Ear acupoint biological signal synchronous acquisition method and system based on multi-modal sensor fusion

Abstract

The application provides an auricular point biological signal synchronous acquisition method and system based on multi-modal sensing fusion, and belongs to the field of intelligent sensing, and comprises the following steps: loading multi-modal contact fingerprint data and combining the multi-modal contact fingerprint data into a multi-modal contact fingerprint feature vector; performing initial wearing offset detection based on a standard anatomical coordinate reference model; repeatedly collecting dynamic contact fingerprint data at a fixed period, and performing dynamic drift real-time tracking; calculating the module length of a real-time drift state vector, and when the module length is greater than or equal to a preset drift threshold, triggering the re-execution of the initial wearing offset detection; driving the coordinate real-time update of a digital twin ear model; and mapping the multi-dimensional feature data collected by each sensing unit to a standard anatomical acupoint coordinate to generate an auricular point health data set. The application solves the problem that the correspondence between sensing units and acupoints is inaccurate due to the wearing offset and sliding of an ear cap in the existing auricular point acquisition system, so that stable, spatially alignable multi-modal auricular point health data is difficult to obtain.

Description

Technical Field

[0001] This invention relates to the field of intelligent sensing, specifically to a method and system for synchronous acquisition of auricular biosignals based on multimodal sensor fusion. Background Technology

[0002] In the synchronous acquisition method of auricular biosignals based on multimodal sensor fusion, the poor repeatability of the wearing position of flexible earmuffs and the dynamic drift problem during use have become key technical bottlenecks restricting the accuracy of long-term dynamic health monitoring. Existing technologies such as CN120997127A, while achieving static projection positioning of auricular acupoints using 3D point cloud models and temperature characteristics, rely on external optical and thermal imaging equipment and cannot solve the problem of real-time earmuff slippage caused by users' daily activities in wearable scenarios. Furthermore, traditional methods lack automatic detection and compensation mechanisms for wearing offsets, making it difficult to compare auricular acupoint data collected at different times longitudinally within the same anatomical coordinate system, thus limiting its application in quantitative assessment of therapeutic effects and monitoring of chronic disease trends.

[0003] Therefore, there is an urgent need for a method that can track and dynamically correct earpiece position drift in real time. Summary of the Invention

[0004] This invention provides a method and system for synchronous acquisition of auricular biosignals based on multimodal sensor fusion, which solves the problem that existing auricular acupoint acquisition systems suffer from inaccurate correspondence between sensing units and acupoints due to ear cover misalignment and slippage, making it difficult to obtain stable and spatially aligned multimodal auricular acupoint health data.

[0005] In view of the above problems, the present invention provides a method and system for synchronous acquisition of auricular biosignals based on multimodal sensor fusion.

[0006] In a first aspect, the present invention provides a method for synchronous acquisition of auricular biosignals based on multimodal sensor fusion, comprising:

[0007] Multimodal contact fingerprint data synchronously collected by all sensing units on the flexible multimodal sensing earmuff is loaded, and the multimodal contact fingerprint data of each sensing unit is combined into a multimodal contact fingerprint feature vector.

[0008] Based on the multimodal contact fingerprint feature vector and the standard anatomical coordinate reference model, initial wearing offset detection is performed to obtain the overall rigid offset.

[0009] During the process of collecting auricular biosignals, dynamic contact fingerprint data is repeatedly collected at a fixed period. Based on the time series of the dynamic contact fingerprint data, dynamic drift real-time tracking is performed to obtain the real-time drift state vector.

[0010] Calculate the magnitude of the real-time drift state vector. When the magnitude is greater than or equal to a preset drift threshold, trigger the re-execution of the initial wearing offset detection to update the overall rigidity offset.

[0011] Based on the overall rigidity offset and the real-time drift state vector, the coordinates of the digital twin ear mold are updated in real time to obtain the standard anatomical acupoint coordinates corresponding to the measurement values ​​of each sensing unit.

[0012] The multidimensional feature data collected by each sensing unit is mapped to the coordinates of the standard anatomical acupoints to generate a dynamically registered auricular acupoint health dataset.

[0013] Secondly, the present invention provides a synchronous acquisition system for auricular biosignals based on multimodal sensor fusion, comprising:

[0014] The multimodal initial fingerprint acquisition and feature construction module is used to load the multimodal contact fingerprint data synchronously acquired by all sensing units on the flexible multimodal sensing ear cover, and combine the multimodal contact fingerprint data of each sensing unit into a multimodal contact fingerprint feature vector.

[0015] The initial wearing offset detection module is used to perform initial wearing offset detection based on the multimodal contact fingerprint feature vector and the standard anatomical coordinate reference model to obtain the overall rigid offset.

[0016] The dynamic fingerprint periodic acquisition and drift tracking module is used to repeatedly acquire dynamic contact fingerprint data at a fixed period during the acquisition of auricular biosignals, and to perform real-time dynamic drift tracking based on the time series of the dynamic contact fingerprint data to obtain a real-time drift state vector.

[0017] The drift state assessment and recalibration trigger module is used to calculate the magnitude of the real-time drift state vector. When the magnitude is greater than or equal to a preset drift threshold, it triggers the re-execution of the initial wearing offset detection to update the overall rigid offset.

[0018] The digital twin ear mold coordinate dynamic mapping module is used to drive the coordinates of the digital twin ear mold to be updated in real time according to the overall rigidity offset and the real-time drift state vector, so as to obtain the standard anatomical acupoint coordinates corresponding to the measurement values ​​of each sensing unit.

[0019] The multimodal dataset dynamic registration and generation module is used to map the multidimensional feature data collected by each sensing unit to the coordinates of the standard anatomical acupoints, and generate a dynamically registered auricular acupoint health dataset.

[0020] One or more technical solutions provided in this invention have at least the following technical effects or advantages:

[0021] The technical solution of this invention first controls the earcups to enter scanning mode after wearing, simultaneously collecting the skin contact impedance amplitude, local skin temperature value, and static contact pressure value of each sensing unit under low-frequency and high-frequency excitation, and combining them into a global multimodal contact fingerprint feature vector according to a preset fixed spatial arrangement order. This comprehensively characterizes the wearing contact state from three physical dimensions: electrical, thermal, and mechanical, providing a richly informative and spatially traceable initial benchmark for subsequent spatial registration.

[0022] Furthermore, the multimodal contact fingerprint feature vector is matched with the expected fingerprint range associated with each acupoint in the pre-constructed standard anatomical coordinate reference model using maximum similarity matching, and the overall rigid offset is solved through iterative optimization. This provides a one-time spatial reference registration for each wear, uniformly mapping the physical coordinates of the sensing unit to the standard anatomical coordinate system, eliminating systematic spatial errors introduced by differences in auricular morphology and inconsistent wearing positions.

[0023] Furthermore, during the acquisition process, high-frequency impedance fluctuations, contact pressure AC components, and temperature drift trends are continuously collected at fixed intervals to form a dynamic contact fingerprint vector. Real-time drift state vectors are obtained through cross-correlation analysis of adjacent time points and time integration. This transforms the microscopic creep and gradual sliding of the earcups into quantifiable spatial deviation data in real time, filling the state monitoring blind spot between initial registration and long-term acquisition. Simultaneously, it integrates contact fingerprint stability index monitoring, automatically pausing acquisition and prompting for re-adhesion when uncontrollable sliding occurs.

[0024] Furthermore, the magnitude of the real-time drift state vector is calculated. When it exceeds the drift threshold dynamically set based on the initial offset magnitude, the initial wearing offset detection is automatically re-executed to update the overall rigid offset. An adaptive closed-loop mechanism from micro-tracking to macro-correction is established to match the recalibration criterion with the actual fit quality of each wear, ensuring spatial consistency of long-term data acquisition.

[0025] Furthermore, the overall rigid offset and real-time drift state vector are applied sequentially to the parameterized 3D auricular digital twin model to drive the real-time update of virtual marker point coordinates. The coordinates of the standard anatomical acupoints corresponding to each sensing unit are obtained through nearest neighbor lookup. By incorporating both spatial error sources into real-time compensation, it is ensured that signals acquired at any given time can be accurately anchored to the correct standard acupoints.

[0026] Finally, the multidimensional feature data of each sensing unit are allocated according to standard acupoint coordinates. Multi-source data for the same acupoint are then weighted and fused using contact pressure as the weight, generating a structured auricular health dataset with timestamps and acupoint labels. This dataset supports cross-date inverse transformation data unification and long-term trend analysis. The spatial registration results are transformed into high-quality, standardized data products that can be directly used for diagnostic analysis, ensuring spatial comparability and fusion of data from different times and different wearing batches.

[0027] In summary, the technical solution of this invention realizes real-time dynamic mapping from physical sensing signals to anatomical acupoint coordinates, and finally outputs a standardized auricular acupoint health dataset with clear spatial semantics and comparable data across time periods. This solves the problem that existing auricular acupoint acquisition systems suffer from inaccurate correspondence between sensing units and acupoints due to ear cover wearing offset and slippage, making it difficult to obtain stable, spatially aligned multimodal auricular acupoint health data. Attached Figure Description

[0028] Figure 1 This is a flowchart illustrating the synchronous acquisition method of auricular biosignals based on multimodal sensor fusion provided by the present invention.

