Tongue fur and finger end microcirculation combined detection device based on multispectral fusion

Abstract

The application discloses a tongue fur and finger end microcirculation combined detection device based on multispectral fusion. The device synchronously applies multi-band optical excitation to the tongue detection area and the finger end detection area through an excitation unit, forming a multispectral excitation sequence; a collection unit obtains the reflection, absorption and scattering response of the tongue under different spectral excitation conditions, generating a tongue multi-layer spectral response data set; a response unit collects the deep optical interference and polarization change of the finger end, and extracts a finger end microcirculation dynamic parameter set; a modeling unit constructs a cross-tissue feature constraint relationship between the tongue tissue response and the finger end microcirculation response, and generates a joint feature constraint vector; a generation unit performs joint mapping calculation on the tongue multi-layer spectral response data set and the finger end microcirculation dynamic parameter set under the limitation of the joint feature constraint vector, and outputs a constitution type identifier and a constitution stability evaluation index, so that the collaborative constraint of multi-site physiological information and reliable constitution determination are realized.

Description

Technical Field

[0001] This invention relates to the field of medical testing technology, and in particular to a device for the combined detection of tongue coating and fingertip microcirculation based on multispectral fusion. Background Technology

[0002] In existing technologies, various optical-based detection devices and methods have emerged for human physical fitness assessment and health status detection. One type of technology uses visible light and near-infrared spectral imaging to analyze the tongue coating or tongue body, obtaining information on coating thickness, color distribution, and surface characteristics. Another type focuses on detecting the microcirculation status of fingertips or other peripheral areas, typically employing photoplethysmography, polarized light imaging, or optical coherence correlation (OCC) techniques to analyze blood flow changes, vascular distribution, or hemodynamic parameters. Furthermore, some existing solutions attempt to combine data from different detection sites to assist in health assessment or physical fitness analysis, but overall, they still primarily focus on single-object detection or simple parallel analysis.

[0003] However, the aforementioned existing technologies generally suffer from the problem of lacking intrinsic correlation modeling between detection sites. Tongue detection and fingertip microcirculation detection are often performed independently at different times and under different stimulation conditions. The resulting data lack a unified benchmark in terms of physical stimulation conditions and physiological response scales, which can easily introduce non-physiologically consistent interference features. At the same time, existing data fusion methods mostly rely on statistical superposition or empirical weight allocation, which makes it difficult to reflect the true response correlation of different tissues under the same stimulation conditions, resulting in deficiencies in the stability, repeatability, and reliability of the physical condition assessment results.

[0004] Therefore, it is necessary to propose a new technical solution to improve the overall mechanism for the collaborative acquisition and joint determination of physiological information from multiple sites. Summary of the Invention

[0005] This application provides a multispectral fusion-based joint detection device for tongue coating and fingertip microcirculation to improve the physiological consistency and stability of physical condition assessment results.

[0006] This application provides a device for the joint detection of tongue coating and fingertip microcirculation based on multispectral fusion, comprising: The excitation unit is used to simultaneously apply multi-band optical excitation to the tongue detection area and the fingertip detection area according to a unified timing control rule within a preset detection cycle, forming a multispectral excitation sequence with time distinguishability. The acquisition unit is used to synchronously acquire the reflection response, absorption response and apparent scattering changes of the tongue detection area under different spectral excitations under the constraint of multispectral excitation time sequence identification, and generate a multi-layer spectral response dataset of the tongue based on the joint spectral-temporal analysis method. The response unit is used to collect the deep optical interference response and polarization maintenance changes generated in the fingertip detection area under the same multispectral excitation conditions under the constraint of multispectral excitation timing identifier, and extract the dynamic parameter set of fingertip microcirculation through the timing consistency reconstruction method. The modeling unit is used to construct cross-tissue feature constraint relationships between tongue tissue response and fingertip microcirculation response based on the correlation between energy transfer consistency and physiological response time caused by multispectral excitation, under the common input of tongue multi-layer spectral response dataset and fingertip microcirculation dynamic parameter set. Under the constraint of cross-tissue feature constraint relationships, a joint feature constraint vector is generated to suppress non-physiological consistency features. The generation unit is used to perform joint mapping calculations on the tongue multilayer spectral response dataset and the fingertip microcirculation dynamic parameter set under the constraint of the joint feature constraint vector, so as to obtain the constitution determination result including at least the constitution type identifier and constitution stability assessment index.

[0007] This application has the following beneficial technical effects: (1) By applying multi-band optical excitation to the tongue and fingertip detection areas simultaneously under a unified temporal control rule, and forming a time-distinguishable multispectral excitation sequence, the physiological responses from different detection sites are established on a completely consistent optical stimulus and time reference. This reduces the systematic bias introduced by inconsistent detection timing or different excitation conditions from the source, and provides a stable and reliable physical basis for subsequent joint analysis of multi-source data. (2) The multi-layer spectral response dataset of the tongue is obtained by using a spectral-temporal joint analysis method, and the dynamic parameter set of fingertip microcirculation is extracted under the same temporal constraints. This allows the macroscopic surface features of the tongue and the deep microcirculation dynamic features of the fingertip to be characterized in a multi-dimensional and temporal form, which effectively enhances the ability to depict tissue level differences and physiological dynamic changes, and avoids the problem that a single static parameter is difficult to reflect the real physiological state. (3) By constructing a cross-tissue feature constraint relationship based on the correlation between energy transfer consistency and physiological response time, and generating a joint feature constraint vector, abnormal or isolated features that do not have physiological consistency in the tongue and fingertip detection data are suppressed, which significantly reduces the interference of noise features and random fluctuations on the constitution determination results and improves the credibility and stability of the multimodal feature fusion process. (4) Under the constraint of the joint feature constraint vector, the joint mapping calculation of the tongue multilayer spectral response dataset and the fingertip microcirculation dynamic parameter set is performed, so that the constitution determination results are generated only based on effective features that meet cross-tissue physiological consistency, thereby improving the accuracy and repeatability of constitution type identification, and being able to output constitution stability assessment indicators simultaneously, enhancing the application value of the detection results in continuous monitoring and health management scenarios. Attached Figure Description

[0008] Figure 1This is a schematic diagram of a multispectral fusion-based joint detection device for tongue coating and fingertip microcirculation provided in the first embodiment of this application. Detailed Implementation

[0009] Many specific details are set forth in the following description to provide a full understanding of this application. However, this application can be implemented in many other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of this application; therefore, this application is not limited to the specific embodiments disclosed below.

[0010] The first embodiment of this application provides a device for the joint detection of tongue coating and fingertip microcirculation based on multispectral fusion. Please refer to... Figure 1 This figure is a schematic diagram of the first embodiment of this application. The following is in conjunction with... Figure 1 The first embodiment of this application provides a joint detection device for tongue coating and fingertip microcirculation based on multispectral fusion.

[0011] The multispectral fusion-based joint detection device for tongue coating and fingertip microcirculation includes an excitation unit 101, a data acquisition unit 102, a response unit 103, a modeling unit 104, and a generation unit 105.

[0012] The excitation unit 101 is used to synchronously apply multi-band optical excitation to the tongue detection area and the fingertip detection area according to a unified timing control rule within a preset detection period, forming a time-distinguishable multispectral excitation sequence, and generating a multispectral excitation timing identifier that corresponds one-to-one with the multispectral excitation sequence.

[0013] In this invention, the excitation unit 101 serves as the basic execution unit of the entire combined detection device. Its core function is to provide the tongue detection area and the fingertip detection area with multi-band optical excitation that is highly consistent in physical conditions and can be accurately distinguished in temporal structure, thereby laying a unified excitation foundation for the comparability and constrainability of the physiological responses of different tissues.

[0014] Specifically, the excitation unit 101 includes at least a multi-band light source component, a timing control component, and a synchronization distribution component. The multi-band light source component generates an optical excitation signal covering multiple spectral ranges. The multi-band optical excitation includes at least the visible light band, the near-infrared band, and the short-wave ultraviolet band. Here, "band" refers to a spectral range with a defined center wavelength and bandwidth. For example, the visible light band can be selected from several discrete bands with center wavelengths between 450nm and 650nm; the near-infrared band can be selected from several discrete bands with center wavelengths between 750nm and 950nm; and the short-wave ultraviolet band can be selected from narrowband bands with center wavelengths between 360nm and 400nm. It should be noted that this invention does not require the use of a continuous spectrum, but rather preferably employs a combination of several discrete bands to reduce system complexity while ensuring physiological sensitivity.

[0015] The timing control component is used to uniformly schedule the emission time, duration, and switching order between adjacent bands of the multi-band light source component. Its core objective is to construct a multispectral excitation sequence with a clear temporal structure within a preset detection period. The "preset detection period" refers to the time interval covered from the start of the first band optical excitation to the end of the last band optical excitation during a complete detection process. This time interval can be set according to the system's sampling capability and the stability of the object under test, for example, in the range of hundreds of milliseconds to several seconds. By allocating a unique time window to each band optical excitation, the timing control component ensures that the excitations of different bands are distinguishable from each other on the time axis but arranged continuously as a whole, thus forming a time-distinguishable multispectral excitation sequence. "Time-distinguishability" means that any two different band optical excitations do not overlap on the time axis, and their start and end times can be accurately identified by the system's internal clock.

[0016] The synchronization distribution component, under the unified scheduling of the timing control component, synchronously applies multi-band optical excitation generated at the same time to the tongue detection area and the fingertip detection area. Synchronous application does not mean that the two detection areas must have completely identical illumination geometry; rather, it means that within the same time window, the optical excitation received by both detection areas belongs to the same band and follows the same on / off sequence. This synchronization distribution method ensures that when tongue tissue and fingertip tissue face the same spectral excitation, the difference in their physiological responses mainly stems from differences in tissue structure and function, rather than differences in external excitation conditions.

[0017] In practical implementation, the excitation unit 101 can also preset and calibrate the optical excitation intensity of different wavelengths according to the detection requirements, so as to avoid signal saturation or insufficient signal-to-noise ratio caused by excessive differences in the light absorption capacity of different tissues. For example, the safe and effective irradiation power range of the surface tissue of the tongue in the near-infrared band can be determined by preliminary experiments, and the output intensity of the corresponding wavelength band can be limited accordingly. The above calibration process falls within the scope of conventional optical safety control, and will not be repeated here.

