constraint verification. This confirms that all core indicators of the output results meet the preset stability threshold requirements. Simultaneously, the standard physiological signals, signal processing parameters, and feature extraction traceability links of this test are reviewed to ensure they are complete and traceable. The review process does not modify any assessment indicator values ​​or signal processing parameters; it only completes the compliance verification. After the review is passed, the standardized report generation stage begins.

[0062] First, solidify the basic information of this test, including the unique identifier of the testing entity, the testing time, the model of the acquisition equipment, the sampling frequency, the duration of a single acquisition segment, the total number of acquisition segments, the number of effective stable signal segments, and the acquisition environment label. All basic information corresponds one-to-one with the original preset parameters of this acquisition process, without any tampering or omission.

[0063] This system provides a complete output of the core TCM health assessment indicators for this test, simultaneously presenting quantitative values ​​for the following organ dimensions: heart deficiency, spleen deficiency, stomach deficiency, lung deficiency, kidney deficiency, liver deficiency, large intestine deficiency, small intestine deficiency, bladder deficiency, and gallbladder deficiency. It also presents the preset normal reference range for each value, highlighting values ​​exceeding the reference range and indicating the degree of deviation. Furthermore, it presents the classification results for constitution and syndrome elements, including balanced constitution, yang deficiency constitution, yin deficiency constitution, qi deficiency, qi stagnation, blood deficiency, blood stasis, body fluid deficiency, body fluid stagnation, cold, and heat. The system also presents the confidence level for each classification result, marking risk warnings for imbalanced constitutions and syndrome elements exceeding warning thresholds and indicating the corresponding risk level. Finally, it presents the overall health index value for this test, along with quantitative indicators and risk level classifications for disease tendencies across each syndrome differentiation dimension, and highlights key health risk points requiring attention.

[0064] The system outputs complete quantitative data on signal quality and stability assessment for this test, including the total number of original acquired signal segments, the percentage of effective stable signal segments, the average consistency score of a single signal segment, the number of abnormal segments removed and the corresponding abnormality reasons. It also presents the effective frequency band energy percentage of the standard physiological signal after calibration, the integrity verification results of the core feature points of the pulse wave, and the quantitative values ​​of stability assessment results, including the comprehensive fluctuation value between the final assessment result and the benchmark result, the relative deviation values ​​of each core assessment indicator, and indicates that all values ​​meet the preset stability threshold requirements, thus achieving a repeatable and reliable quantitative presentation of the assessment results.

[0065] Based on the core results of the output of visceral deficiency and excess and body constitution classification, corresponding TCM syndrome differentiation interpretation content is generated, and the correspondence between the syndrome differentiation results and the core characteristics of the standard physiological signals is marked. Then, based on the syndrome differentiation results, corresponding personalized conditioning directions are matched. The conditioning directions are divided into two categories: basic conditioning and special conditioning. Basic conditioning includes dietary advice, recipes, acupoint therapy, and health exercises. Special conditioning includes recommendations for food and medicine with the same origin, health teas, moxibustion, massage, foot baths, and suggestions for the appropriate use of Chinese patent medicines. All conditioning suggestions correspond one-to-one with the syndrome differentiation results of this test, with no generic or undifferentiated content.

[0066] The system synchronously outputs traceable information for the entire testing process, including signal component separation parameters, consistency evaluation model operating parameters, fixed parameters of the TCM health assessment model, closed-loop feedback optimization iteration count, and parameter adjustment records. This ensures that the entire assessment process is traceable and reproducible. Additionally, a usage note is included at the end of the report, indicating that this report is for TCM health status assessment reference only and does not constitute a Western medicine disease diagnosis. Related treatment plans must be implemented under the guidance of a licensed TCM physician. A standardized TCM health assessment report is generated after full content verification.

[0067] This application also provides an embodiment of an electronic device. The electronic device is manifested in the form of a general-purpose computing device. The components of the electronic device may include, but are not limited to: one or more processors or processing units, memory, and buses connecting different components (including memory and processing units).

[0068] A bus refers to one or more of several bus architectures, including memory buses or memory controllers, peripheral buses, graphics acceleration ports, processors, or local buses using any of the various bus architectures. Examples of these architectures include, but are not limited to, Industry Standard Architecture (ISA) buses, Micro Channel Architecture (MCA) buses, Enhanced ISA buses, Video Electronics Standards Association (VESA) local buses, and Peripheral Component Interconnect (PCI) buses.

[0069] Electronic devices typically include a variety of computer-readable media. These media can be any available media that can be accessed by the electronic device, including volatile and non-volatile media, and removable and non-removable media.

[0070] The memory may include computer-readable media in the form of volatile memory, such as random access memory (RAM) and / or cache memory. Electronic devices may further include other removable / non-removable, volatile / non-volatile computer device storage media. By way of example only, the storage system may be used to read and write non-removable, non-volatile magnetic media.