[0029] Figure 2 This is a schematic diagram illustrating the construction process of the ear acupoint health dataset in the ear acupoint biosignal synchronous acquisition method based on multimodal sensor fusion provided by the present invention.

[0030] Figure 3 This is a schematic diagram of the structure of the auricular biosignal synchronous acquisition system based on multimodal sensor fusion provided by the present invention.

[0031] In the attached diagram, the labels representing each component are as follows:

[0032] The module includes: a multimodal initial fingerprint acquisition and feature construction module 11, an initial wearing offset detection module 12, a dynamic fingerprint periodic acquisition and drift tracking module 13, a drift state evaluation and recalibration trigger module 14, a digital twin ear mold coordinate dynamic mapping module 15, and a multimodal dataset dynamic registration and generation module 16. Detailed Implementation

[0033] This invention provides a method and system for synchronous acquisition of auricular biosignals based on multimodal sensor fusion, which solves the problem that existing auricular acupoint acquisition systems suffer from inaccurate correspondence between sensing units and acupoints due to ear cover misalignment and slippage, making it difficult to obtain stable and spatially aligned multimodal auricular acupoint health data.

[0034] It should be noted that the terms "comprising" and "having" are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or modules that are not explicitly listed or that are inherent to these processes, methods, products, or devices.

[0035] Example 1, as Figure 1 As shown, this invention provides a method for synchronous acquisition of auricular biosignals based on multimodal sensor fusion, the method comprising:

[0036] S100: Loads multimodal contact fingerprint data synchronously collected by all sensing units on the flexible multimodal sensing ear cover, and combines the multimodal contact fingerprint data of each sensing unit into a multimodal contact fingerprint feature vector.

[0037] This step performs the first multimodal sensor fingerprint acquisition after system initialization: After the user puts on the flexible multimodal sensor earmuff, all the sensor units distributed on the earmuff are driven synchronously to sense the contact state and acquire multimodal contact fingerprint data such as impedance, capacitance, and micro-strain at each sensor unit; then, the system splices and combines the multiple modal data corresponding to each sensor unit into a unified multidimensional vector that represents the contact characteristics of that point, i.e., the multimodal contact fingerprint feature vector.

[0038] Step S100 in the method provided by the present invention includes:

[0039] The flexible multimodal sensing earmuff is controlled to enter the contact fingerprint scanning mode, and the skin contact impedance amplitude of each sensing unit under low-frequency AC signal and high-frequency AC signal is measured in active mode respectively;

[0040] Local skin temperature values ​​are read by miniature temperature sensors integrated in each sensing unit;

[0041] The static contact pressure value is read by a micro-motion pressure sensor integrated in each sensing unit;

[0042] The skin contact impedance amplitude, local skin temperature value, and static contact pressure value of the same sensing unit are combined to form the multimodal contact fingerprint data of that sensing unit, and the multimodal contact fingerprint data of all sensing units are arranged in a preset order to form a multimodal contact fingerprint feature vector.

[0043] In this step, a control command is first issued to put the flexible multimodal sensing earpiece into contact fingerprint scanning mode. In this mode, low-frequency AC signals and high-frequency AC signals are sequentially injected into each sensing unit through active excitation, and the contact impedance amplitude of the skin at that sensing unit is measured respectively. The low-frequency signal can penetrate deeper tissues and reflect the overall conductivity state of the electrode-skin interface; the high-frequency signal is more sensitive to the water content of the stratum corneum and the microscopic contact area. The combination of the two can more precisely characterize the electrical properties of the contact interface.

[0044] For example, in a certain scan, the No. 3 sensor unit located in the cochlear region first received a 1kHz low-frequency AC excitation and measured a skin contact impedance amplitude of 52kΩ; then it switched to a 100kHz high-frequency AC excitation and measured a contact impedance amplitude of 18kΩ. The two impedance values ​​of this unit [52, 18] were recorded.

[0045] Secondly, while each sensing unit performs impedance measurement, a miniature temperature sensor integrated at the center of the unit initiates a reading operation to acquire the local skin surface temperature value within the area covered by the sensing unit. The local skin surface temperature value reflects the heat distribution characteristics of the auricular acupoint area due to differences in local metabolism and blood perfusion. For example, the miniature temperature sensor of sensing unit number 3 completes the reading and outputs a local skin temperature value of 35.2℃.

[0046] Furthermore, during the acquisition of data from each sensing unit, the micro-motion pressure sensor integrated within that unit simultaneously measures the static contact pressure value between the sensing unit and the skin. The static contact pressure value characterizes the degree of normal pressure exerted on that local point when the earcup is worn. For example, the micro-motion pressure sensor of sensing unit number 3 reads a current static contact pressure value of 0.28 N.

[0047] Finally, for each sensing unit, the measured impedance amplitude, local skin temperature, and static contact pressure values—a total of four dimensions—are combined in a fixed order to form the multimodal contact fingerprint data for that sensing unit. Subsequently, the multimodal contact fingerprint data of all sensing units are sequentially spliced ​​according to the preset spatial arrangement order of the sensing units on the earcup, ultimately generating a global multimodal contact fingerprint feature vector containing all modal information of all sensing units.

[0048] The preset spatial arrangement order is a fixed numbering and arrangement rule followed by all sensing units on the flexible multimodal sensing earmuff when generating global feature vectors. This rule is pre-fixed during the design phase and strictly corresponds to the physical layout of the earmuff and the position index in the standard anatomical coordinate reference model. During setup, the auricle can be divided into several anatomical regions, such as the earlobe region, antitragus region, concha region, triangular fossa region, and helix region. Each region is assigned a fixed region number, which determines the macroscopic arrangement order of the data in the global vector.

[0049] For example, a certain earmuff integrates a total of 12 sensing units, distributed in three anatomical regions: the earlobe region, the cochlea region, and the triangular fossa region. The earlobe region contains three sensing units, namely sensing unit 1, sensing unit 2, and sensing unit 3; the cochlea region contains five sensing units; and the triangular fossa region contains four sensing units, resulting in a total of 12 sensing units in sequence.

[0050] For example, the multimodal contact fingerprint data of the third sensing unit is combined into [52,18,35.2,0.28]. There are a total of 12 sensing units on the earcup. The data of units 1 to 12 are arranged in sequence to finally generate a multimodal contact fingerprint feature vector with a length of 48, that is, 12 units × 4 modes, which serves as the initial reference fingerprint of the entire earcup during this wearing.

[0051] It should be noted that the above values ​​are for illustrative purposes only and do not constitute a limitation on the present invention.

[0052] In summary, this step comprehensively characterizes the contact state between each sensing unit and the skin of the auricle from three physical dimensions: electrical, thermal, and mechanical. This forms a wearing benchmark representation with rich information dimensions and traceable spatial location, effectively solving the problem that single-modal fingerprints are sensitive to interference factors such as skin moisture and pressure tightness.

[0053] S200: Based on the multimodal contact fingerprint feature vector and the standard anatomical coordinate reference model, perform initial wearing offset detection to obtain the overall rigid offset.

[0054] This step calls a pre-built standard anatomical coordinate reference model, matches and compares the multimodal fingerprint data of each sensing unit in the feature vector with the expected fingerprint features at the corresponding positions in the reference model, and calculates the six-degree-of-freedom overall rigid offset representing the current wearing position of the earmuff relative to the standard anatomical position by solving the spatial transformation parameters that minimize the overall matching error.

[0055] Step S200 in the method provided by the present invention includes:

[0056] Load a pre-built standard anatomical coordinate reference model, wherein each standard acupoint coordinate point in the standard anatomical coordinate reference model is associated with a standard contact fingerprint expected range;

[0057] The multimodal contact fingerprint feature vector is matched with the expected range of the standard contact fingerprint to obtain the rigid transformation parameters that minimize the overall matching error. The rigid transformation parameters include horizontal displacement, vertical displacement and rotation angle, which are set as the overall rigid offset.

[0058] In this step, the standard anatomical coordinate reference model corresponding to the current earbud model is first loaded from a pre-built model library. This standard anatomical coordinate reference model was calibrated using extensive clinical data during the construction phase. Each standard acupoint coordinate point not only includes its three-dimensional spatial position in the standard auricular coordinate system but also stores the expected range of multimodal contact fingerprints that should be presented under normal wearing conditions. The expected range is obtained from the mean values ​​of four dimensions: low-frequency impedance, high-frequency impedance, local skin temperature, and static contact pressure, expressed as mean ± standard deviation.

[0059] For example, in the loaded standard anatomical coordinate reference model, the coordinates of the 5th standard acupoint located in the cochlear region have the following associated standard contact fingerprint expected ranges: low-frequency impedance [40,60]kΩ, high-frequency impedance [12,22]kΩ, temperature [34.5,36.5]℃, and pressure [0.20,0.40]N.