[0018] Through the aforementioned structure and control method, the excitation unit 101 ultimately outputs not only simple illumination conditions, but also a multispectral excitation sequence containing both spectral and temporal information. This multispectral excitation sequence is assigned a unique temporal identifier within the system, which is used as a time reference by the subsequent acquisition unit 102 and response unit 103 during data acquisition and analysis. Therefore, the excitation unit 101 plays a crucial role in the entire technical solution by unifying excitation conditions, aligning time references, and constraining the comparability of multi-source physiological responses.

[0019] Furthermore, the excitation unit is specifically used for: Before the start of the preset detection cycle, at least one reference optical excitation is applied to the tongue detection area and the fingertip detection area to obtain the corresponding initial optical response, and a timing initialization parameter characterizing the current detection stable state of the object under test is generated based on the initial optical response. Based on the timing initialization parameters, the excitation duration of each spectral band within the preset detection period and the excitation interval between adjacent spectral bands are determined, thereby constructing a multispectral excitation sequence with non-equidistant temporal distribution characteristics. After the multispectral excitation sequence is constructed, a corresponding multispectral excitation timing identifier is generated for the excitation time window of each spectral band in the multispectral excitation sequence. The multispectral excitation timing identifier is used to characterize the one-to-one correspondence between the spectral band and the excitation time window. Multi-band optical excitation is applied synchronously to the tongue detection area and the fingertip detection area according to the multi-spectral excitation sequence and its corresponding multi-spectral excitation timing identifier.

[0020] In this embodiment, the multispectral excitation formation mechanism is further defined and refined. The purpose is to establish an excitation timing basis that matches the current state of the object being tested before the formal detection begins, thereby ensuring the comparability and consistency of subsequent tongue detection and fingertip detection in the time and spectral dimensions.

[0021] In actual operation, the excitation unit first enters the initialization phase before the start of the preset detection cycle. During this phase, the excitation unit does not immediately execute complete multispectral excitation, but instead applies at least one reference optical excitation to the tongue detection area and the fingertip detection area. The "reference optical excitation" referred to here is a standard form of optical excitation with a pre-fixed spectral composition and duration, and a low degree of tissue stimulation. Its purpose is not to extract specific physiological characteristics, but rather to observe the basic stability of the test subject's response to optical stimulation under the current detection environment and physical state. The reference optical excitation can be a single spectral band or a representative band from multiple spectral bands. The specific band selection does not affect the core idea of ​​this invention, as long as the excitation serves as a unified reference within the system.

[0022] When a reference optical excitation is applied to the tongue and fingertip detection areas, the excitation unit, in conjunction with the corresponding optical detection structure, acquires the corresponding initial optical response. The "initial optical response" referred to here is the reflection intensity, absorption degree, or other response signals that can be sensed by the optical sensor produced by the tongue and fingertip tissues under the reference optical excitation. This initial optical response is not used for subsequent physical assessment but rather to reflect the stability of the tissue response under the current detection conditions. For example, under the same reference excitation conditions, if the response of the tongue or fingertip fluctuates little within a short period, it indicates that the current state is relatively stable; conversely, if the response exhibits significant fluctuations, it indicates the presence of unstable factors such as posture changes, local blood flow fluctuations, or ambient light interference.

[0023] Based on the initial optical response described above, the excitation unit generates temporal initialization parameters to characterize the current stable state of the measured object. These "temporal initialization parameters" are a set of parameters obtained by organizing and quantizing the initial optical response over time, and they at least reflect the fluctuation amplitude and short-term consistency level of the initial response. For example, during the duration of the reference excitation, multiple initial optical responses can be compared. If the difference between adjacent sampling points is small, the corresponding temporal initialization parameters represent stability; if the difference is large, the corresponding temporal initialization parameters represent instability. The specific calculation method for these parameters can be completed using conventional statistical methods, and those skilled in the art can set them according to the sampling frequency and noise level of the actual system.

[0024] After obtaining the timing initialization parameters, the excitation unit enters the multispectral excitation sequence construction stage. In this stage, the excitation unit does not use a fixed excitation timing sequence, but rather determines the excitation duration of each spectral band within a preset detection period and the excitation interval between adjacent spectral bands based on the aforementioned timing initialization parameters. Here, "excitation duration" refers to the length of time for which optical excitation is continuously applied to a certain spectral band within the detection period; "excitation interval" refers to the time interval between the end of excitation of the previous spectral band and the start of excitation of the next spectral band. When the timing initialization parameters indicate that the measured object is in a relatively stable state, the excitation interval can be relatively shortened and the excitation duration extended to improve the effective acquisition efficiency per unit time; when the timing initialization parameters indicate poor stability, the excitation interval can be appropriately extended to avoid mutual interference between responses of different spectral excitations.

[0025] Through the above method, the excitation unit ultimately constructs a multispectral excitation sequence with non-uniform temporal distribution characteristics. The "non-uniform temporal distribution characteristics" refer to the fact that the excitation duration and interval for different spectral bands are not entirely the same, but are differentiated based on the temporal initialization parameters. This multispectral excitation sequence not only contains information in the spectral dimension but also implicitly contains information in the temporal structure, thus providing a foundation for the temporal alignment of subsequent multi-source optical responses.

[0026] After the multispectral excitation sequence is constructed, the excitation unit further generates a corresponding multispectral excitation timing identifier for the excitation time window of each spectral band in the multispectral excitation sequence. The "excitation time window" refers to the time interval covered by a certain spectral band from the start to the end of excitation; the "multispectral excitation timing identifier" is an information carrier used to clearly characterize the one-to-one correspondence between a spectral band and its corresponding excitation time window. This timing identifier can include spectral band identification information and time window identification information, enabling the system to accurately identify which spectral band and within which time window a certain optical response was generated in subsequent processing.

[0027] Finally, during the formal detection phase, the excitation unit synchronously applies multi-band optical excitation to the tongue and fingertip detection areas according to the constructed multi-spectral excitation sequence and its corresponding multi-spectral excitation timing identifier. "Synchronous application" here does not require the optical path structure or incident angle of the two detection areas to be completely identical, but rather means that within the same excitation time window, the tongue and fingertip detection areas receive optical excitation in the same spectral band and share the same multi-spectral excitation timing identifier. In this way, the subsequent acquisition and response units can directly perform time alignment based on this timing identifier when acquiring and reconstructing the optical response, thereby ensuring the consistency of the tongue multi-layer spectral response dataset and the fingertip microcirculation dynamic parameter set in the time dimension.

[0028] Through the above complete process, the excitation unit not only completes the application of multi-band optical excitation, but also introduces stability assessment before the detection begins, and uses the results to construct and generate excitation timing, thereby providing a clear, implementable and logically constrained excitation basis for the entire joint detection process based on multispectral fusion.

[0029] The acquisition unit 102 is used to synchronously acquire the reflection response, absorption response and apparent scattering changes of the tongue detection area under different spectral excitations under the constraint of multispectral excitation time sequence identification, and generate a multi-layer spectral response dataset of the tongue based on the spectral-temporal joint analysis method, which characterizes the surface coverage of the tongue coating, the oxygen content of the superficial blood of the tongue, and the water migration characteristics of the tongue tissue.

[0030] In this invention, the acquisition unit 102 is a key unit for accurately acquiring and structurally expressing the optical response generated in the tongue detection area based on the multispectral excitation sequence formed by the excitation unit 101. Its function is not simply image or signal acquisition, but rather to systematically construct the multi-layer physiological response characteristics of the tongue tissue under different spectra and different time windows, so as to ensure that the subsequent modeling and judgment process has a sufficient, reliable and reproducible data foundation.

[0031] Specifically, the acquisition unit 102 must strictly adhere to the multispectral excitation timing identifier generated by the excitation unit 101 during operation. This multispectral excitation timing identifier, as previously defined, is uniquely calibrated on the time axis for each optical excitation band, including the start time, duration, and end time of each band's optical excitation. Under this constraint, the acquisition unit 102 performs synchronous acquisition, meaning that its sampling trigger time corresponds one-to-one with the corresponding spectral excitation time window. This ensures that the acquired optical response signal can be clearly mapped back to the specific spectral excitation conditions, preventing aliasing or time mismatch between different band responses.

[0032] The acquisition unit 102 includes at least an optical imaging or optical detection component, a synchronous sampling control component, and a spectral-temporal joint analysis component. The optical imaging or optical detection component receives the returned light signal from the tongue detection area. This returned light signal is a comprehensive result of various optical effects of the tongue tissue under specific spectral excitation, including reflection, absorption, and apparent scattering of incident light. Here, "reflection response" refers to the change in the intensity of the returned light signal after specular or diffuse reflection of incident light on the tongue coating or tongue body surface; "absorption response" refers to the attenuation characteristics of the returned light intensity caused by the degree of absorption of specific wavelength light energy by different components in the tongue tissue; and "apparent scattering change" refers to the change in scattering characteristics exhibited by light path deflection and energy redistribution caused by factors such as the granular structure of the tongue coating and the inhomogeneity of the tongue tissue microstructure. In actual acquisition, these three types of responses are not forcibly physically separated but are uniformly acquired by the acquisition unit 102 as a comprehensive optical response, and subsequently distinguished and characterized through analysis.

[0033] The synchronous sampling control component is used to control the optical imaging or optical detection component to complete at least one effective sampling within each spectral excitation time window, under the constraint of the multispectral excitation timing identifier. In a preferred embodiment, multiple high-speed samplings can be performed within each spectral excitation time window, and the sampling results are aggregated over time to reduce the impact of instantaneous noise. For example, within a 10-millisecond excitation time window for a certain near-infrared band, 10 frames of response data can be acquired at a sampling period of 1 millisecond, and then these 10 frames of data can be averaged or medianized to obtain a stable response value for that band within that time window. Those skilled in the art can adjust the sampling frequency and aggregation method according to actual hardware capabilities and signal-to-noise ratio requirements; such adjustments do not affect the technical essence of the present invention.

[0034] The spectral-temporal joint analysis component is used to further process synchronously acquired tongue optical response data. Its core lies in incorporating both spectral and temporal dimensions into the analysis framework, rather than analyzing based solely on a single spectrum or a single moment. The "spectral-temporal joint analysis method" refers to considering both the response differences corresponding to different optical excitations at different wavelengths and the response trends over time under the same wavelength excitation. This joint analysis method allows for the differentiation between the rapid response characteristics of the tongue surface and the relatively slow physiological response characteristics of the superficial tissues of the tongue.