[0071] The electronic device can also communicate with one or more external devices (e.g., keyboard, pointing device, camera, etc.), may include a display, and may communicate with one or more devices that enable a user to interact with the electronic device, and / or with any device that enables the electronic device to communicate with one or more other computing devices (e.g., network card, modem, etc.). This communication can be performed via an input / output (I / O) interface. Furthermore, the electronic device can communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN)) and / or public networks, such as the Internet) via a network adapter. The network adapter communicates with other modules of the electronic device via a bus. The processor executes various functional applications and data processing by running programs stored in memory, such as implementing the traditional Chinese medicine health assessment method based on physiological signal enhancement provided in the above embodiments of this application.

[0072] This application also discloses a traditional Chinese medicine health assessment system based on physiological signal enhancement, comprising: a signal acquisition module for receiving pulse physiological signals from a finger acquisition device and generating an ordered raw signal dataset; a signal decomposition module for decomposing each raw signal segment into three independent components: stable physiological components, transient interference components, and low-frequency drift components; a signal preprocessing module for removing transient interference components, correcting low-frequency drift components, performing feature-protected smoothing processing on stable physiological components, and generating a preprocessed effective signal set; a model building module for establishing a signal consistency evaluation model; and a signal reconstruction module for calling pre-trained signal consistency evaluation models. The evaluation model calculates the consistency score of each effective signal segment, removes abnormal segments with scores below a preset threshold, and reconstructs the remaining effective signal segments by weighted superposition according to the scores to generate a standard physiological signal. The evaluation feedback module generates a TCM health assessment result based on the standard physiological signal, compares the fluctuation value of the output assessment result with the baseline result, and if the fluctuation exceeds a preset threshold, automatically adjusts the component processing parameters and repeats the signal decomposition, consistency score calculation, standard physiological signal generation, and TCM health assessment result generation until the result meets the stability requirements. The report output module outputs a TCM health assessment report that meets the stability requirements.

[0073] Finally, it should be noted that the above description is only a preferred embodiment of this application and is not intended to limit this application. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A traditional Chinese medicine health assessment method based on physiological signal enhancement, characterized in that, Includes the following steps: The pulse physiological signal is collected in a single detection process using a finger sampling device, generating an ordered raw signal dataset. Each original signal segment is decomposed into three independent components: stable physiological components, transient interference components, and low-frequency drift components. This process involves: for each original single-segment signal in the original signal dataset, time-domain amplitude normalization is first performed. The normalized single-segment original signal is then subjected to layer-by-layer adaptive decomposition to obtain several intrinsic mode function components and one residual component. Based on the inherent frequency band characteristics of the human fingertip pulse pressure physiological signal, frequency domain features are extracted and clustered on the decomposed components. These components are then superimposed and integrated into three independent components: stable physiological components, transient interference components, and low-frequency drift components. After the three-segment component division of a single signal is completed, component orthogonality verification is performed. Once the verification passes, the component separation of that signal segment is completed. The same process is then used to complete the component decomposition of all original single-segment signals in the dataset. The process involves removing transient interference components, correcting low-frequency drift components, and performing feature-protective smoothing on stable physiological components to generate a preprocessed effective signal set. This includes: first, zeroing out transient interference components from each segment of the original signal after orthogonality verification, without touching the effective frequency band data of stable physiological components; then, performing baseline correction and trend removal on low-frequency drift components, fully preserving the extremely low-frequency effective physiological characteristics; performing feature-protective smoothing on stable physiological components, locking the core feature points of the pulse wave to exclude them from smoothing calculations, smoothing the transition intervals between feature points, filtering out residual noise while fully preserving the core morphological characteristics of the pulse wave related to TCM syndrome differentiation; after completing the differentiated preprocessing of the three types of components, superimposing the processed stable physiological components with the effective features preserved by the corrected low-frequency drift components to generate a single segment of preprocessed effective signal, and arranging them according to the original acquisition sequence to generate an ordered preprocessed effective signal set. Establish a signal consistency evaluation model; The pre-trained signal consistency evaluation model is invoked to calculate the consistency score of each effective signal segment. Abnormal segments with scores below the preset qualified threshold are removed. The remaining effective signal segments are reconstructed by weighted superposition according to the scores to generate a standard physiological signal. This includes: weighted superposition reconstruction based on the consistency score of each effective stable signal segment as the weight allocation basis. After completing the normalized weight calculation and aligning with the pulse cycle benchmark, the initial reconstructed physiological signal is generated by weighted superposition according to the weights. After passing the integrity and validity verification, the standard physiological signal is generated. If the verification fails, the cycle alignment and weighted superposition calculation are re-executed. Based on standard physiological signals, a TCM health assessment result is generated. The fluctuation value of the output assessment result is compared with that of the baseline result. If the fluctuation of the two exceeds the preset threshold, the component processing parameters are automatically adjusted and the signal decomposition, consistency score calculation, standard physiological signal and TCM health assessment result are repeated until the result meets the stability requirements. Output a TCM health assessment report that meets stability requirements.