[0060] Furthermore, the generated global multimodal contact fingerprint feature vector is traversed and matched against the expected fingerprint range associated with all standard acupoint coordinate points in the standard anatomical coordinate reference model. For each sensing unit, the deviation of its actual fingerprint data from falling into each dimension of the corresponding expected range is calculated, and the overall similarity objective function is constructed by combining the matching errors of all sensing units.

[0061] The similarity objective function uses an iterative optimization algorithm to find the rigid transformation parameters that minimize the overall matching error. These rigid transformation parameters describe the offset of the earpiece's current wearing posture relative to its standard position, including horizontal displacement Δx, vertical displacement Δy, and rotation angle Δθ around the normal. The minimum rigid transformation parameter obtained is the overall rigid offset.

[0062] Preferably, the above iterative optimization algorithm can employ the Levenberg-Marquardt algorithm, defining the overall matching error as the sum of squared normalized deviations of all sensing units in each modal dimension. The deviation is defined as follows: if the measured fingerprint data falls within the expected range, the normalized deviation is 0; if it does not fall within the expected range, the normalized deviation is defined as the difference between the measured value and the nearest boundary value divided by the dimensional tolerance. Setting the tolerance to 50% of the expected range filters out fluctuations within the controllable range of the wearing condition, avoiding misjudgments of wearing misalignment due to temporary interference, while retaining sensitivity to genuine structural misalignment.

[0063] The convergence condition of the above iterative optimization algorithm is set as follows: convergence is determined when the absolute value of the difference between the objective function values ​​of two adjacent iterations is less than a preset threshold or when a preset maximum number of iterations is reached. The preset threshold is set based on the fact that it is less than the contribution of the minimum permissible deviation of each sensing unit's single mode to the objective function, to avoid premature termination of the algorithm when there is still substantial registration benefit. An example value can be set to 1×102. -6 The preset value for the maximum number of iterations is set to ensure that, under conditions of sensor size and hardware computing power, the worst-case convergence path can still be completed within the time limit required for real-time operation. An example value can be set to 50 iterations.

[0064] For example, in the current wearing scenario, the fingerprint data actually measured by sensor unit 5 is [48, 17, 35.8, 0.33]. All of these values ​​fall within the expected range for the acupoint, indicating a small matching error. However, the measured data from sensor unit 8 deviates significantly from the expected range. Considering the matching results of all 12 units, an iterative optimization algorithm is used to solve the problem, ultimately yielding the overall rigid offset as: Δx = +1.2mm, Δy = -0.8mm, and Δθ = +3.5°. These parameters represent the output overall rigid offset.

[0065] It should be noted that the above values ​​are for illustrative purposes only and do not constitute a limitation on the present invention.

[0066] In summary, this step provides a one-time spatial reference registration for each wear, which can uniformly map the measurement coordinates of all sensing units to the standard anatomical coordinate system, fundamentally eliminating the systematic spatial errors introduced by differences in auricular shape and inconsistent wearing positions.

[0067] S300: During the process of collecting auricular biosignals, dynamic contact fingerprint data is repeatedly collected at fixed intervals. Based on the time series of the dynamic contact fingerprint data, dynamic drift real-time tracking is performed to obtain a real-time drift state vector.

[0068] After continuously acquiring multimodal dynamic contact fingerprint data from each sensing unit at fixed intervals, this step involves real-time tracking and analysis of the time-series changes in this data. By comparing the spatial distribution patterns of the dynamic fingerprints at adjacent time points, a real-time drift state vector reflecting the cumulative minute slippage and rotation that occurs after the earcup is initially worn is continuously calculated and output.

[0069] Step S300 in the method provided by the present invention includes:

[0070] With a preset fixed sampling period, the high-frequency impedance fluctuation, contact pressure AC component and temperature drift trend of each sensing unit are continuously collected to form a dynamic contact fingerprint vector at each sampling moment.

[0071] Calculate the cross-correlation function between the dynamic contact fingerprint vectors at adjacent sampling times, and determine the overall displacement increment from the previous time to the current time based on the peak position of the cross-correlation function;

[0072] The overall displacement increment is integrated over time to obtain the cumulative horizontal displacement, cumulative vertical displacement, and cumulative rotation angle, which are then combined into a real-time drift state vector.

[0073] In this step, three dynamic characteristic quantities of all sensing units are first collected synchronously at a preset fixed sampling period: high-frequency impedance fluctuation, contact pressure AC component, and temperature drift trend. Among them, high-frequency impedance fluctuation, contact pressure AC component, and temperature drift trend are all instantaneous changes in the contact state of each sensing unit relative to the previous moment or steady-state value. The combination of the three constitutes the dynamic contact fingerprint vector at the current sampling moment.

[0074] The preset fixed sampling period is a standardized sampling time interval used for real-time tracking of dynamic drift during the acquisition of auricular biosignals. This setting is based on a balance between the rate of change of dynamic contact fingerprint data and the system's computing power: a period that is too long will cause a lag in drift tracking response, failing to capture transient slippage of the earcup caused by facial movements in a timely manner; a period that is too short will introduce high-frequency measurement noise and increase unnecessary computational load. Considering the typical time constant of changes in earcup wearing status, an example value for the preset fixed sampling period can be set to 2 seconds.

[0075] For example, during continuous data acquisition, a dynamic contact fingerprint scan is triggered every 2 seconds to sequentially acquire dynamic data at times such as 0 seconds, 2 seconds, and 4 seconds, which serves as a unified time reference for subsequent cross-correlation calculations and drift integral tracking.

[0076] For example, 120 seconds after the start of auricular biosignal acquisition, a dynamic fingerprint acquisition is triggered. At this time, the 5th sensor unit located in the cochlea region measures: high-frequency impedance fluctuation of +0.6kΩ, contact pressure AC component of -0.03N, and temperature drift trend of +0.04℃ / min. The dynamic contact fingerprint vector of the current unit is then recorded as [+0.6, -0.03, +0.04]. All 12 sensor units simultaneously complete acquisition, obtaining 12 sets of three-dimensional dynamic contact fingerprint vectors.

[0077] Secondly, the dynamic contact fingerprint vectors at the current sampling time and the previous sampling time are considered as two spatially distributed signals, and the spatial cross-correlation function between them is calculated. Specifically, the dynamic fingerprint vectors of each sensing unit at the previous sampling time are virtually translated along the candidate displacement direction of the auricle surface. The similarity between the fingerprint distribution at the previous sampling time and the actual fingerprint distribution at the current sampling time is calculated for each displacement, and the spatial displacement that maximizes the correlation coefficient is searched. The spatial displacement coordinates corresponding to the peak of the cross-correlation function are the overall displacement increment of the sensing unit array, reflecting the microscopic rigid sliding that occurs in the ear cup during the sampling period.

[0078] Preferably, the similarity is calculated using the Pearson correlation coefficient, which is calculated by dividing the covariance of the two sets of data by the product of their respective standard deviations. The value range is [-1, +1]: +1 indicates a perfect positive correlation, meaning that the two sets of data have completely identical change patterns; 0 indicates no linear correlation; and -1 indicates a perfect negative correlation.

[0079] For example, the similarity of the dynamic contact fingerprint vectors from cycles 59 and 60 is calculated, i.e., the Pearson correlation coefficient is calculated. After a comprehensive search, the cross-correlation function reaches a peak of 0.94 at the candidate displacement coordinates (Δx=+0.12mm, Δy=-0.07mm, Δθ=+0.10°). The results show that within the 2-second interval from cycle 59 to cycle 60, the earpiece slides 0.12mm to the right, 0.07mm downwards, and rotates 0.10° counterclockwise; this set of values ​​represents the overall displacement increment for this cycle.

[0080] Finally, starting from the initial registration time, the overall displacement increment calculated each time is continuously integrated over time, i.e., accumulated periodically, to obtain the cumulative drift amount relative to the initial wearing position at the current sampling time. The integration result includes three components: cumulative horizontal displacement, cumulative vertical displacement, and cumulative rotation angle, which are combined into a real-time drift state vector. The real-time drift state vector quantitatively describes the overall evolution of the earpiece wearing posture since the start of data acquisition.

[0081] For example, from cycle 1 to cycle 60, 59 displacement increment calculations have been completed consecutively. These 59 increments are then accumulated: the cumulative horizontal displacement is +1.25 mm, the cumulative vertical displacement is -0.85 mm, and the cumulative rotation angle is +3.2°. Therefore, the real-time drift state vector output at the 60th sampling time is [+1.25, -0.85, +3.2].

[0082] Step S400 in the method provided by the present invention further includes:

[0083] During the real-time tracking of dynamic drift, the time integral of the standard deviation of the high-frequency impedance fluctuation of all sensing units within a preset time window is calculated and set as the contact fingerprint stability index.