[0035] In constructing the analytical results, the tongue multilayer spectral response dataset generated by acquisition unit 102 includes at least three levels of feature representation. The first type of feature characterizes the surface coverage of the tongue coating, reflecting its thickness, density, and surface wettability. This primarily originates from changes in reflection and scattering responses under visible and short-wave ultraviolet light. The second type of feature characterizes changes in oxygen content in the superficial blood of the tongue, reflecting the selective absorption characteristics of hemoglobin in the superficial capillaries of the tongue to different wavelengths of light. This primarily originates from differences in absorption responses under visible and near-infrared light. The third type of feature characterizes the water migration characteristics of the tongue tissue, reflecting changes in water distribution within a short timescale. This primarily originates from changes in response to water-sensitive wavelengths under near-infrared light. The aforementioned "multilayer" is not a strict anatomical stratification but rather a functional hierarchical division of features based on optical response depth and physiological meaning.

[0036] To facilitate subsequent use by the modeling unit 104, the acquisition unit 102 ultimately organizes the above analysis results according to a unified data structure, forming a tongue multi-layer spectral response dataset. Each set of data in this dataset is clearly associated with a corresponding spectral band identifier and time window identifier, enabling subsequent units to directly use this data for cross-organizational feature constraint modeling without needing to re-infer the acquisition conditions.

[0037] Furthermore, the acquisition unit is specifically used for: Under the constraint of multispectral excitation timing identifier, within the excitation time window corresponding to each spectral band, the optical response acquisition of the tongue detection area is triggered, and the original optical response data corresponding to the excitation time window is generated one by one. Based on the original optical response data, the reflection component, absorption component and apparent scattering component of the tongue detection area under the corresponding spectral band are distinguished and processed to form tongue optical response data with spectral component labeling. After obtaining the tongue optical response data with spectral component labels, the tongue optical responses under different spectral bands are time-aligned and recombined according to the multispectral excitation time sequence identifier to generate a tongue spectral time response sequence containing the correlation between spectral and time dimensions. Based on the time response sequence of the tongue's spectral response, the variation characteristics of the tongue's optical response under different spectra and time stages are analyzed hierarchically, thereby generating a multi-layer spectral response dataset of the tongue that can distinguish the response characteristics of the surface layer of the tongue coating from the response characteristics of the superficial tissue of the tongue.

[0038] In practical operation, the acquisition unit is strictly constrained by the multispectral excitation timing identifier. This multispectral excitation timing identifier, generated by the excitation unit during the construction of the multispectral excitation sequence, describes the excitation time window corresponding to each spectral band. It includes at least the spectral band identifier and the start and end time range of the excitation for that band. The acquisition unit does not continuously acquire the tongue's optical response throughout the entire detection cycle; instead, it triggers the acquisition action only within the excitation time window defined by the multispectral excitation timing identifier. Specifically, when the excitation time window for a certain spectral band arrives, the acquisition unit activates the corresponding optical detection component to acquire the optical response generated in the tongue detection area within that time window; when the time window ends, the acquisition action stops. In this way, the acquisition unit ensures that each set of acquired optical response data corresponds one-to-one with a unique spectral band and a unique time window, thereby generating raw optical response data that corresponds one-to-one with the excitation time window.

[0039] The raw optical response data described here refers to the tongue-returned light signal data directly acquired by the optical detector before component differentiation and time reconstruction. This data can be expressed as light intensity values, grayscale values, or equivalent electrical signals. The raw optical response data itself is the result of the superposition of multiple optical effects, including the reflection of incident light by the tongue surface, the absorption of light energy by the tongue tissue, and the scattering effect caused by the microscopic inhomogeneity of the tongue coating structure and tissue. To ensure that subsequent analysis has clear physical meaning, the acquisition unit does not directly use the raw optical response data but processes it further.

[0040] After obtaining the raw optical response data, the acquisition unit distinguishes between the reflection, absorption, and apparent scattering components of the tongue detection area in the corresponding spectral bands. This "distinction" does not require a completely physical separation of the three components, but rather involves analyzing the spatial distribution characteristics, intensity variation characteristics, and response characteristics with spectral variations of the raw optical response data to extract feature information that can characterize the three types of optical behavior. For example, for the reflection component, the focus can be on the high-intensity return signal near the tongue surface; for the absorption component, it can be characterized by comparing the differences in light intensity attenuation in different spectral bands; and for the apparent scattering component, it can be characterized by analyzing the degree of light intensity diffusion or irregular variation trends in local areas. After this distinction, the acquisition unit generates tongue optical response data with clearly defined spectral component labels. Each response data point not only corresponds to a specific spectral band but also clearly indicates whether it primarily reflects reflection, absorption, or apparent scattering characteristics.

[0041] After obtaining the tongue optical response data with spectral component labels, the acquisition unit further performs time alignment and recombination of the tongue optical responses under different spectral bands according to the multispectral excitation time sequence identifier. "Time alignment" here refers to mapping the optical response data acquired under different spectral bands onto a unified time axis, using the excitation time window recorded in the multispectral excitation time sequence identifier as a reference, so that the tongue response under different spectral conditions can correspond to the same time position. "Recombination" refers to rearranging the response data of each spectral band in chronological order to form a data sequence with a continuous time structure. Through the above time alignment and recombination operations, the acquisition unit generates a tongue spectral time response sequence, which contains both spectral and temporal dimension information, reflecting the temporal response changes of the tongue tissue to multispectral excitation.

[0042] After obtaining the tongue's spectral time response sequence, the acquisition unit performs stratified analysis of the changes in the tongue's optical response under different spectra and time stages based on this sequence. This "stratified analysis" does not rely on additional spatial segmentation or imaging depth measurement, but rather on distinguishing different response levels of the tongue tissue based on the different characteristic patterns exhibited by the optical response as it changes with time and spectrum. Generally, the surface layer of the tongue coating responds quickly to optical excitation, with changes concentrated in the initial stage of excitation. In contrast, the superficial tissues of the tongue, due to the involvement of blood oxygenation and tissue water changes, typically exhibit a relatively slow response that gradually appears during the sustained or post-excitation phase. By analyzing the changing trends of various spectral components in the tongue's spectral time response sequence at different time stages, the acquisition unit can distinguish between the response characteristics primarily reflecting the surface layer of the tongue coating and those primarily reflecting the superficial tissues of the tongue.

[0043] Finally, the acquisition unit organizes and processes the results obtained from the hierarchical analysis to generate a multi-layer spectral response dataset of the tongue. This dataset clearly distinguishes between the surface response features of the tongue coating and the response features of the superficial tissues of the tongue, and maintains the correspondence between them and the time sequence identifiers of the multispectral excitation. This allows the subsequent modeling unit to accurately call upon tongue response data with clear physiological meaning and temporal context when performing cross-tissue feature constraint modeling.

[0044] The response unit 103 is used to collect the deep optical interference response and polarization maintenance changes generated in the fingertip detection area under the same multispectral excitation conditions under the constraint of multispectral excitation timing identifier, and extract the fingertip microcirculation dynamic parameter set characterizing the changes in dermal microvascular perfusion, hemodynamic response delay and vascular wall compliance fluctuations through a timing consistency reconstruction method.

[0045] In this invention, the response unit 103 is the core unit that acquires, analyzes and structures the optical response related to microcirculation inside the fingertip detection area under the constraint of the multispectral excitation timing identifier provided by the excitation unit 101. Its purpose is to extract a set of parameters that can truly reflect the functional state and dynamic changes of dermal microvessels from the fingertip, a representative peripheral part, thereby providing reliable input for subsequent cross-tissue feature constraint modeling.

[0046] When the response unit 103 is working, it must first strictly adhere to the time reference defined by the multispectral excitation timing identifier, which, as previously explained, is a unified calibration information for different spectral excitation time windows. "Acquisition under the constraint of the multispectral excitation timing identifier" means that the signal acquisition trigger, duration, and sampling sequence of the response unit 103 for the fingertip detection area all correspond one-to-one with the corresponding spectral excitation time window. This ensures that the optical response of the fingertip tissue and the optical response of the tongue tissue are under completely consistent excitation conditions and time scales, avoiding incomparable physiological differences introduced due to different stimulation conditions.

[0047] The response unit 103 includes at least a deep optical response acquisition component, a polarization state detection component, a synchronous sampling and reconstruction control component, and a microcirculation parameter extraction component. The deep optical response acquisition component is used to acquire the deep optical interference response generated by fingertip tissue under multispectral excitation conditions. The "deep optical interference response" referred to here refers to the phase change and interference characteristic change caused by blood flow within microvessels, vessel wall structure, and differences in refractive index of surrounding tissues after incident light penetrates the fingertip epidermis and enters the dermis. This response, compared to surface reflection signals, can more sensitively reflect the microvascular perfusion state and blood flow dynamics. The deep optical response acquisition component can be implemented using existing mature low-coherence interference detection principles. This invention does not limit the specific optical path structure, as long as it can stably acquire time-varying interference intensity or phase-correlation signals.

[0048] The polarization state detection component is used to detect polarization maintenance changes in the returned light signal. The "polarization maintenance change" referred to here means the change in the degree or direction of polarization of the incident light after passing through the fingertip tissue, due to tissue anisotropy, changes in blood vessel alignment, and alterations in blood flow. Polarization maintenance changes are crucial for distinguishing between static tissue structures and dynamic blood flow effects; therefore, the response unit 103, by introducing polarization-related information, can further enhance its ability to perceive dynamic changes in microcirculation. The polarization state detection component can be implemented by setting up a polarizer, analyzer, or equivalent polarization analysis structure, and its output is recorded synchronously with the deep optical interference response.

[0049] The synchronous sampling and reconstruction control component is used to synchronously sample the deep optical interference response and polarization maintenance changes under the constraint of multispectral excitation timing identifiers, and to reconstruct the temporally consistent data. The "temporally consistent reconstruction method" described here refers to the strict reorganization of the acquired fingertip optical response signals according to the spectral excitation time window during data processing, ensuring that the response sequences under different spectral excitations maintain an alignment relationship on the time axis. For example, within a complete detection cycle, when the visible, near-infrared, and short-wave ultraviolet bands are excited sequentially, the response unit 103 acquires the interference and polarization response sequences within the corresponding time windows, and then rearranges these sequences into a unified time response curve according to the excitation timing identifier, thereby enabling the analysis of the response differences and time delay relationships of the same physiological process under different spectral conditions.