2. The TCM health assessment method based on enhanced physiological signals according to claim 1, characterized in that, Using a finger sampling device, pulse physiological signals are collected in a single detection process to generate an ordered raw signal dataset. This includes: collecting pulse pressure physiological signals from the fingertip of the human body using a finger sampling device, performing multiple consecutive short-term physiological signal acquisitions, using a fixed sampling frequency during the acquisition process, controlling the interval between adjacent acquisition segments, and generating an ordered raw signal dataset according to the chronological order of the acquisitions after the entire acquisition process is completed.

3. The method for TCM health assessment based on enhanced physiological signals according to claim 1, characterized in that, The establishment of the signal consistency evaluation model includes: constructing a hierarchical dataset based on labeled clinical and socially collected pulse physiological signal samples, including populations of different health statuses across all age groups; constructing a consistency feature system for preprocessed effective signals, including basic signal consistency features and TCM syndrome differentiation correlation features; generating consistency feature vectors as model inputs through standardization; constructing a signal consistency evaluation model; optimizing model parameters through cross-validation; solidifying model parameters and judgment rules after training and clinical validation; embedding the model into the system signal processing link; and independently outputting the consistency score and classification label for the corresponding signal segment.

4. The TCM health assessment method based on physiological signal enhancement according to claim 3, characterized in that, The pre-trained signal consistency evaluation model is invoked to calculate the consistency score of each effective signal segment. This includes: extracting two types of features, namely basic signal consistency and TCM syndrome differentiation, from each signal segment in the pre-processed effective signal set according to the original acquisition time sequence. After standardization, feature vectors matching the input dimensions of the pre-trained signal consistency evaluation model are generated. The model outputs the consistency score of the corresponding signal segment and stores the score in conjunction with the original acquisition time sequence.

5. The TCM health assessment method based on physiological signal enhancement according to claim 1, characterized in that, If the fluctuation values ​​of the comparison output evaluation results and the benchmark results exceed a preset threshold, the component processing parameters are automatically adjusted and the signal decomposition, consistency score calculation, standard physiological signal generation, and TCM health assessment result generation are repeated until the results meet the stability requirements. This includes: setting the comprehensive fluctuation threshold of the TCM health assessment results and the sub-item fluctuation threshold of the core assessment indicators; reconstructing the benchmark physiological signal by superimposing the effective signal segments in the preprocessed effective signal set with equal weights; generating a dimension-matched benchmark result using a TCM health assessment model with fixed input parameters; calculating the relative deviation between the initial judgment result and the benchmark result; obtaining the comprehensive result fluctuation value according to the preset weight ratio; comparing it with the corresponding threshold; if it meets the constraint, the initial judgment result is marked as the result to be output; if it exceeds the threshold, closed-loop feedback optimization is triggered, iteratively adjusting the component processing parameters and preprocessing process, and re-executing the entire signal processing and evaluation process until the results meet the stability requirements. If the maximum number of iterations is still not met, the iteration is terminated and an abnormal acquisition prompt is triggered.

6. The TCM health assessment method based on physiological signal enhancement according to claim 1, characterized in that, The output of a TCM health assessment report that meets stability requirements includes: performing a final review of the full-dimensional compliance and consistency of the output results that have completed stability constraint verification; solidifying the basic information of this test after the review is passed; fully outputting the core TCM health assessment indicators, signal quality and assessment stability quantification data of this test; generating corresponding traceable TCM syndrome interpretation content and matching personalized conditioning direction based on the syndrome differentiation results; simultaneously outputting traceable information of the entire process of this test; marking the confirmed report usage prompts; and generating a standardized TCM health assessment report after completing full content verification.

7. A system utilizing the traditional Chinese medicine health assessment method based on physiological signal enhancement as described in claim 1, characterized in that, include: The signal acquisition module is used to receive the pulse physiological signals from the finger acquisition device and generate an ordered raw signal dataset; The signal decomposition module is used to decompose each segment of the original signal into three independent components: stable physiological components, transient interference components, and low-frequency drift components. The signal preprocessing module is used to remove transient interference components, correct low-frequency drift components, perform feature-protected smoothing on stable physiological components, and generate a preprocessed effective signal set. The model building module is used to build signal consistency evaluation models; The signal reconstruction module is used to call the pre-trained signal consistency evaluation model, calculate the consistency score of each effective signal segment, remove abnormal segments with scores lower than the preset qualified threshold, and reconstruct the remaining effective signal segments by weighted superposition according to the scores to generate standard physiological signals. The assessment feedback module is used to generate TCM health assessment results based on standard physiological signals, compare the fluctuation values ​​of the output assessment results with the benchmark results, and if the fluctuations of the two exceed the preset threshold, automatically adjust the component processing parameters and repeat the signal decomposition, calculation of consistency score, generation of standard physiological signals and TCM health assessment results until the results meet the stability requirements. The report output module is used to output TCM health assessment reports that meet stability requirements.

Images