[0084] When the contact fingerprint stability index exceeds a preset safety threshold, it is determined that the flexible multimodal sensor earmuff has experienced continuous uncontrollable slippage, the ear acupoint biosignal acquisition is automatically paused, and a prompt to re-fit is issued through the user prompt unit.

[0085] Specifically, during the continuous operation of dynamic drift real-time tracking, high-frequency impedance fluctuation values ​​of all sensing units are collected within a preset time window. The standard deviation is calculated to measure the spatial dispersion of the fluctuation amplitude of each unit. If the earcup is generally stable, the fluctuation difference between units is small, and the standard deviation remains low. If the earcup experiences local or overall uncontrollable slippage, the fluctuation of some units increases dramatically, leading to a significant increase in the standard deviation. This standard deviation is then integrated over time, and the standard deviation values ​​of multiple consecutive windows are summed to obtain the contact fingerprint stability index, which reflects the cumulative instability of the contact state over a period of time.

[0086] The preset time window is a fixed-duration sliding interval used to define the statistical range of the standard deviation of high-frequency impedance fluctuations when calculating the contact fingerprint stability index. Its setting is based on the following: an excessively long window will introduce too much historical stable data, diluting the abnormal signal of the current sudden slippage and causing a delay in the judgment of uncontrollable slippage; an excessively short window may produce false alarms due to brief fluctuations caused by instantaneous physiological disturbances. Considering that the time scale for uncontrollable slippage of the earcup is usually several seconds to more than ten seconds, an example value for the preset time window can be set to 10 seconds.

[0087] For example, if the preset time window is set to 10 seconds, within the time window from the 120th to the 130th second, the high-frequency impedance fluctuation range of the 3 units in the earlobe area is [-0.3, +0.4] kΩ, the range of the 5 units in the cochlea area is [-0.5, +0.6] kΩ, the range of the 4 units in the triangular fossa area is [-0.4, +0.3] kΩ, and the standard deviation of the fluctuation of all 12 units is 0.22 kΩ.

[0088] Furthermore, the contact fingerprint stability index is compared with a preset safety threshold. When the stability index is lower than the safety threshold, it indicates that the contact state is within a controllable range, and normal collection continues; when the stability index reaches or exceeds the safety threshold, it is determined that the earpiece has experienced continuous uncontrollable slippage, at which point dynamic drift tracking and coordinate compensation can no longer reliably maintain the acupoint positioning accuracy.

[0089] The preset safety threshold is a critical value for the contact fingerprint stability index used to determine whether the sensing earcup is experiencing continuous uncontrollable slippage. Its setting is based on the following: under normal and stable wearing conditions, the contact fingerprint stability index remains at a low level; when the contact fingerprint stability index rapidly increases due to continuous slippage of the earcup and exceeds a certain limit, it indicates that the compensation capability of dynamic drift tracking has reached its limit, and continued acquisition will lead to serious inaccuracies in acupoint positioning. An example value for the preset safety threshold can be set to 8.0, meaning that when the cumulative standard deviation of multiple consecutive time windows reaches 8.0, the contact state is determined to be uncontrollable, automatically triggering a pause in acquisition and a prompt for re-adhesion.

[0090] For example, starting from the 120th second, due to the user's continuous talking causing frequent ear movements, the high-frequency impedance fluctuations in multiple units of the cochlea region continuously increase. By the 180th second, the standard deviations of six consecutive time windows have been accumulated, and the contact fingerprint stability index has reached 8.5, exceeding the preset safety threshold of 8.0. Therefore, it is determined that the slippage of the earcup has exceeded the compensation range of dynamic tracking.

[0091] Finally, when continuous uncontrollable slippage is detected, the current ear acupoint biosignal acquisition process is automatically paused to prevent subsequent data acquisition from introducing erroneous information due to severe acupoint positioning inaccuracies. Furthermore, through user prompting units, such as indicator lights on the wearable device, vibration modules, or pop-ups on the host computer software interface, a prompt to readjust the fit is issued to the user, guiding the user to remove the ear tips and readjust them correctly. Data acquisition will resume only after the initial wear offset detection is completed again.

[0092] For example, after the system determines that the slippage is uncontrollable at 180 seconds, the ear acupoint signal acquisition module immediately stops working, and the acquisition interface displays that the acquisition has been paused. At the same time, the LED indicator on the earcup changes from solid green to flashing red, and the host computer software pops up a prompt box indicating that the earcup has slipped too much. Please put the earcup back on and ensure a tight fit before continuing. After the user puts the earcup back on and completes the initial registration as prompted, the acquisition process resumes normally at 210 seconds.

[0093] It should be noted that the above values ​​are for illustrative purposes only and do not constitute a limitation on the present invention.

[0094] In summary, this step transforms the microscopic creep and gradual sliding of the earcup that are imperceptible in traditional methods into spatial deviation data that can be quantified and tracked by the system in real time, filling the monitoring blind spot of changes in wearing status between the initial one-time registration and long-term continuous acquisition.

[0095] S400: Calculate the magnitude of the real-time drift state vector. When the magnitude is greater than or equal to a preset drift threshold, trigger the re-execution of the initial wearing offset detection to update the overall rigidity offset.

[0096] This step calculates the modulus of the output real-time drift state vector, comprehensively reflecting the cumulative spatial offset of the earcup since initial registration. The modulus value is compared with a preset drift threshold in real time. Once the modulus reaches or exceeds the threshold, the initial wearing offset detection process is automatically re-executed to generate a new overall rigid offset to replace the old value, completing the update and calibration of the wearing reference.

[0097] In this step, the preset drift threshold is dynamically set according to the modulus of the overall rigidity offset, wherein the preset drift threshold is obtained by multiplying the modulus of the overall rigidity offset by a preset proportional coefficient.

[0098] Specifically, after each initial wearing offset detection and output of the overall rigid offset, the modulus of this offset is calculated simultaneously. The modulus is calculated as a weighted combination of the displacement and angular components, i.e., modulus = , where k is the angle-to-displacement conversion coefficient, which equates the rotation angle to the linear displacement of the outer edge of the sensor array. The larger the modulus value, the more significant the deviation between the initial position of the device and the standard anatomical coordinate reference model.

[0099] The angle-displacement conversion coefficient k is calculated as follows: assuming the equivalent radius of the area covered by the sensor unit array is r, approximately 15~25mm, then when the rotation angle is in degrees, k=r×(π / 180), approximately 0.26~0.44mm / °; for ease of engineering implementation, k can be simplified to a fixed constant within this range, such as k taking the value of 0.35mm / °.

[0100] For example, in a certain wearing scenario, the overall rigidity offset output is Δx = +1.2mm, Δy = -0.8mm, Δθ = +3.5°, and the angle-displacement conversion coefficient k = 0.35mm / °, then the modulus = ≈1.89mm

[0101] Furthermore, the obtained overall rigidity offset modulus is multiplied by a preset scaling factor to dynamically generate a preset drift threshold applicable to this wearing process.

[0102] The proportional coefficient is set based on the following: when the initial wearing deviation is large, the correspondence between the sensing unit and the acupoint has significant initial uncertainty, and the allowable margin for dynamic drift tracking should be narrowed accordingly; when the initial wearing deviation is small and the fit is good, the allowable margin can be appropriately widened. The preset proportional coefficient is usually between 0.3 and 0.8, with a default recommended value of 0.5, meaning the drift threshold is set to half the initial offset modulus length. For example, based on the modulus length of 1.89mm, with a preset proportional coefficient of 0.5, the dynamically calculated preset drift threshold is 1.89 × 0.5 ≈ 0.95mm.

[0103] Finally, during dynamic drift tracking, the magnitude of the real-time drift state vector is continuously calculated and compared with a dynamically generated preset drift threshold. When the cumulative drift magnitude reaches or exceeds this dynamic threshold, it indicates that the actual spatial offset of the earcup has accumulated to an unacceptable level based on the initial deviation. In this case, the initial wearing offset detection is automatically re-executed to update the overall rigid offset, and the cumulative drift vector in dynamic drift tracking is cleared to zero, starting a new round of drift monitoring.

[0104] For example, if the real-time drift state vector is [+1.0, -0.5, +2.8] at the 350th second of data acquisition, and the angular component k = 0.35 mm / °, then the real-time drift modulus = If the deviation is approximately 1.49mm, exceeding the preset drift threshold of 0.95mm, recalibration is immediately triggered. After re-performing the initial wearing offset detection, new overall rigidity offsets are obtained: Δx = +0.3mm, Δy = -0.2mm, and Δθ = +0.8°. The modulus length is approximately 0.55mm. Multiplying this new modulus length by 0.5 yields the updated drift threshold of 0.28mm. The cumulative dynamic tracking value is then reset to zero, and data acquisition continues.

[0105] It should be noted that the above values ​​are for illustrative purposes only and do not constitute a limitation on the present invention.

[0106] In summary, this step avoids the accuracy decay or tracking loss that occurs when dynamic drift tracking is subjected to excessive cumulative offset. Furthermore, by dynamically setting a threshold based on the initial offset modulus, the recalibration trigger criterion is matched with the actual fit quality of each wear, ensuring that the mapping relationship between the sensing unit and the standard anatomical acupoint coordinates remains within a clinically acceptable range throughout the entire long-term acquisition process.