[0050] After completing the temporal consistency reconstruction, the microcirculation parameter extraction component analyzes the reconstructed response signal to generate a dynamic parameter set for fingertip microcirculation. This "dynamic parameter set for fingertip microcirculation" refers to a set of parameters that can quantitatively describe the functional state of microvessels in the dermis of the fingertip and their time-varying characteristics. This parameter set includes at least microvascular perfusion changes, hemodynamic response delay, and vascular wall compliance fluctuations. Microvascular perfusion changes describe the volume fraction of microvessels participating in blood flow per unit time or the equivalent degree of blood filling, which can be obtained by analyzing the amplitude of the change in the intensity of deep optical interference signals over time. Hemodynamic response delay describes the lag time of the fingertip microcirculation in response to optical excitation or physiological pulsation stimulation, which can be determined by comparing the time difference of the peak values ​​of interference signals under different spectral excitation conditions. Vascular wall compliance fluctuations describe the ability of the vascular wall to undergo periodic expansion and contraction under pulse-driven conditions, which can be obtained by analyzing the periodic variation components in the interference signal or polarization signal that are synchronized with the cardiac cycle.

[0051] The extraction of the above parameters does not require complex physiological modeling; rather, calculations can be performed based on signal change trends and temporal relationships. For example, if the intensity of the interference signal exhibits significant periodic fluctuations within a cardiac cycle, the amplitude of these fluctuations can be used as a measure of vascular wall compliance fluctuations. If the peak values ​​of these fluctuations show a stable sequential relationship under different spectral excitation conditions, the time difference can be used as a measure of hemodynamic response delay. Those skilled in the art can perform parameter calculations based on the above description and in conjunction with conventional signal processing methods.

[0052] Finally, the response unit 103 organizes the various microcirculation dynamic parameters obtained from the analysis according to a unified data structure to form a fingertip microcirculation dynamic parameter set, and outputs this parameter set to the modeling unit 104. Since this parameter set is always constrained by the multispectral excitation time sequence identifier during the acquisition and reconstruction process, it is naturally aligned with the tongue multilayer spectral response dataset in terms of time scale and excitation conditions, thus providing a solid, feasible, and physiologically meaningful data foundation for the subsequent construction of cross-tissue feature constraint relationships.

[0053] Modeling unit 104 is used to construct cross-tissue feature constraint relationships between tongue tissue response and fingertip microcirculation response based on the correlation between energy transfer consistency and physiological response time caused by multispectral excitation, under the common input of tongue multi-layer spectral response dataset and fingertip microcirculation dynamic parameter set, and generate joint feature constraint vectors for suppressing non-physiological consistency features under the constraints of cross-tissue feature constraint relationships.

[0054] In this invention, the modeling unit 104 is the core unit that transforms the tongue multilayer spectral response dataset and the fingertip microcirculation dynamic parameter set from "two sets of detection results existing in parallel" into "a joint physiological representation that mutually constrains and verifies each other". Its role is not to directly give a conclusion on physical condition, but to build a constraint framework that conforms to the real physiological laws before the judgment, so as to screen, limit and standardize the effective feature range that can be used in subsequent calculations, thereby avoiding interference from non-physiological consistency information on the judgment result.

[0055] The modeling unit 104 receives the tongue multi-layer spectral response dataset output by the acquisition unit 102 and the fingertip microcirculation dynamic parameter set output by the response unit 103. Since both types of data are obtained under the same multispectral excitation temporal identifier constraint, they naturally have a one-to-one correspondence on the time axis. The modeling unit 104 first utilizes this time alignment basis to jointly read the two types of data, ensuring that for any tongue response data at any given time and under any spectral condition, a corresponding fingertip microcirculation response data can be found, thus forming a cross-tissue, cross-modal foundational data pair.

[0056] Building upon this, modeling unit 104 introduces the concept of "energy transfer consistency." Energy transfer consistency, as described here, refers to the principle that, under the same spectral excitation conditions, the changes in physiological response intensity caused by optical excitation input should follow a fundamentally consistent energy change trend across different tissues. Specifically, when optical excitation in a certain spectral band causes significant absorption enhancement or scattering changes in the tongue tissue, the changes in parameters related to blood perfusion or hemodynamics in the fingertip microcirculation should not exhibit completely opposite or unrelated trends; otherwise, it can be determined that the response of at least one of them mainly originates from noise, posture changes, or incidental interference. Modeling unit 104 establishes constraints at the energy transfer level by comparing the characteristic quantities reflecting energy changes in the tongue's multi-layer spectral response data set with the characteristic quantities reflecting changes in blood flow or vascular state in the fingertip microcirculation dynamic parameters set, to determine whether they have consistent directions of increase or decrease or correlation under the same spectral conditions.

[0057] Meanwhile, modeling unit 104 also introduces the concept of "physiological response time correlation." Physiological response time correlation, as described here, refers to the fact that while the physiological responses of different tissues may differ in absolute time after being subjected to the same optical stimulus, the order and time delay of their responses should be within a reasonable and physiologically explainable range. For example, the response of superficial tongue tissue to optical stimulus typically exhibits rapid changes in optical properties, while the response of the dermal microcirculation in the fingertips may show a stable time lag due to hemodynamic processes. Modeling unit 104 analyzes the time difference between the time points when changes in the tongue's spectral response characteristics occur and when changes in the fingertip microcirculation parameters occur to determine whether this time difference is within a preset physiologically reasonable range. If the time difference is too large or exhibits random fluctuations, the corresponding feature pair can be considered to lack reliable physiological correlation.

[0058] In the specific implementation process, the modeling unit 104 can construct feature pairs by combining each set of features in the multi-layer spectral response dataset of the tongue with the corresponding features in the dynamic parameter set of the fingertip microcirculation, and calculate their consistency index in the direction of energy change and their correlation index in the time response. The consistency index can be understood as a measure of the degree of consistency between the changing trends of the two sets of features; the correlation index can be understood as a measure of the reasonableness of the temporal relationship between the changing trends of the two sets of features. The calculation of the above indicators does not require a complex mathematical model; those skilled in the art can use conventional correlation analysis, time difference statistics, or threshold judgment methods. For example, when the intensity of a certain spectral response of the tongue increases relative to the baseline, and the fingertip microvascular perfusion parameter also increases within the corresponding time window, and the time difference between the peak values ​​of the two changes stably falls within a preset range, then the feature pair can be considered to meet the physiological consistency requirements.

[0059] After completing the above analysis, the modeling unit 104 constructs a cross-tissue feature constraint relationship between the tongue tissue response and the fingertip microcirculation response. The cross-tissue feature constraint relationship mentioned here refers to a set of rules that clearly define which feature combinations can be considered physiologically consistent and which feature combinations should be considered abnormal or unreliable. This set of rules is directly derived from the joint judgment result of the correlation between energy transfer consistency and physiological response time.

[0060] Furthermore, modeling unit 104 generates a joint feature constraint vector under the constraints of the cross-organizational feature constraint relationship. This joint feature constraint vector is a constraint expression used to identify the validity of features. It can be represented by assigning a label to each feature indicating whether it can participate in subsequent calculations, or by assigning constraint coefficients with different weight levels. Features that satisfy the cross-organizational feature constraint relationship are assigned a valid label in the joint feature constraint vector; features that do not satisfy the constraint relationship are suppressed or eliminated, thereby achieving proactive suppression of non-physiologically consistent features.

[0061] In this way, the modeling unit 104 does not directly judge the constitution, but completes a feature selection and constraint modeling based on real physiological laws before the constitution judgment. This ensures that the subsequent generation unit 105 processes the joint mapping calculation based only on features that have been verified for cross-tissue consistency. Thus, the modeling unit 104 plays a crucial role in the entire technical solution, inheriting all the information from the front-end multispectral tongue detection and fingertip microcirculation detection, and improving the scientificity and stability of the overall judgment process through a strict physiological constraint mechanism.

[0062] To help those skilled in the art to more intuitively understand and implement the entire process of generating joint feature constraint vectors for suppressing non-physiological consistency features, the following specific example will be used to further illustrate the process.

[0063] In a complete detection, the excitation unit 101 outputs optical excitations in the visible light band, near-infrared band, and short-wave ultraviolet band sequentially according to a predetermined timing sequence within the same preset detection cycle, with each band corresponding to a specific time window. The acquisition unit 102 acquires multi-layer spectral response datasets of the tongue within these time windows, including, for example, absorption response feature value A1 reflecting changes in oxygen content in the superficial blood of the tongue under the near-infrared band, and scattering response feature value A2 reflecting the surface coverage state of the tongue coating within the same time window. Simultaneously, the response unit 103 acquires a set of dynamic parameters for fingertip microcirculation within the exact same spectral excitation time window, including microvascular perfusion change parameter B1 extracted under the near-infrared band, and periodic change amplitude parameter B2 reflecting fluctuations in vascular wall compliance.

[0064] The modeling unit 104 first pairs A1 with B1 and A2 with B2 based on the multispectral excitation timing identifier, forming cross-tissue feature pairs. Then, it performs an energy transfer consistency judgment on each feature pair. For example, in the near-infrared band, if A1 shows a significant increase relative to its baseline value, it indicates enhanced absorption of light energy in this band by the superficial layer of the tongue. Simultaneously, the modeling unit 104 checks whether B1 also shows a trend of enhanced perfusion within the corresponding time window. If B1 also increases relative to its baseline value, the feature pair is judged to be consistent in the direction of energy change; if B1 changes in the opposite direction or shows no significant change, the feature pair is marked as having inconsistent energy transfer.

[0065] After completing the energy transfer consistency judgment, the modeling unit 104 further performs a physiological response time correlation judgment. Specifically, by comparing the time point when the change in A1 reaches its peak with the time point when the change in B1 reaches its peak, the time difference Δt between the two is calculated. If Δt stably falls within a preset physiologically reasonable range, such as tens to hundreds of milliseconds, and is consistent across multiple detection cycles, then the feature pair is considered to be correlated in terms of time response; if Δt fluctuates significantly or exceeds the physiologically reasonable range, then it is considered to lack reliable time correlation.

[0066] After both judgment steps are completed, the modeling unit 104 marks feature pairs that simultaneously satisfy energy transfer consistency and physiological response time correlation as "physiologically consistent feature pairs"; and marks feature pairs that do not satisfy either condition as "non-physiologically consistent feature pairs". Subsequently, the modeling unit 104 generates a joint feature constraint vector based on the marking results. For example, for A1 in the tongue multilayer spectral response dataset and B1 in the fingertip microcirculation dynamic parameter set, if they are marked as physiologically consistent feature pairs, a valid identifier value is assigned to the corresponding position in the joint feature constraint vector; for A2 and B2, if they are marked as non-physiologically consistent feature pairs, an inhibition identifier value is assigned to the corresponding position in the joint feature constraint vector.