[0107] S500: Based on the overall rigidity offset and the real-time drift state vector, drive the coordinates of the digital twin ear mold to update in real time, and obtain the standard anatomical acupoint coordinates corresponding to the measurement values ​​of each sensing unit.

[0108] This step uses a pre-constructed digital twin earmold as a three-dimensional spatial carrier. The output overall rigidity offset is applied as a reference transformation parameter to the overall coordinate system of the earmold. Simultaneously, the continuously output real-time drift state vector is superimposed as a dynamic correction on the position attributes of each sensing unit, driving the spatial coordinates of the virtual marker points corresponding to each sensing unit in the digital twin earmold to update in real time according to the wearing state. Thus, the raw physical coordinates directly collected by the sensing units are converted into standard acupoint coordinates with anatomical significance.

[0109] Step S500 in the method provided by the present invention includes:

[0110] Load a pre-established parametric three-dimensional digital twin model of the auricle, wherein the parametric three-dimensional digital twin model of the auricle has a coordinate system consistent with the standard anatomical coordinate reference model;

[0111] The overall rigid offset is used as the initial transformation matrix, and the real-time drift state vector is used as the dynamic compensation matrix. These are applied sequentially to the parameterized three-dimensional auricle digital twin model to obtain the dynamic calibration coordinates corresponding to the physical coordinates of each sensing unit.

[0112] The coordinates of the closest standard acupoint in the standard anatomical coordinate reference model are found based on the dynamic calibration coordinates, and these coordinates are used as the standard anatomical acupoint coordinates corresponding to the measurement value of the corresponding sensing unit at the current moment.

[0113] In this step, a parametric 3D digital twin model of the auricle, matching the current user's auricle morphology, is first loaded from a pre-built model library. This model is based on a standard auricle template and is parametrically driven by extracting key anatomical morphological parameters of the auricle. These key parameters include geometric features such as the total length and width of the auricle, the depth of the cochlea, the length of the earlobe, the angle of the triangular fossa, and the radius of curvature of the helix. During construction, the normal distribution of these parameters is statistically analyzed from clinical 3D auricle scan data to form a parametric deformation algorithm. In use, by inputting the measured morphological parameters of a specific individual, the standard template is driven to deform accordingly, generating a personalized 3D mesh model that highly matches that individual's auricle.

[0114] The parametric 3D auricular digital twin model includes a 3D curved mesh composed of tens of thousands of triangular facets, which accurately describes the geometry of the auricle. The mesh density is increased in areas with dense acupoints, such as the cochlea, to ensure the spatial resolution of coordinate mapping. A virtual marker array is also included, with spatial coordinate points on the mesh surface that correspond one-to-one with the physical earpiece sensing unit. Initially, each marker point is located at the designed wearing position. The coordinate system calibration information is also included, with the model's coordinate system origin, axis, and scale strictly aligned with the standard anatomical coordinate reference model to ensure a unified benchmark for coordinate transformation and registration calculations.

[0115] For example, the loaded parametric 3D auricular digital twin model contains a complete auricular surface mesh including the earlobe area, cochlea area, and triangular fossa area. The virtual marker point corresponding to the 5th sensing unit is initially located near the cochlea area's central acupoint, with coordinates (12.5, 8.3, -3.1), which is consistent with the standard coordinates (12.5, 8.3, -3.1) of the central acupoint in the standard anatomical coordinate reference model.

[0116] Furthermore, the overall rigid offset output is used as the initial transformation matrix to apply a one-time rigid transformation to the parameterized 3D auricular digital twin model, causing the model to translate and rotate as a whole, eliminating initial wearing deviations. Subsequently, the real-time output drift state vector is used as the dynamic compensation matrix to apply a small correction based on the initial transformation, reflecting the cumulative drift of the earpiece relative to the initial calibration position at the current moment. After the two transformations are superimposed, the physical coordinates of each virtual marker point in the model are converted into dynamic calibration coordinates.

[0117] The transformation matrix employs a two-dimensional rigid transformation formula, and the projection onto the Z-axis normal direction can be approximated as a two-dimensional planar transformation. After rotating around the origin by an angle θ, the relationship between the new coordinates (x', y') and the original coordinates (x, y) is: x' = x·cosθ - y·sinθ, y' = x·sinθ + y·cosθ. The translation components are directly superimposed.

[0118] For example, if the overall rigid offset of the output is Δx = +1.2mm, Δy = -0.8mm, and Δθ = +3.5°, then an initial transformation matrix is ​​generated to perform overall translation and rotation on the digital twin model. cos(3.5°) ≈ 0.9981, sin(3.5°) ≈ 0.0610, then x1 = (12.5 × 0.9981 - 8.3 × 0.0610) + 1.2 = 13.170. y1 = (12.5 × 0.0610 + 8.3 × 0.9981) - 0.8 = 8.247. At the 350th second, the output real-time drift state vector is [+1.0, -0.5, +2.8°], then a dynamic compensation matrix is ​​generated for further correction. cos(2.8°)≈0.9988, sin(2.8°)≈0.0488, x2=(13.170×0.9988-8.247×0.0488)+1.0=13.752. y2=(13.170×0.0488+8.247×0.9988)-0.5=8.380. Finally, after two transformations, the dynamic calibration coordinates of the 5th sensor unit were updated from the initial (12.5,8.3,-3.1) to (13.752,8.380,-3.1).

[0119] Finally, after obtaining the current dynamic calibration coordinates of each sensing unit, a nearest neighbor search is performed in the standard anatomical coordinate reference model. That is, using the calibration coordinates as the query point, all standard acupoint coordinate points in the model are traversed, the spatial Euclidean distance is calculated, and the standard acupoint coordinate point with the smallest distance is selected as the matching result. This sensing unit is then assigned as the spatial label of the current measurement value.

[0120] For example, the dynamic calibration coordinates of sensor unit 5 are (13.752, 8.380, -3.1). Searching for the nearest standard acupoint coordinates in the standard anatomical coordinate reference model, the standard coordinates of the Heart acupoint are found to be (14.0, 7.0, -2.7), with an Euclidean distance of only about 0.22 mm, thus considered the best match. Meanwhile, the dynamic calibration coordinates of sensor unit 6 in the cochlea region are closest to the standard coordinates of the Lung acupoint, and the standard coordinates of sensor unit 2 in the earlobe region are closest to the standard coordinates of the Eye acupoint. The impedance, temperature, and other biosignal data collected by each sensor unit at the current moment are all assigned corresponding standard acupoint coordinates, achieving spatial registration.

[0121] It should be noted that the above values ​​are for illustrative purposes only and do not constitute a limitation on the present invention.

[0122] In summary, this step incorporates both initial wearing deviation and long-term acquisition drift, two sources of spatial error, into real-time compensation. This ensures that multimodal biosignals acquired at any given time can be accurately anchored to the correct standard acupoint coordinates, providing a spatially semantically clear and cross-time-period comparable positioning foundation for the subsequent generation of auricular health datasets.

[0123] S600: Map the multidimensional feature data collected by each sensing unit to the coordinates of the standard anatomical acupoints to generate a dynamically registered auricular acupoint health dataset.

[0124] like Figure 2 As shown, after outputting the standard anatomical acupoint coordinates corresponding to each sensing unit in real time, this step binds and maps the multidimensional feature data collected by each sensing unit at the same time with the corresponding standard acupoint coordinates according to the timestamp, and finally integrates them to generate a structured ear acupoint health dataset in which all data have eliminated wearing space errors and achieved dynamic registration of acupoint positions.

[0125] Step S600 in the method provided by the present invention includes:

[0126] Acquire multidimensional feature data collected by each sensing unit, including skin potential difference in passive mode and multi-frequency bioelectrical impedance amplitude and phase angle in active mode;

[0127] The multidimensional feature data is assigned to the corresponding standard acupoints according to the standard anatomical acupoint coordinates corresponding to each sensing unit;

[0128] When the same standard acupoint is mapped by multiple sensing units, the weighted average value is calculated using the contact pressure value of each sensing unit as the weight, and this average value is used as the final feature value of the standard acupoint.

[0129] The final feature values ​​and timestamps of all standard acupoints are organized into a dynamically registered auricular acupoint health dataset.

[0130] In this step, multidimensional feature data collected by all sensing units are acquired synchronously at each sampling time, including four types of biosignal parameters across two operating modes. In passive mode, the skin surface potential difference at the sensing unit is measured using a high input impedance differential amplifier, reflecting the intensity of spontaneous bioelectrical activity at the acupoint. In active mode, microampere-level sinusoidal excitation currents at multiple frequencies are injected sequentially, and the amplitude and phase angle of the bioelectrical impedance at each frequency are simultaneously demodulated and measured, reflecting the conductivity characteristics and cell membrane integrity of the local tissue at the acupoint.