[0067] The resulting joint feature constraint vector, as a set of explicit constraint information, is passed to the generation unit 105 to limit the subsequent joint mapping calculation process to allow only features with valid labels to participate in the physical condition determination operation, while features with suppression labels are not allowed to participate or have their weights significantly reduced.

[0068] Furthermore, the modeling unit is specifically used for: Based on multispectral excitation time sequence identification, data from the tongue multilayer spectral response dataset and the fingertip microcirculation dynamic parameter set that are in the same spectral band and correspond to the same excitation time window are paired to form time-aligned cross-tissue joint response data; Based on cross-organizational joint response data, the response trend characteristics of the tongue multilayer spectral response with the change of excitation energy and the response trend characteristics of the fingertip microcirculation dynamic parameters with the corresponding excitation energy were extracted, and cross-organizational energy consistency judgment results were generated based on the response trend characteristics. Based on cross-tissue joint response data, the temporal relationship and response interval characteristics between changes in tongue spectral response and changes in fingertip microcirculation parameters are analyzed to generate cross-tissue time correlation determination results. Based on the cross-organizational energy consistency determination results and the cross-organizational time correlation determination results, a cross-organizational feature constraint relationship between the tongue tissue response and the fingertip microcirculation response is constructed, and a joint feature constraint vector is generated to suppress non-physiological consistency features.

[0069] The so-called multispectral excitation timing identifier is a set of information generated by the excitation unit when forming a time-distinguishable multispectral excitation sequence. It uniquely indicates which time window in the detection cycle the excitation of a certain spectral band occurs within; in other words, it contains both spectral band information and time window information. The so-called excitation time window refers to the time interval covered by a certain spectral band from the start of excitation to the end of excitation. This time interval has a clearly defined start and end time point within the system and can be precisely located by a unified timing control rule. The so-called cross-tissue joint response data refers to the data structure formed by pairing and time-aligning the response data of the tongue and fingertip within the same spectral band and the same excitation time window. It is not new detection data, but rather the result of recombining previously collected data according to the "same excitation condition." The so-called excitation energy refers to the optical energy intensity level input by the excitation unit to the tissue within a certain spectral band and excitation time window, and its distribution over time. In engineering implementation, it can be characterized by known light source drive current settings, duty cycle settings, light source calibration coefficients, and excitation duration, without requiring additional sensors. The so-called response trend characteristics refer to the characteristics of the response quantity as a function of time or as a function of the excitation energy, such as the direction, magnitude and rate of change. They are used to characterize whether the response changes with the excitation and whether the change in response conforms to the expected physiological laws.

[0070] In the specific implementation process, the modeling unit first performs a pairing process. This pairing process uses the multispectral excitation time sequence identifier as the sole basis to retrieve data segments that are in the same spectral band and correspond to the same excitation time window from the tongue multi-layer spectral response dataset and the fingertip microcirculation dynamic parameter set. The tongue multi-layer spectral response dataset typically contains various optical response information related to the response characteristics of the tongue surface and the response characteristics of the superficial tissues of the tongue, while the fingertip microcirculation dynamic parameter set typically contains time-series parameters related to changes in microvascular perfusion, hemodynamic response delays, and fluctuations in vascular wall compliance. After the modeling unit locks the two into the same spectral band and the same time window according to the time sequence identifier, it combines the tongue response segment and the fingertip parameter segment within that window into a cross-tissue joint response data entry, and repeats this process for all bands and all time windows, thereby forming complete time-aligned cross-tissue joint response data. Time alignment here does not mean that the sampling points on both sides must correspond one-to-one, but rather that they share the same time window boundary and the same excitation condition description. If the sampling frequencies are different, the data can be unified to the same time axis resolution through timestamp mapping. For example, the data within the window can be resampled or interpolated with a fixed time step, so that subsequent trend extraction has a consistent data format.

[0071] After completing the pairing and obtaining cross-tissue joint response data, the modeling unit extracts response trend features and generates cross-tissue energy consistency determination results. The essence of energy consistency determination is to judge whether the tongue response and fingertip response exhibit a "consistent direction and compatible intensity" change relationship under the same spectral excitation. Specifically, the modeling unit first extracts the change trends of the tongue multilayer spectral response and the fingertip microcirculation dynamic parameters within the time window. Taking the tongue multilayer spectral response as an example, if the superficial tissue response features of the tongue show a continuous increase or decrease during the excitation input, and this change is synchronous with the temporal distribution of the excitation energy, then the tongue response can be extracted as a set of response trend features, such as the direction of change, the location of the peak, and the level of change amplitude. Taking the fingertip microcirculation dynamic parameters as an example, if the microvascular perfusion change parameters show a pattern of increasing with increasing excitation and decreasing with decreasing excitation within the same time window, then the corresponding response trend features can also be extracted. The modeling unit compares tongue trend features with fingertip trend features to generate a cross-tissue energy consistency determination result. The output can be a clear classification, such as consistent, inconsistent, or uncertain, or a more granular ranking, such as highly consistent, basically consistent, weakly consistent, or inconsistent. Here's an example: when the tongue trend direction is the same as the fingertip trend direction, and both peak values ​​appear near the high-intensity range of the excitation energy, it is determined to be consistent; when the tongue trend direction is opposite to the fingertip trend direction, or one side has almost no response while the other side responds significantly, it is determined to be inconsistent; when both directions are the same but the peak positions differ greatly, it is temporarily marked as uncertain, leaving the time-related determination result for joint adjudication.

[0072] After generating cross-tissue energy consistency determination results, the modeling unit analyzes the temporal sequence and response interval characteristics based on the same cross-tissue joint response data to generate cross-tissue temporal correlation determination results. The purpose of temporal correlation determination is to determine whether there is a temporal relationship between changes in tongue spectral response and changes in fingertip microcirculation parameters that conforms to physiological conduction laws. The so-called temporal sequence refers to whether the significant onset time of tongue response change or the significant onset time of fingertip parameter change occurs first or last within the same excitation time window; the so-called response interval characteristic refers to the time difference between the significant onset times of the two changes, and the stability of this time difference across multiple excitation windows or multiple detection cycles. The modeling unit can obtain the significant onset time by identifying the "change start point" in the response curve. The change start point can be defined as the first time point when the response value exceeds the baseline fluctuation range after multiple consecutive samplings. The baseline fluctuation range can be determined by the reference optical response before detection or the stable segment at the beginning of the time window. To avoid ambiguity in the definition of "exceeding the range," this invention allows for implementation with a clear engineering threshold, such as using the maximum-minimum difference of the baseline segment as the fluctuation range. If subsequent response values ​​exceed this range for several consecutive sampling points, the change is considered to have begun. After identifying the starting points of changes in the tongue and fingertips, the modeling unit can obtain the temporal sequence and response interval, and generate a cross-tissue temporal correlation determination result accordingly. For example, if the starting point of changes in the fingertip microcirculation parameters always appears after the starting point of changes in the superficial tissue response of the tongue, and the interval between the two remains within a relatively stable range across multiple wavelength bands, then the temporal correlation is determined to be valid; if the sequence is frequently reversed or the interval is extremely large, then the temporal correlation is determined to be invalid; if data noise makes it difficult to stably identify the starting point of changes, then the result is determined to be uncertain.

[0073] Once the cross-organizational energy consistency and cross-organizational temporal correlation results are obtained, the modeling unit constructs cross-organizational feature constraint relationships and generates a joint feature constraint vector. The cross-organizational feature constraint relationships described here are a set of rules used to define which tongue features and which fingertip features can be considered "physiologically consistent and can jointly participate in subsequent joint mapping." These can be expressed as permissible pairings between features, feature admission conditions, or feature inhibition conditions. For example, if a set of superficial tongue tissue response features and a set of fingertip perfusion change features are determined to be consistent in the energy consistency determination and also valid in the temporal correlation determination, the cross-organizational feature constraint relationship will mark this combination as permissible for joint mapping; conversely, if either the energy consistency or temporal correlation determination fails, the combination is marked as disallowed or needs to be inhibited. The joint feature constraint vector is a vectorized representation of the aforementioned constraint relationships. It encodes the admission state of each type of feature or each pair of features into constraint information that can be directly used for subsequent calculations. This allows the generation unit to directly reference this constraint information for feature selection and weight suppression during joint mapping calculations, thereby achieving the goal of suppressing non-physiologically consistent features. Here, "suppression" does not necessarily require the complete deletion of features. In engineering implementation, either elimination or reducing their participation to a very low level can be used. However, regardless of the method used, the joint feature constraint vector must be able to clearly distinguish between the sets of features that are "allowed to participate" and those that are "suppressed," thus ensuring the determinism and repeatability of subsequent calculations.

[0074] The following simplified but complete example illustrates how the "energy consistency judgment result" and the "temporal correlation judgment result" work together to generate the joint feature constraint vector. Assume that the excitation time window corresponding to a certain spectral band is a fixed duration. Within this window, the response characteristics of the superficial tissue of the tongue gradually increase, reaching a maximum in the middle of the window. The fingertip microvascular perfusion change parameter also shows a significant increase and reaches a peak near the middle of the window. In this case, the energy consistency judgment result can be marked as consistent. Simultaneously, the starting point of the tongue response change appears in the early part of the window, while the starting point of the fingertip perfusion change stably appears within a short time interval after the tongue, and this interval does not change significantly across multiple detection cycles. In this case, the temporal correlation judgment result can be marked as valid. Based on this, the modeling unit marks the combination of the tongue feature and the fingertip parameter as allowed to enter subsequent joint mapping and sets the position corresponding to this combination as an allowed marker in the joint feature constraint vector. If, in another band, there is an increase in the tongue response but no significant change in the fingertip parameter, or the starting point of the fingertip change is earlier than that of the tongue and unstable, then this combination is marked as an inhibited marker in the joint feature constraint vector. The generating unit only reads the features corresponding to the allowed labels or assigns them a higher degree of participation during subsequent calculations, thereby achieving systematic suppression of non-physiologically consistent features.

[0075] Furthermore, the modeling unit is also used for: After generating time-aligned cross-organization joint response data, the excitation energy distribution description corresponding to each cross-organization joint response data is determined based on the spectral band information and excitation time window information corresponding to the multispectral excitation time sequence identifier, which is used to characterize the change profile of optical excitation energy within the time window. Based on the description of excitation energy distribution, the changes in the multi-layer spectral response data of the tongue and the dynamic parameters of the microcirculation of the fingertips within the corresponding time window are normalized to generate tongue response change sequences and fingertips response change sequences that eliminate dimension differences. Based on the tongue response change sequence and the fingertip response change sequence, the change direction and change magnitude features of the two in the excitation energy rise phase, excitation energy maintenance phase and excitation energy decay phase are extracted, and the features are combined to form a cross-organization energy response alignment description, which is used to generate cross-organization energy consistency judgment results. After obtaining the cross-tissue energy consistency determination results, based on the tongue response change sequence and the fingertip response change sequence, the significant starting time of the tongue spectral response change and the significant starting time of the fingertip microcirculation parameter change are determined respectively, and the response time interval between the two is calculated accordingly to generate a cross-tissue response time difference description. Based on the distribution of cross-organizational response time difference descriptions across multiple spectral bands and multiple excitation time windows, it is determined whether the response time difference maintains a stable sequence and variation range, thereby generating cross-organizational time correlation determination results. The cross-organizational time correlation determination results and the cross-organizational energy consistency determination results are used together as inputs for constructing cross-organizational feature constraint relationships.

[0076] After completing the time alignment process, the modeling unit has obtained a one-to-one correspondence between the tongue's multi-layer spectral response data and the fingertip microcirculation dynamic parameters within the same spectral band and the same excitation time window. This correspondence constitutes the so-called time-aligned cross-tissue joint response data. The term "time alignment" here does not refer to simple timestamp matching, but rather to the fact that both the tongue and fingertip response data strictly originate from the same optical excitation time window. This eliminates non-physiological differences introduced by inconsistencies in excitation start and end times, ensuring that subsequent analysis only reflects the physiological response characteristics of the tested object itself.

[0077] Based on this, the modeling unit first determines the excitation energy distribution description corresponding to each set of cross-tissue joint response data according to the spectral band information and excitation time window information contained in the multispectral excitation timing identifier. The excitation energy distribution description is a quantitative characterization of the change of optical excitation energy over time within a certain excitation time window. It is not equivalent to a single energy value, but rather reflects the possible rise, maintenance, and decay processes of excitation energy from initiation to termination. For example, within an excitation time window lasting several milliseconds, the light source may experience a process of gradually increasing from low power to the target power, maintaining stable output, and then gradually decreasing power until it is turned off. The energy change profile corresponding to this process constitutes the excitation energy distribution description for that time window. This description can be obtained by analyzing the light source drive signal, current changes, or preset power control curves, with the aim of providing a unified energy reference for the subsequent changes in tongue and fingertip responses.

[0078] After obtaining the description of the excitation energy distribution, the modeling unit normalizes the changes in the tongue multi-layer spectral response data and the fingertip microcirculation dynamic parameters within the corresponding excitation time window. The purpose of normalization is to eliminate the inherent differences in physical dimensions, numerical ranges, and measurement methods between the tongue optical response and the fingertip microcirculation parameters, enabling comparison of the two types of data at the same scale of change. Specifically, for the tongue multi-layer spectral response data, the minimum response value within the time window is used as the starting point, mapping subsequent response changes to relative proportions; similarly, for the fingertip microcirculation dynamic parameters, the baseline state within the corresponding time window is used as a reference, converting changes in blood flow-related parameters into relative amplitudes. Through these processes, tongue response change sequences and fingertip response change sequences are formed, both describing the evolution of the response over time in a dimensionless form.

[0079] After generating the tongue response change sequences and fingertip response change sequences, the modeling unit further divides the entire excitation time window into an excitation energy rise phase, an excitation energy maintenance phase, and an excitation energy decay phase based on the excitation energy distribution description. This division is not a fixed ratio but is automatically determined based on the energy change trend in the excitation energy distribution description. For example, the rise phase corresponds to a continuous increase in excitation energy over time, the maintenance phase corresponds to a stable energy change, and the decay phase corresponds to a decrease in energy. Subsequently, within each of these phases, the modeling unit extracts change direction features and change amplitude features from the tongue and fingertip response change sequences. The change direction features describe whether the response increases or decreases with energy change, while the change amplitude features describe the relative intensity of the response change. Combining the direction and amplitude features extracted from different phases forms a cross-tissue energy response alignment description. This description reflects whether the tongue tissue and fingertip microcirculation exhibit consistent or interpretable response trends when facing the same energy change process, thereby generating a cross-tissue energy consistency determination result. If the two types of responses show a matching direction and reasonable amplitude relationship in the energy rise, maintenance and decay phases, then their energy consistency is considered to be high; otherwise, it is considered that there is non-physiological consistency at the energy level.

[0080] After obtaining the cross-tissue energy consistency determination results, the modeling unit further determines the significant onset time of each type of response based on the tongue response change sequence and the fingertip response change sequence. The significant onset time refers to the moment when the response change first exceeds a preset change threshold and remains there for a certain period of time; it is used to characterize the starting point of a recognizable physiological response of the tissue to the stimulus. For example, when the relative change amplitude of the tongue spectral response is higher than the baseline noise range for multiple consecutive sampling points, the corresponding time point can be identified as the significant onset time of the tongue spectral response change; similarly, the significant onset time of the fingertip microcirculation parameter change can be determined. By comparing the two, the time difference between the tongue response and the fingertip response, or the time difference between the fingertip response and the tongue response, is calculated; this time difference constitutes the cross-tissue response time difference description.

[0081] Subsequently, the modeling unit summarized and analyzed the aforementioned cross-tissue response time difference descriptions across multiple spectral bands and excitation time windows, observing their distribution. If the response time differences consistently maintain a stable sequence across different spectral bands and excitation time windows—for example, the tongue response always precedes the fingertip response—and the time difference always falls within a relatively fixed range, then a stable physiological temporal correlation can be determined between the two. Conversely, if the response sequence frequently reverses or the time difference fluctuates abnormally, it indicates a lack of reliable physiological temporal correlation between the two types of responses. Based on the above analysis, the modeling unit generates a cross-tissue temporal correlation determination result.

[0082] Ultimately, the cross-organizational temporal correlation determination result and the aforementioned cross-organizational energy consistency determination result are used together as inputs to construct cross-organizational feature constraint relationships. In the subsequent joint feature constraint vector generation process, features that do not conform to energy consistency or temporal correlation are suppressed or eliminated, thereby ensuring that the feature set entering the physical fitness determination process meets the physiological logic consistency requirements.

[0083] The generation unit 105 is used to perform joint mapping calculations on the tongue multilayer spectral response dataset and the fingertip microcirculation dynamic parameter set under the constraint of the joint feature constraint vector, so as to obtain the constitution determination result including at least the constitution type identifier and constitution stability assessment index.

[0084] In this invention, the generation unit 105 is a functional unit that performs the final joint calculation of multi-source features after strict physiological consistency constraints, based on the completion of the aforementioned excitation, acquisition, response and modeling processes. Its role is to further transform the tongue multi-layer spectral response dataset and the fingertip microcirculation dynamic parameter set from "objective description at the level of physical and physiological signals" into "judgment results with clear medical and physical significance", and to ensure that the judgment results have good stability and interpretability in repeated tests and among different individuals.

[0085] When the generation unit 105 is working, it first receives the joint feature constraint vector output by the modeling unit 104. This joint feature constraint vector, as defined above, is a set of constraint information used to identify whether each tongue feature and fingertip feature meets the cross-tissue physiological consistency requirements; essentially, it is a mechanism for limiting the eligibility of features to participate. Before performing any calculations, the generation unit 105 must use this joint feature constraint vector as a mandatory constraint condition, meaning that only features marked as physiologically consistent are allowed to enter the subsequent joint mapping calculation process, while features marked as non-physiologically consistent are not involved or are significantly weakened, thereby preventing them from affecting the final physical condition assessment result.

[0086] In the specific implementation process, the generation unit 105 first jointly reads the tongue multi-layer spectral response dataset and the fingertip microcirculation dynamic parameter set, and then filters the original feature set according to the joint feature constraint vector to form a set of constrained effective features. The "joint mapping calculation" mentioned here refers to the process of uniformly representing and comprehensively calculating features from different detection sites and different physical mechanisms within the same computational framework. Its purpose is not simple superposition, but rather to transform multi-source features into an intermediate representation space that can be used for constitution assessment through mapping relationships. This intermediate representation space can be understood as an abstract constitution-related feature space, where each dimension corresponds to a certain type of physiologically significant comprehensive indicator.

[0087] In the joint mapping calculation, the generation unit 105 can process the features using rule-driven, model-driven, or a combination of both methods. This invention does not limit the use of a specific algorithm, but requires that the mapping process be strictly based on a set of valid features after constraint filtering, and that the calculation process maintains determinism and repeatability. For example, features reflecting the superficial blood state of the tongue in a specific band of the tongue's multi-layer spectral response can be jointly mapped with features reflecting microvascular perfusion levels in the dynamic parameters of fingertip microcirculation to generate a comprehensive index characterizing overall blood circulation and tissue oxygenation. Similarly, features related to the surface coverage of the tongue can be jointly mapped with vascular wall compliance fluctuations to generate a comprehensive index reflecting the body's metabolism and tissue elasticity. Those skilled in the art can design the mapping relationship reasonably according to actual application needs without departing from the technical essence of this invention.

[0088] After completing the joint mapping calculation, the generation unit 105 outputs the constitution determination result based on the mapping result. The constitution determination result mentioned here includes at least a constitution type identifier and a constitution stability assessment index. The constitution type identifier indicates the constitution category of the tested object under a preset constitution classification system. This classification system can be established based on traditional constitution classification methods or modern medical constitution assessment systems. This invention does not limit specific classification standards, but requires that the classification rules be preset in the system and remain consistent. The generation process of the constitution type identifier is based on the matching relationship between the comprehensive index obtained from the joint mapping calculation and the preset constitution determination rules. For example, when several key comprehensive indicators simultaneously fall within the threshold range corresponding to a certain constitution type, the corresponding constitution type identifier can be output.

[0089] The physical fitness stability assessment index is used to reflect the reliability of the physical fitness assessment result in both the time and feature dimensions. This index can be obtained by analyzing the number of valid features involved in the assessment, the consistency between features, and the sensitivity of the joint mapping result to small changes in the input features. For example, when the number of valid features in the joint feature constraint vector is large, and these features show a consistent trend in multiple tests, the generation unit 105 can output a high physical fitness stability assessment index; conversely, when the number of valid features is small or the consistency between features is weak, a low index is output. This index does not directly change the physical fitness type identifier, but it serves as important supplementary information to the reliability of the assessment result.

[0090] Through the above processing method, the generation unit 105 does not simply classify the original detection data, but performs ordered and interpretable joint mapping calculations on the multi-source physiological information of the tongue and fingertips under the strict constraints of the joint feature constraint vector, thereby outputting a constitution determination result that has both clear physical significance and stability and repeatability.