[0131] For example, at the 350-second sampling time, the No. 5 sensor unit located in the cochlear region completed multi-dimensional feature acquisition: in passive mode, the skin potential difference was measured to be 8.3mV; in active mode, the bioelectrical impedance amplitudes at three frequency points of 1kHz, 10kHz, and 100kHz were measured to be 52kΩ, 31kΩ, and 18kΩ, respectively, with corresponding phase angles of -15°, -22°, and -28°. The multi-dimensional feature data of this unit is recorded as [8.3,52,31,18,-15,-22,-28].

[0132] Secondly, based on the real-time mapping relationship between each sensor unit and the coordinates of the standard anatomical acupoint, the collected multidimensional feature data is assigned to the corresponding standard acupoint. The multidimensional feature data vector of each sensor unit carries the spatial attribution label at the current moment and is written into the temporary data buffer of the corresponding standard acupoint, awaiting further fusion processing.

[0133] For example, at 350 seconds, the dynamic calibration coordinates of sensor unit 5 are output as (13.752, 8.380, -3.1). Through nearest neighbor lookup, these coordinates are found to be closest to the standard coordinates of the Heart acupoint in the standard anatomical coordinate reference model (14.0, 7.0, -2.7) in Euclidean distance, indicating a successful match. The multidimensional feature data [8.3, 52, 31, 18, -15, -22, -28] of sensor unit 5 are assigned to the Heart acupoint label. Simultaneously, sensor unit 6 in the cochlea region is assigned to the Lung acupoint, sensor unit 2 in the earlobe region is assigned to the Eye acupoint, and sensor unit 9 in the triangular fossa region is assigned to the Shenmen acupoint.

[0134] Furthermore, when a standard acupoint in the standard anatomical coordinate reference model is simultaneously mapped by multiple sensing units, the weighted average of each dimension of the mapped data for that acupoint is calculated using the contact pressure value of each sensing unit as the weight, and this average is used as the final feature value of the standard acupoint. Sensing units with larger contact pressure values ​​are considered to have closer contact with the skin and higher signal quality, and therefore have a greater weight in the fusion process. If an acupoint is mapped by only a single sensing unit, then the feature value of that unit is directly used as the final feature value. Specifically, the final feature value = (mapped data × corresponding contact pressure value) / sum of contact pressure values ​​of all units.

[0135] For example, at 350 seconds, the acupoint is simultaneously mapped by two sensing units, number 5 and number 7. Unit 5 has a contact pressure of 0.33 N and a low-frequency impedance amplitude of 52 kΩ in its characteristic data; unit 7 has a contact pressure of 0.21 N and a low-frequency impedance amplitude of 48 kΩ. Therefore, the weighted average of the low-frequency impedance amplitude of the acupoint is calculated as: (0.33 × 52 + 0.21 × 48) / (0.33 + 0.21) ≈ 50.4 kΩ. The characteristic values ​​of other dimensions are calculated similarly to obtain the final characteristic value of the acupoint. The eye acupoint is mapped only by unit number 2, so the characteristic value of that unit is directly used as the final characteristic value of the eye acupoint.

[0136] Finally, the final feature values ​​of all standard acupoints, along with the current sampling timestamp, are organized into a registered auricular acupoint health data record according to a preset data structure. As the data acquisition progresses, data records at each sampling time are continuously added in chronological order, ultimately forming a complete dynamically registered auricular acupoint health dataset. Each record includes a timestamp, standard acupoint identifier, multidimensional feature values, and fusion weight information for each modality.

[0137] For example, the data record at the 350th second is organized as follows: Timestamp = 350s, Heart Point = [Potential difference 8.3mV, Impedance amplitude (50.4, 30.2, 17.8)kΩ, Phase angle (-16.1°, -21.5°, -28.0°)], Lung Point = [Potential difference 6.1mV, Impedance amplitude (58.2, 35.4, 20.1)kΩ, Phase angle (-12.3°, - [18.7°, -24.5°], Eye acupoint = [potential difference 4.2mV, impedance amplitude (63.5, 38.1, 23.4)kΩ, phase angle (-10.5°, -16.2°, -21.8°)], Shenmen acupoint = [potential difference 5.8mV, impedance amplitude (71.3, 42.6, 26.9)kΩ, phase angle (-9.8°, -15.1°, -20.3°)]. These records, together with the records from previous and subsequent times, constitute a complete ear acupoint health dataset.

[0138] Step S600 in the method provided by the present invention further includes:

[0139] When a user performs multiple wear detections on different dates, the historical overall rigid offset and historical real-time drift state vector corresponding to each detection are loaded respectively.

[0140] For the auricular health dataset obtained from each detection, based on the overall rigid offset and the inverse transformation of the real-time drift state vector of that detection, all feature values ​​are uniformly mapped to the same benchmark digital twin auricular coordinate system.

[0141] Under the reference digital twin ear mold coordinate system, the changing trends of multidimensional feature values ​​of the same standard acupoint on different dates are compared to generate a long-term dynamic health analysis report.

[0142] Specifically, after a user performs multiple wear tests on different dates and completes a single data collection, the spatial registration parameters corresponding to each test are loaded sequentially from the historical database, including the overall rigid offset and the real-time drift state vector time series.

[0143] For example, User A underwent ear acupuncture testing three times over three consecutive weeks. The overall rigidity offset of the first test was Δx = +1.2mm, Δy = -0.8mm, and Δθ = +3.5°, with a real-time drift state vector mean of [+0.3, -0.2, +0.8°]; the overall rigidity offset of the second test was Δx = -0.6mm, Δy = +1.1mm, and Δθ = -2.1°, with a drift mean of [+0.2, +0.4, -0.5°]; and the overall rigidity offset of the third test was Δx = +0.4mm, Δy = -0.3mm, and Δθ = +1.0°, with a drift mean of [-0.1, -0.2, +0.3°].

[0144] Furthermore, a unified baseline digital twin ear mold coordinate system is defined, perfectly aligned with the coordinate system of the standard anatomical coordinate reference model. For each ear acupoint health dataset obtained from each test, an inverse transformation matrix is ​​calculated using the overall rigidity offset and real-time drift state vector corresponding to that test. This matrix is ​​then applied inversely to the coordinates of each standard acupoint and its associated multidimensional feature values ​​in the dataset, eliminating spatial distortions introduced by wearing deviations and drift, and uniformly mapping all feature values ​​to the standard acupoint positions in the baseline coordinate system.

[0145] The inverse transformation matrix is ​​calculated as follows: x = (x'-Δx)·cos(-θ) - (y'-Δy)·sin(-θ) = (x'-Δx)·cosθ + (y'-Δy)·sinθ, and similarly, y = (x'-Δx)·sinθ + (y'-Δy)·cosθ. For the overall rigid offset and real-time drift state vector detected each time, their equivalent mapping position in the reference coordinate system is calculated according to the above inverse transformation formula. Based on this, it is determined that the actual skin area covered by each sensing unit in this detection corresponds to the standard acupoint in the reference coordinate system.

[0146] For example, the overall rigid offset measured in the first test is Δx = +1.2mm, Δy = -0.8mm, and Δθ = +3.5°, indicating that the earcup is shifted to the right and downward and rotated counterclockwise. During the inverse transformation, the coordinates of the sensor unit in the wearing state are first reduced by the translation, and then rotated 3.5° in the opposite direction. The dynamic calibration coordinates of sensor unit #5 in the wearing state are (13.752, 8.380), and the inverse transformation is calculated as follows: x = (13.752 - 1.2) × 0.9981 + (8.380 - (-0.8)) × (0.0610) = 13.088, y = -(13.752 - 1.2) × (0.0610) + (8.380 - (-0.8)) × 0.9981 = 8.397.

[0147] Finally, within the coordinate system of the benchmark digital twin ear mold, all detection data were spatially aligned using standard acupoints as the unit. Multidimensional feature values ​​of the same standard acupoint were extracted over time series from different detection dates, and the amount of change, rate of change, and statistical trend indicators were calculated dimension by dimension. Combined with preset physiological reference ranges, the long-term evolution of the biophysical characteristics of acupoints was analyzed. Ultimately, a long-term dynamic health analysis report was generated, including multidimensional feature trend curves for each acupoint, abnormal fluctuation markers, and health interpretation suggestions.

[0148] The pre-defined physiological reference range is constructed based on auricular biosignal acquisition data from healthy individuals. Healthy volunteers of different ages and genders were recruited to collect multi-dimensional characteristic data of each standard acupoint under standard wearing conditions, including skin potential difference in passive mode, amplitude and phase angle of multi-frequency bioelectrical impedance in active mode, etc. After statistical processing, the mean ± 1.96 times the standard deviation, i.e., the 95% confidence interval, was used as the normal reference range for each modality of each acupoint, and was pre-stored in the standard anatomical coordinate reference model for later use.