[0091] The following is a detailed example. This example takes a complete detection as the unit, assuming that the tongue multilayer spectral response dataset already contains several feature entries related to the surface coverage of the tongue coating, changes in oxygen content of the superficial blood of the tongue, and water migration characteristics of the tongue tissue; the fingertip microcirculation dynamic parameter set already contains several feature entries related to changes in microvascular perfusion, hemodynamic response delay, and fluctuations in vascular wall compliance; and the joint feature constraint vector already gives each feature a clear label of "allowed to participate in joint mapping" or "inhibited from participating in joint mapping".

[0092] In this example, the generation unit first reads the joint feature constraint vector and uses it to filter the two sets of input data. Assume the tongue multilayer spectral response dataset originally contained nine feature entries: three related to the surface coverage of the tongue coating, three related to changes in oxygen content in the superficial blood of the tongue, and three related to the water migration characteristics of the tongue tissue. The fingertip microcirculation dynamic parameter set originally contained six feature entries: two related to changes in microvascular perfusion, two related to hemodynamic response delay, and two related to fluctuations in vascular wall compliance. After the joint feature constraint vector labels each of the above fifteen features, the generation unit finds that ten are marked as "allowed to participate in joint mapping," and the remaining five are marked as "suppressed to participate in joint mapping." The first thing the generation unit does is to directly discard or freeze the five suppressed features, ensuring they are no longer read in any subsequent calculations, thus guaranteeing that the input for joint mapping comes only from valid features whose physiological consistency has been verified.

[0093] Next, the generation unit groups the remaining ten valid features according to a pre-defined joint mapping rule. This joint mapping rule is a set of correspondences fixed at the time of device manufacturing or algorithm deployment, specifying which tongue features and which fingertip features should be used together to characterize the same type of constitution-related state. To make the process clearer, this example divides the joint mapping into three "constitution-related intermediate representations," used to describe circulatory oxygenation-related states, body fluid and hydration-related states, and vascular elasticity and stress-related states, respectively. Note that these "intermediate representations" are only internal quantities in the calculation process and are not directly output as constitution types, but rather serve as the basis for judgment.

[0094] Taking the state related to circulatory oxygenation as an example, the generation unit selects two tongue features most relevant to changes in superficial blood oxygenation from the effective features, one fingertip feature most relevant to changes in microvascular perfusion, and one hemodynamic response delay feature as the basis for dynamic correction. In specific calculations, the generation unit first performs uniform scaling on these four features to eliminate the unfairness caused by differences in dimensions. Uniform scaling can be explicitly implemented without a formula; the method is to perform "interval mapping" on each feature using its statistical range in the normal population within the device's built-in reference database. For example, for a tongue oxygenation-related feature, the device database records that it typically falls within a certain interval when within the normal range. The generation unit compares the current detection value with this interval; if it falls in the middle of the interval, it gives a "neutral contribution"; if it is above the upper bound, it gives a "positive contribution"; if it is below the lower bound, it gives a "negative contribution," and further assigns a contribution intensity level based on the deviation magnitude. The same processing is applied to fingertip perfusion change characteristics; for hemodynamic response delay characteristics, "excessive delay, normal delay, and insufficient delay" are used as classifications, with "excessive delay" considered a weakening factor on the reliability of oxygen supply status. After completing the unified scaling process, the generation unit merges the contributions of the two tongue oxygenation-related characteristics to obtain a tongue oxygen supply contribution value; it also obtains a fingertip oxygen supply contribution value from the fingertip perfusion change characteristics; then, it dynamically adjusts the two contribution values ​​according to the delay classification. For example, when the delay is in the normal range, the weights of the two are kept balanced; when the delay is excessive, the weight of the fingertip contribution in this mapping is reduced and the uncertainty label is increased. Finally, the generation unit obtains an intermediate representation result of the cyclic oxygen supply-related status, and simultaneously obtains an internal label of "confidence of this intermediate representation".

[0095] Taking the state related to body fluids and moisture as an example, the generation unit selects two tongue features most relevant to the water migration characteristics of tongue tissue from the effective features, and then selects one fingertip feature related to the fluctuation of vascular wall compliance as lateral evidence. This is because water migration often produces corresponding signs in the tissue elasticity of the peripheral microcirculation. The generation unit also first performs uniform scaling on the three features, then merges the two tongue water-related features to obtain the tongue moisture contribution value, and then uses the fingertip compliance fluctuation feature to check for any obvious contradictions. A contradiction refers to a tongue moisture contribution value showing a significant bias towards dryness or wetness, but the fingertip compliance fluctuation falls at an extreme level completely inconsistent with this. If there is no contradiction, the generation unit directly uses the tongue moisture contribution value as the main output of this intermediate representation, and uses the fingertip feature as a consistency bonus; if there is a contradiction, the output direction is not changed, but the credibility of the intermediate representation is reduced, and this will be reflected as a source of deduction for "insufficient cross-tissue consistency" in subsequent physical stability assessments.

[0096] Taking the relationship between vascular elasticity and stress as an example, the generation unit selects two fingertip features related to vascular wall compliance fluctuations as primary features from the effective features, and one tongue feature related to the tongue coating coverage as an auxiliary feature. The logic here is that vascular elasticity and stress status often have a stable associated relationship with certain tongue coating coverage characteristics, but since the tongue coating is a surface feature, it is only used as an auxiliary feature, not the primary one. The generation unit still performs uniform scaling, then first merges the two fingertip elasticity-related features to obtain the fingertip elasticity contribution value, and then uses the tongue coating coverage feature for consistency confirmation. If the consistency is good, the credibility of this intermediate representation is increased; if the consistency is average, the intermediate representation value is maintained but the credibility is decreased.

[0097] Once all three intermediate physical characteristics have been calculated, the generation unit enters the physical type identification stage. A pre-defined physical condition identification rule base is used here. This rule base stores the identification logic for each physical type in a "condition combination" manner, with each logic described by a combination of intermediate characteristics. For example, one physical type might require a low level of circulatory oxygenation and a dry level of body fluid and moisture, while also having weak vascular elasticity and stress-related states, and all three must have a certain minimum confidence level. Another physical type might require a high level of circulatory oxygenation but large fluctuations in vascular elasticity and stress-related states, while also having a wet level of body fluid and moisture. The generation unit matches each of the three intermediate characteristics against the rule base. If a physical type satisfies the highest degree and exceeds the preset minimum matching threshold, that physical type is output as the physical type identification. If two physical types have similar matching degrees, the one with stronger consistency with the circulatory oxygenation state is prioritized, or a combination identification of "main type + tendency type" is output, provided that this combination output is allowed in the device's rule base.

[0098] After obtaining the body type identifier, the generation unit calculates the body stability assessment index. This index can be calculated without relying on a formula and is determined by three types of factors. The first factor is effective feature coverage, which is the proportion of allowed features in the joint feature constraint vector to the total number of features. A higher proportion indicates more features passing cross-organizational consistency, resulting in higher stability. The second factor is cross-organizational consistency strength, which is the number of times consistency bonus items and contradiction deduction items appear during the calculation of the three intermediate representations. More contradiction deductions reduce stability. The third factor is dynamic credibility summary, which is the combined result of the credibility labels of the three intermediate representations. If any one of these credibility values ​​is below the minimum standard, the overall stability cannot be rated as high. The generation unit converts these three factors into grade scores. For example, coverage can be divided into high, medium, and low levels; consistency strength into strong, average, and weak levels; and credibility summary into reliable, requiring retesting, and unreliable levels. These three are then combined to obtain the final body stability assessment index. For example, if the coverage of effective features is high, the cross-organizational consistency is strong, and the credibility of the three intermediate representations is reliable, the output of the physical stability assessment index is high; if the coverage of effective features is moderate and there is a significant contradiction in the score, but the minimum credibility standard is still met, the output is medium and a retest is recommended; if the coverage of effective features is low and the credibility of at least one intermediate representation is unreliable, the output is low and the result is indicated as being for reference only.

[0099] Furthermore, the generation unit is specifically used for: Under the constraint of the joint feature constraint vector, tongue features and fingertip features marked as allowed to participate in joint mapping are selected from the tongue multi-layer spectral response dataset and the fingertip microcirculation dynamic parameter set, forming an effective set of joint features that satisfies the cross-tissue feature constraint relationship; Based on an effective set of joint features, tongue features and fingertip features are combined in accordance with the spectral dimension, time dimension and tissue origin dimension to generate joint feature representation results that reflect the relationship between the state of tongue tissue and the state of fingertip microcirculation. Based on the joint feature representation results, the consistency of tongue features and fingertip features under multiple excitation time windows is comprehensively evaluated to generate an intermediate physical condition assessment representation to describe the overall physical condition of the test subject. Based on the intermediate physical fitness assessment, a physical fitness type identifier is output, and combined with the changes of the intermediate physical fitness assessment under different stimulus time windows, a physical fitness stability assessment index is generated to reflect the stability of the physical fitness assessment results.

[0100] In practice, the generation unit first filters the tongue multi-layer spectral response dataset and the fingertip microcirculation dynamic parameter set under the constraint of the joint feature constraint vector. The joint feature constraint vector, generated by the modeling unit based on cross-tissue energy consistency and cross-tissue temporal correlation determination results, is essentially a clear marker indicating whether various tongue and fingertip features are allowed to participate in subsequent joint mapping calculations. After reading this constraint vector, the generation unit selects only the tongue features marked as allowed from the tongue multi-layer spectral response dataset and only the fingertip features marked as allowed from the fingertip microcirculation dynamic parameter set, merging them to form a valid joint feature set. Each feature or feature pair in this valid joint feature set has passed the prior cross-tissue consistency test, possesses clear physiological rationality and temporal correspondence, and thus constitutes the sole input source for subsequent joint mapping calculations.

[0101] After obtaining a valid set of joint features, the generation unit performs a structured combination of tongue and fingertip features based on this set. This combination is not a simple concatenation, but rather a correspondence and organization of features according to spectral, temporal, and tissue origin dimensions. The spectral dimension distinguishes features originating from different spectral bands, such as visible light, near-infrared, or other bands; the temporal dimension distinguishes the excitation time window or different detection stages corresponding to the features; and the tissue origin dimension clearly distinguishes between tongue tissue features and fingertip microcirculation features. Through the corresponding combination in these three dimensions, the generation unit constructs a joint feature representation result that clearly reflects the correlation between the tongue tissue state and the fingertip microcirculation state under the same spectral conditions and temporal background. The joint feature representation result does not require a fixed data structure; it can be a vector, matrix, or other structured representation method, as long as it clearly expresses the "correspondence between a certain tongue feature and a certain fingertip feature under a certain time window." Those skilled in the art can choose according to the system implementation requirements.

[0102] After generating the joint feature representation, the generation unit comprehensively evaluates the consistency of tongue and fingertip features across multiple excitation time windows, thus forming an intermediate representation for physical fitness assessment. "Consistency" here refers to whether the state changes of tongue and fingertip features exhibit a relatively stable and repeatable pattern across different spectral bands and excitation time windows. The generation unit can determine the level of consistency by comparing the direction, magnitude, and relative order of change of the same type of joint feature across multiple time windows. For example, if the superficial tissue response features of the tongue and the microcirculation perfusion features of the fingertips consistently show a synchronous increasing or decreasing trend across multiple excitation time windows, it indicates high consistency of the joint feature over time; if this trend frequently reverses or exhibits significant randomness across different time windows, it indicates low consistency. The generation unit summarizes the consistency assessment results of various joint features to form an intermediate representation for physical fitness assessment, describing the overall physical fitness status of the tested subject. This intermediate representation can be understood as a comprehensive state description. It does not directly give the name of the constitution category, but reflects the overall state characteristics of multi-source features in the current detection period in a structured way.

[0103] After obtaining the intermediate representation for constitution assessment, the generation unit outputs a constitution type identifier based on this intermediate representation. The constitution type identifier described here is a categorical description of the current constitution state of the tested subject, and its classification is based on the results of the joint feature representation and its consistency assessment. The specific classification method for constitution types can be completed according to pre-set judgment rules, such as determining the final constitution type based on the proportion of various joint features in the intermediate representation and the direction of the dominant feature. This invention does not limit the specific constitution classification system, but only requires that the generation process of the constitution type identifier has clear rules and can be repeatedly executed. Those skilled in the art can choose an appropriate constitution classification standard based on actual application needs.

[0104] While outputting the constitution type identifier, the generation unit also combines the changes in intermediate representations of constitution assessment under different stimulus time windows to generate a constitution stability assessment index. This constitution stability assessment index reflects the reliability of the constitution assessment results over time; it focuses not on the constitution type itself, but on the consistency of the constitution type across multiple time windows or testing cycles. The generation unit obtains this index by comparing the similarity of intermediate representations of constitution assessment under different stimulus time windows. For example, when the intermediate representation structures generated under multiple stimulus time windows are highly similar, and the final output constitution type identifier remains consistent, the constitution stability assessment index is judged to be high; when the intermediate representations under different time windows differ significantly, leading to frequent changes in the constitution type identifier, the constitution stability assessment index is judged to be low. This stability assessment index can be expressed in rank, numerical, or interval form, as long as it clearly distinguishes between "stable" and "unstable" constitution assessment results.

[0105] Although this application discloses preferred embodiments as described above, it is not intended to limit this application. Any person skilled in the art can make possible changes and modifications without departing from the spirit and scope of this application. Therefore, the scope of protection of this application should be determined by the scope defined in the claims of this application.

Claims

1. A device for joint detection of tongue coating and fingertip microcirculation based on multispectral fusion, characterized in that, include: The excitation unit is used to simultaneously apply multi-band optical excitation to the tongue detection area and the fingertip detection area according to a unified timing control rule within a preset detection cycle, forming a multispectral excitation sequence with time distinguishability. The acquisition unit is used to synchronously acquire the reflection response, absorption response and apparent scattering changes of the tongue detection area under different spectral excitations under the constraint of multispectral excitation time sequence identification, and generate a multi-layer spectral response dataset of the tongue based on the joint spectral-temporal analysis method. The response unit is used to collect the deep optical interference response and polarization maintenance changes generated in the fingertip detection area under the same multispectral excitation conditions under the constraint of multispectral excitation timing identifier, and extract the dynamic parameter set of fingertip microcirculation through the timing consistency reconstruction method. The modeling unit is used to construct cross-tissue feature constraint relationships between tongue tissue response and fingertip microcirculation response based on the correlation between energy transfer consistency and physiological response time caused by multispectral excitation, under the common input of tongue multi-layer spectral response dataset and fingertip microcirculation dynamic parameter set. Under the constraint of cross-tissue feature constraint relationships, a joint feature constraint vector is generated to suppress non-physiological consistency features. The generation unit is used to perform joint mapping calculations on the tongue multilayer spectral response dataset and the fingertip microcirculation dynamic parameter set under the constraint of the joint feature constraint vector, so as to obtain the constitution determination result including at least the constitution type identifier and constitution stability assessment index.

2. The multispectral fusion-based joint detection device for tongue coating and fingertip microcirculation according to claim 1, characterized in that, The excitation unit is specifically used for: Before the start of the preset detection cycle, at least one reference optical excitation is applied to the tongue detection area and the fingertip detection area to obtain the corresponding initial optical response, and a timing initialization parameter characterizing the current detection stable state of the object under test is generated based on the initial optical response. Based on the timing initialization parameters, the excitation duration of each spectral band within the preset detection period and the excitation interval between adjacent spectral bands are determined, thereby constructing a multispectral excitation sequence with non-equidistant temporal distribution characteristics. After the multispectral excitation sequence is constructed, a corresponding multispectral excitation timing identifier is generated for the excitation time window of each spectral band in the multispectral excitation sequence. The multispectral excitation timing identifier is used to characterize the one-to-one correspondence between the spectral band and the excitation time window. Multi-band optical excitation is applied synchronously to the tongue detection area and the fingertip detection area according to the multi-spectral excitation sequence and its corresponding multi-spectral excitation timing identifier.

3. The multispectral fusion-based joint detection device for tongue coating and fingertip microcirculation according to claim 1, characterized in that, The acquisition unit is specifically used for: Under the constraint of multispectral excitation timing identifier, within the excitation time window corresponding to each spectral band, the optical response acquisition of the tongue detection area is triggered, and the original optical response data corresponding to the excitation time window is generated one by one. Based on the original optical response data, the reflection component, absorption component and apparent scattering component of the tongue detection area under the corresponding spectral band are distinguished and processed to form tongue optical response data with spectral component labeling. After obtaining the tongue optical response data with spectral component labels, the tongue optical responses under different spectral bands are time-aligned and recombined according to the multispectral excitation time sequence identifier to generate a tongue spectral time response sequence containing the correlation between spectral and time dimensions. Based on the time response sequence of the tongue's spectral response, the variation characteristics of the tongue's optical response under different spectra and time stages are analyzed hierarchically, thereby generating a multi-layer spectral response dataset of the tongue that can distinguish the response characteristics of the surface layer of the tongue coating from the response characteristics of the superficial tissue of the tongue.

4. The multispectral fusion-based joint detection device for tongue coating and fingertip microcirculation according to claim 1, characterized in that, The modeling unit is specifically used for: Based on multispectral excitation time sequence identification, data from the tongue multilayer spectral response dataset and the fingertip microcirculation dynamic parameter set that are in the same spectral band and correspond to the same excitation time window are paired to form time-aligned cross-tissue joint response data; Based on cross-organizational joint response data, the response trend characteristics of the tongue multilayer spectral response with the change of excitation energy and the response trend characteristics of the fingertip microcirculation dynamic parameters with the corresponding excitation energy were extracted, and cross-organizational energy consistency judgment results were generated based on the response trend characteristics. Based on cross-tissue joint response data, the temporal relationship and response interval characteristics between changes in tongue spectral response and changes in fingertip microcirculation parameters are analyzed to generate cross-tissue time correlation determination results. Based on the cross-organizational energy consistency determination results and the cross-organizational time correlation determination results, a cross-organizational feature constraint relationship between the tongue tissue response and the fingertip microcirculation response is constructed, and a joint feature constraint vector is generated to suppress non-physiological consistency features.

5. The multispectral fusion-based joint detection device for tongue coating and fingertip microcirculation according to claim 1, characterized in that, The generation unit is specifically used for: Under the constraint of the joint feature constraint vector, tongue features and fingertip features marked as allowed to participate in joint mapping are selected from the tongue multi-layer spectral response dataset and the fingertip microcirculation dynamic parameter set, forming an effective set of joint features that satisfies the cross-tissue feature constraint relationship; Based on an effective set of joint features, tongue features and fingertip features are combined in accordance with the spectral dimension, time dimension and tissue origin dimension to generate joint feature representation results that reflect the relationship between the state of tongue tissue and the state of fingertip microcirculation. Based on the joint feature representation results, the consistency of tongue features and fingertip features under multiple excitation time windows is comprehensively evaluated to generate an intermediate physical condition assessment representation to describe the current overall physical condition of the test subject. Based on the intermediate physical fitness assessment, a physical fitness type identifier is output, and combined with the changes of the intermediate physical fitness assessment under different stimulus time windows, a physical fitness stability assessment index is generated to reflect the stability of the physical fitness assessment results.

6. The multispectral fusion-based joint detection device for tongue coating and fingertip microcirculation according to claim 4, characterized in that, The modeling unit is also used for: After generating time-aligned cross-organization joint response data, the excitation energy distribution description corresponding to each cross-organization joint response data is determined based on the spectral band information and excitation time window information corresponding to the multispectral excitation time sequence identifier, which is used to characterize the change profile of optical excitation energy within the time window. Based on the description of excitation energy distribution, the changes in the multi-layer spectral response data of the tongue and the dynamic parameters of the microcirculation of the fingertips within the corresponding time window are normalized to generate tongue response change sequences and fingertips response change sequences that eliminate dimension differences. Based on the tongue response change sequence and the fingertip response change sequence, the change direction and change magnitude features of the two in the excitation energy rise phase, excitation energy maintenance phase and excitation energy decay phase are extracted, and the features are combined to form a cross-organization energy response alignment description, which is used to generate cross-organization energy consistency judgment results. After obtaining the cross-tissue energy consistency determination results, based on the tongue response change sequence and the fingertip response change sequence, the significant starting time of the tongue spectral response change and the significant starting time of the fingertip microcirculation parameter change are determined respectively, and the response time interval between the two is calculated accordingly to generate a cross-tissue response time difference description. Based on the distribution of cross-organizational response time difference descriptions across multiple spectral bands and multiple excitation time windows, it is determined whether the response time difference maintains a stable sequence and variation range, thereby generating cross-organizational time correlation determination results. The cross-organizational time correlation determination results and the cross-organizational energy consistency determination results are used together as inputs for constructing cross-organizational feature constraint relationships.

Images