[0149] For example, at the central acupoint on the baseline coordinate system, three test results were summarized: Day 1: low-frequency impedance amplitude 50.4kΩ, potential difference 8.3mV; Day 8: low-frequency impedance amplitude 46.2kΩ, potential difference 9.1mV; Day 15: low-frequency impedance amplitude 43.8kΩ, potential difference 9.8mV. The low-frequency impedance amplitude at the Heart acupoint showed a continuous decreasing trend, with a decrease of approximately 13%, while the potential difference showed an increasing trend, with an increase of approximately 18%. According to the clinical interpretation rules of the Heart acupoint corresponding to the heart reflex zone, the above trends may indicate a gradual change in the functional state of the cardiovascular system. The report marked the Heart acupoint entry with a progressive decrease in impedance and a progressive increase in potential, suggesting attention to cardiovascular health, and generated a visual trend curve. Other standard acupoints were processed similarly, and finally summarized into a long-term dynamic health analysis report covering all 12 standard acupoints in the earlobe area, cochlea area, and triangular fossa area.

[0150] It should be noted that the above values ​​are for illustrative purposes only and do not constitute a limitation on the present invention.

[0151] In summary, this step transforms the results of all previous spatial registration and drift compensation work into high-quality data products that can be directly used for diagnostic analysis. This ensures that multimodal biosignals collected at different times and in different wearing batches have unified spatial semantics and traceable acupoint attribution, eliminating the impact of wearing deviations on data consistency. It provides a standardized data foundation that is accurately located and comparable across time periods for acupoint-based health status identification, disease-assisted diagnostic model training, and long-term dynamic health trend analysis.

[0152] In summary, this invention achieves real-time dynamic mapping from physical sensor signals to anatomical acupoint coordinates, and ultimately outputs a standardized auricular acupoint health dataset with clear spatial semantics and comparability across time periods, providing a reliable data foundation for accurate health status identification and long-term dynamic monitoring based on auricular acupoint biosignals.

[0153] Example 2, as Figure 3 As shown, this invention provides a synchronous acquisition system for auricular biosignals based on multimodal sensor fusion, the system comprising:

[0154] The multimodal initial fingerprint acquisition and feature construction module 11 is used to load the multimodal contact fingerprint data synchronously acquired by all sensing units on the flexible multimodal sensing ear cover, and combine the multimodal contact fingerprint data of each sensing unit into a multimodal contact fingerprint feature vector.

[0155] Specifically, the system loads multimodal contact fingerprint data synchronously collected by all sensing units on the flexible multimodal sensing earmuff, and combines the multimodal contact fingerprint data of each sensing unit into a multimodal contact fingerprint feature vector, including:

[0156] The flexible multimodal sensing earmuff is controlled to enter the contact fingerprint scanning mode, and the skin contact impedance amplitude of each sensing unit under low-frequency AC signal and high-frequency AC signal is measured in active mode respectively;

[0157] Local skin temperature values ​​are read by miniature temperature sensors integrated in each sensing unit;

[0158] The static contact pressure value is read by a micro-motion pressure sensor integrated in each sensing unit;

[0159] The skin contact impedance amplitude, local skin temperature value, and static contact pressure value of the same sensing unit are combined to form the multimodal contact fingerprint data of that sensing unit, and the multimodal contact fingerprint data of all sensing units are arranged in a preset order to form a multimodal contact fingerprint feature vector.

[0160] The initial wearing offset detection module 12 is used to perform initial wearing offset detection based on the multimodal contact fingerprint feature vector and the standard anatomical coordinate reference model to obtain the overall rigid offset.

[0161] Specifically, based on the multimodal contact fingerprint feature vector and the standard anatomical coordinate reference model, initial wearing offset detection is performed to obtain the overall rigid offset, including:

[0162] Load a pre-built standard anatomical coordinate reference model, wherein each standard acupoint coordinate point in the standard anatomical coordinate reference model is associated with a standard contact fingerprint expected range;

[0163] The multimodal contact fingerprint feature vector is matched with the expected range of the standard contact fingerprint to obtain the rigid transformation parameters that minimize the overall matching error. The rigid transformation parameters include horizontal displacement, vertical displacement and rotation angle, which are set as the overall rigid offset.

[0164] The dynamic fingerprint periodic acquisition and drift tracking module 13 is used to repeatedly acquire dynamic contact fingerprint data at a fixed period during the acquisition of auricular biosignals, and perform real-time dynamic drift tracking based on the time series of the dynamic contact fingerprint data to obtain a real-time drift state vector.

[0165] Specifically, dynamic drift real-time tracking is performed based on the time series of the dynamic contact fingerprint data to obtain a real-time drift state vector, including:

[0166] With a preset fixed sampling period, the high-frequency impedance fluctuation, contact pressure AC component and temperature drift trend of each sensing unit are continuously collected to form a dynamic contact fingerprint vector at each sampling moment.

[0167] Calculate the cross-correlation function between the dynamic contact fingerprint vectors at adjacent sampling times, and determine the overall displacement increment from the previous time to the current time based on the peak position of the cross-correlation function;

[0168] The overall displacement increment is integrated over time to obtain the cumulative horizontal displacement, cumulative vertical displacement, and cumulative rotation angle, which are then combined into a real-time drift state vector.

[0169] This also includes:

[0170] During the real-time tracking of dynamic drift, the time integral of the standard deviation of the high-frequency impedance fluctuation of all sensing units within a preset time window is calculated and set as the contact fingerprint stability index.

[0171] When the contact fingerprint stability index exceeds a preset safety threshold, it is determined that the flexible multimodal sensor earmuff has experienced continuous uncontrollable slippage, the ear acupoint biosignal acquisition is automatically paused, and a prompt to re-fit is issued through the user prompt unit.

[0172] The drift state assessment and recalibration trigger module 14 is used to calculate the magnitude of the real-time drift state vector. When the magnitude is greater than or equal to a preset drift threshold, it triggers the re-execution of the initial wearing offset detection to update the overall rigidity offset.

[0173] The preset drift threshold is dynamically set based on the modulus of the overall rigidity offset, wherein the preset drift threshold is obtained by multiplying the modulus of the overall rigidity offset by a preset proportional coefficient.

[0174] The digital twin ear mold coordinate dynamic mapping module 15 is used to drive the coordinates of the digital twin ear mold to be updated in real time according to the overall rigidity offset and the real-time drift state vector, so as to obtain the standard anatomical acupoint coordinates corresponding to the measurement values ​​of each sensing unit.

[0175] Specifically, based on the overall rigidity offset and the real-time drift state vector, the coordinates of the digital twin earmold are updated in real time to obtain the standard anatomical acupoint coordinates corresponding to the measurement values ​​of each sensing unit, including:

[0176] Load a pre-established parametric three-dimensional digital twin model of the auricle, wherein the parametric three-dimensional digital twin model of the auricle has a coordinate system consistent with the standard anatomical coordinate reference model;

[0177] The overall rigid offset is used as the initial transformation matrix, and the real-time drift state vector is used as the dynamic compensation matrix. These are applied sequentially to the parameterized three-dimensional auricle digital twin model to obtain the dynamic calibration coordinates corresponding to the physical coordinates of each sensing unit.

[0178] The coordinates of the closest standard acupoint in the standard anatomical coordinate reference model are found based on the dynamic calibration coordinates, and these coordinates are used as the standard anatomical acupoint coordinates corresponding to the measurement value of the corresponding sensing unit at the current moment.

[0179] The multimodal dataset dynamic registration and generation module 16 is used to map the multidimensional feature data collected by each sensing unit to the standard anatomical acupoint coordinates to generate a dynamically registered auricular acupoint health dataset.

[0180] Specifically, the multidimensional feature data collected by each sensing unit is mapped to the coordinates of the standard anatomical acupoints to generate a dynamically registered auricular acupoint health dataset, including:

[0181] Acquire multidimensional feature data collected by each sensing unit, including skin potential difference in passive mode and multi-frequency bioelectrical impedance amplitude and phase angle in active mode;

[0182] The multidimensional feature data is assigned to the corresponding standard acupoints according to the standard anatomical acupoint coordinates corresponding to each sensing unit;

[0183] When the same standard acupoint is mapped by multiple sensing units, the weighted average value is calculated using the contact pressure value of each sensing unit as the weight, and this average value is used as the final feature value of the standard acupoint.

[0184] The final feature values ​​and timestamps of all standard acupoints are organized into a dynamically registered auricular acupoint health dataset.

[0185] This also includes:

[0186] When a user performs multiple wear detections on different dates, the historical overall rigid offset and historical real-time drift state vector corresponding to each detection are loaded respectively.

[0187] For the auricular health dataset obtained from each detection, based on the overall rigid offset and the inverse transformation of the real-time drift state vector of that detection, all feature values ​​are uniformly mapped to the same benchmark digital twin auricular coordinate system.

[0188] Under the reference digital twin ear mold coordinate system, the changing trends of multidimensional feature values ​​of the same standard acupoint on different dates are compared to generate a long-term dynamic health analysis report.

[0189] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0190] This specification and accompanying drawings are merely illustrative examples of the invention and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of the invention. Clearly, those skilled in the art can make various alterations and modifications to the invention without departing from its scope. Therefore, if such modifications and modifications fall within the scope of the invention and its equivalents, the invention is intended to include these modifications and modifications.

Claims

1. A method for synchronous acquisition of auricular biosignals based on multimodal sensor fusion, characterized in that, include: Multimodal contact fingerprint data synchronously collected by all sensing units on the flexible multimodal sensing earmuff is loaded, and the multimodal contact fingerprint data of each sensing unit is combined into a multimodal contact fingerprint feature vector. Based on the multimodal contact fingerprint feature vector and the standard anatomical coordinate reference model, initial wearing offset detection is performed to obtain the overall rigid offset. During the process of collecting auricular biosignals, dynamic contact fingerprint data is repeatedly collected at a fixed period. Based on the time series of the dynamic contact fingerprint data, dynamic drift real-time tracking is performed to obtain the real-time drift state vector. Calculate the magnitude of the real-time drift state vector. When the magnitude is greater than or equal to a preset drift threshold, trigger the re-execution of the initial wearing offset detection to update the overall rigidity offset. Based on the overall rigidity offset and the real-time drift state vector, the coordinates of the digital twin ear mold are updated in real time to obtain the standard anatomical acupoint coordinates corresponding to the measurement values ​​of each sensing unit. The multidimensional feature data collected by each sensing unit is mapped to the coordinates of the standard anatomical acupoints to generate a dynamically registered auricular acupoint health dataset.

2. The method for synchronous acquisition of auricular biosignals based on multimodal sensor fusion as described in claim 1, characterized in that, Multimodal contact fingerprint data synchronously collected by all sensing units on the flexible multimodal sensing ear cup is loaded, and the multimodal contact fingerprint data of each sensing unit is combined into a multimodal contact fingerprint feature vector, including: The flexible multimodal sensing earmuff is controlled to enter the contact fingerprint scanning mode, and the skin contact impedance amplitude of each sensing unit under low-frequency AC signal and high-frequency AC signal is measured in active mode respectively; Local skin temperature values ​​are read by miniature temperature sensors integrated in each sensing unit; The static contact pressure value is read by a micro-motion pressure sensor integrated in each sensing unit; The skin contact impedance amplitude, local skin temperature value, and static contact pressure value of the same sensing unit are combined to form the multimodal contact fingerprint data of that sensing unit, and the multimodal contact fingerprint data of all sensing units are arranged in a preset order to form a multimodal contact fingerprint feature vector.

3. The method for synchronous acquisition of auricular biosignals based on multimodal sensor fusion as described in claim 1, characterized in that, Based on the multimodal contact fingerprint feature vector and the standard anatomical coordinate reference model, initial wearing offset detection is performed to obtain the overall rigid offset, including: Load a pre-built standard anatomical coordinate reference model, wherein each standard acupoint coordinate point in the standard anatomical coordinate reference model is associated with a standard contact fingerprint expected range; The multimodal contact fingerprint feature vector is matched with the expected range of the standard contact fingerprint to obtain the rigid transformation parameters that minimize the overall matching error. The rigid transformation parameters include horizontal displacement, vertical displacement and rotation angle, which are set as the overall rigid offset.

4. The method for synchronous acquisition of auricular biosignals based on multimodal sensor fusion as described in claim 1, characterized in that, Based on the time series of the dynamic contact fingerprint data, perform real-time dynamic drift tracking to obtain a real-time drift state vector, including: With a preset fixed sampling period, the high-frequency impedance fluctuation, contact pressure AC component and temperature drift trend of each sensing unit are continuously collected to form a dynamic contact fingerprint vector at each sampling moment. Calculate the cross-correlation function between the dynamic contact fingerprint vectors at adjacent sampling times, and determine the overall displacement increment from the previous time to the current time based on the peak position of the cross-correlation function; The overall displacement increment is integrated over time to obtain the cumulative horizontal displacement, cumulative vertical displacement, and cumulative rotation angle, which are then combined into a real-time drift state vector.

5. The method for synchronous acquisition of auricular biosignals based on multimodal sensor fusion as described in claim 4, characterized in that, Also includes: During the real-time tracking of dynamic drift, the time integral of the standard deviation of the high-frequency impedance fluctuation of all sensing units within a preset time window is calculated and set as the contact fingerprint stability index. When the contact fingerprint stability index exceeds a preset safety threshold, it is determined that the flexible multimodal sensor earmuff has experienced continuous uncontrollable slippage, the ear acupoint biosignal acquisition is automatically paused, and a prompt to re-fit is issued through the user prompt unit.

6. The method for synchronous acquisition of auricular biosignals based on multimodal sensor fusion as described in claim 1, characterized in that, The preset drift threshold is dynamically set based on the modulus of the overall rigidity offset, wherein the preset drift threshold is obtained by multiplying the modulus of the overall rigidity offset by a preset proportional coefficient.

7. The method for synchronous acquisition of auricular biosignals based on multimodal sensor fusion as described in claim 1, characterized in that, Based on the overall rigidity offset and the real-time drift state vector, the coordinates of the digital twin earmold are updated in real time to obtain the standard anatomical acupoint coordinates corresponding to the measurement values ​​of each sensing unit, including: Load a pre-established parametric three-dimensional digital twin model of the auricle, wherein the parametric three-dimensional digital twin model of the auricle has a coordinate system consistent with the standard anatomical coordinate reference model; The overall rigid offset is used as the initial transformation matrix, and the real-time drift state vector is used as the dynamic compensation matrix. These are applied sequentially to the parameterized three-dimensional auricle digital twin model to obtain the dynamic calibration coordinates corresponding to the physical coordinates of each sensing unit. The coordinates of the closest standard acupoint in the standard anatomical coordinate reference model are found based on the dynamic calibration coordinates, and these coordinates are used as the standard anatomical acupoint coordinates corresponding to the measurement value of the corresponding sensing unit at the current moment.

8. The method for synchronous acquisition of auricular biosignals based on multimodal sensor fusion as described in claim 1, characterized in that, The multidimensional feature data collected by each sensing unit is mapped to the coordinates of the standard anatomical acupoints to generate a dynamically registered auricular acupoint health dataset, including: Acquire multidimensional feature data collected by each sensing unit, including skin potential difference in passive mode and multi-frequency bioelectrical impedance amplitude and phase angle in active mode; The multidimensional feature data is assigned to the corresponding standard acupoints according to the standard anatomical acupoint coordinates corresponding to each sensing unit; When the same standard acupoint is mapped by multiple sensing units, the weighted average value is calculated using the contact pressure value of each sensing unit as the weight, and this average value is used as the final feature value of the standard acupoint. The final feature values ​​and timestamps of all standard acupoints are organized into a dynamically registered auricular acupoint health dataset.

9. The method for synchronous acquisition of auricular biosignals based on multimodal sensor fusion as described in claim 8, characterized in that, Also includes: When a user performs multiple wear detections on different dates, the historical overall rigid offset and historical real-time drift state vector corresponding to each detection are loaded respectively. For the auricular health dataset obtained from each detection, based on the overall rigid offset and the inverse transformation of the real-time drift state vector of that detection, all feature values ​​are uniformly mapped to the same benchmark digital twin auricular coordinate system. Under the reference digital twin ear mold coordinate system, the changing trends of multidimensional feature values ​​of the same standard acupoint on different dates are compared to generate a long-term dynamic health analysis report.

10. A synchronous acquisition system for auricular biosignals based on multimodal sensor fusion, characterized in that, The method for synchronous acquisition of auricular biosignals based on multimodal sensor fusion as described in any one of claims 1 to 9 includes: The multimodal initial fingerprint acquisition and feature construction module is used to load the multimodal contact fingerprint data synchronously acquired by all sensing units on the flexible multimodal sensing ear cover, and combine the multimodal contact fingerprint data of each sensing unit into a multimodal contact fingerprint feature vector. The initial wearing offset detection module is used to perform initial wearing offset detection based on the multimodal contact fingerprint feature vector and the standard anatomical coordinate reference model to obtain the overall rigid offset. The dynamic fingerprint periodic acquisition and drift tracking module is used to repeatedly acquire dynamic contact fingerprint data at a fixed period during the acquisition of auricular biosignals, and to perform real-time dynamic drift tracking based on the time series of the dynamic contact fingerprint data to obtain a real-time drift state vector. The drift state assessment and recalibration trigger module is used to calculate the magnitude of the real-time drift state vector. When the magnitude is greater than or equal to a preset drift threshold, it triggers the re-execution of the initial wearing offset detection to update the overall rigid offset. The digital twin ear mold coordinate dynamic mapping module is used to drive the coordinates of the digital twin ear mold to be updated in real time according to the overall rigidity offset and the real-time drift state vector, so as to obtain the standard anatomical acupoint coordinates corresponding to the measurement values ​​of each sensing unit. The multimodal dataset dynamic registration and generation module is used to map the multidimensional feature data collected by each sensing unit to the coordinates of the standard anatomical acupoints, and generate a dynamically registered auricular acupoint health dataset.

Images