A multi-type blood glucose test paper recognition system

Abstract

The present application relates to the technical field of pattern recognition, in particular to a multi-type blood glucose test paper recognition system, comprising four modules of feature perception, intelligent identification, excitation execution and compensation calculation, first, collecting high-frequency impedance phase sequence, charge-discharge relaxation time constant and sample transient surge signal through intelligent sensor, using recursive graph quantitative analysis to extract coupling feature vector; then, projecting the coupling feature vector to the attribute subspace of the preset matrix library, identifying the test paper reaction mediator potential interval, enzyme layer load abundance and hematocrit non-linear coefficient, and dynamically synthesizing the target excitation waveform containing the ladder-shaped pre-polarization segment and the load modulation sampling segment; then, driving the biochemical reaction according to the identification result to capture the Faraday steady-state response current; finally, using the weight matrix to execute space charge layer correction and diffusion flux gain adjustment, and outputting the detection result. The present application realizes accurate detection of heterogeneous test paper through multi-dimensional feature identification.

Description

Technical Field

[0001] This invention relates to the field of pattern recognition technology, specifically to a multi-type blood glucose test strip recognition system. Background Technology

[0002] With the rapid popularization of mobile healthcare technology, portable blood glucose monitoring systems have become a core tool for diabetes management. To meet users' needs for consumable compatibility and ease of operation, the industry is gradually developing blood glucose test strip recognition systems that support multiple brands, specifications, and batches. This trend towards multi-type compatibility aims to reduce users' healthcare expenses and improve the flexibility of chronic disease monitoring.

[0003] In current testing procedures, systems typically require specific detection logic to be set for different types of test strips. Existing technologies often distinguish test strip models by setting fixed identification information (such as resistance codes or optical codes) on the surface of the test strip and then calling a preset static calibration curve based on the identified model. During the detection process, the system usually applies a uniform excitation signal to the test strip and captures the current data generated by the biochemical reaction to calculate the concentration.

[0004] However, in the complex application scenarios of mobile healthcare, various types of blood glucose test strips exhibit significant heterogeneity at the microscopic level. Specifically, different brands or batches of test strips show subtle differences in the physical geometry of the microchannels, the electrochemical polarization characteristics of the electrode interfaces, and the kinetic characteristics of the biochemical reaction system. Due to the dynamic changes in environmental temperature and humidity and the storage state of the test strips, these heterogeneous properties can generate complex nonlinear interferences during the detection process, making it difficult for the system to achieve consistent performance of cross-type test results through a single static processing mode when acquiring biochemical response signals.

[0005] Therefore, how to accurately identify the heterogeneous physicochemical properties of various types of blood glucose test strips through internal system sensing without relying on manual presets, and thereby achieve adaptive performance optimization of the detection process, has become a key technical problem that urgently needs to be solved in the field of mobile medical blood glucose monitoring.

[0006] Therefore, a multi-type blood glucose test strip recognition system is proposed. Summary of the Invention

[0007] The purpose of this invention is to provide a multi-type blood glucose test strip recognition system that achieves accurate detection of heterogeneous test strips through multi-dimensional feature identification. It comprises four main modules: feature perception, intelligent identification, excitation execution, and compensation calculation. First, a smart sensor collects high-frequency impedance phase sequences, charge / discharge relaxation time constants, and sample injection transient surge signals. A recursive graph is used to quantitatively analyze and extract coupling feature vectors. Then, the coupling feature vectors are projected onto the attribute subspace of a preset matrix library to identify the test strip reaction mediator potential range, enzyme layer loading abundance, and erythrocyte hematocrit nonlinear coefficient, and a target excitation waveform containing a stepped pre-polarization segment and a load-modulated sampling segment is dynamically synthesized. Next, based on the identification results, a weight matrix is ​​invoked to drive the biochemical reaction and capture the Faraday steady-state response current. Finally, the weight matrix is ​​used to perform space charge layer correction and diffusion flux gain adjustment to output the detection results.

[0008] To achieve the above objectives, the present invention provides the following technical solution: A multi-type blood glucose test strip recognition system, comprising: The feature perception module acquires the high-frequency impedance phase sequence, charge-discharge relaxation time constant, and transient surge envelope signal during the sample introduction stage collected by the intelligent sensor; it extracts the frequency feature point offset in the high-frequency impedance phase sequence and calculates the recursive graph of the transient surge envelope signal for quantitative analysis of features, generating a coupled feature vector. The intelligent identification module inputs the coupling feature vector into a preset matrix library to identify the redox potential range of the reaction mediator, the effective loading abundance of the enzyme layer, and the nonlinear coefficient of hematocrit interference of the current test paper. Based on the identified redox potential range of the reaction mediator and the effective loading abundance of the enzyme layer, the module synthesizes the target excitation waveform sequence in the preset matrix library, which includes a prepolarized voltage segment with step-like growth and a constant potential acquisition segment modulated by the loading abundance. The excitation execution module, based on the identified nonlinear coefficients, calls a weight matrix from a preset matrix library to control the intelligent sensor to drive the biochemical reaction process according to the target excitation waveform sequence and acquire the steady-state response current signal; The compensation solution module uses a weight matrix to correct the space charge layer thickness and adjust the diffusion flux gain of the steady-state response current signal, performs concentration conversion, and outputs the detection results.

[0009] Preferably, the process of the smart sensor acquiring data includes: applying a multi-frequency AC excitation voltage covering a preset frequency range, and simultaneously acquiring the phase shift value of the current relative to the voltage in the sensing circuit to construct the high-frequency impedance phase sequence; applying a transient step pulse to the sensing electrode and capturing the charge response curve of the double layer at the electrode interface, extracting the characteristic time required for the current signal to decay from the peak to a preset steady-state threshold, and determining the charge-discharge relaxation time constant; at the injection trigger moment when the blood sample enters the microchannel, recording the original current peak value generated at the sensing interface and the subsequent morphological evolution trajectory through high-frequency sampling to generate a transient surge envelope signal.

[0010] Preferably, the process of generating the coupling feature vector includes: identifying the phase extreme frequencies of the high-frequency impedance phase sequence within a preset frequency sweep interval, comparing the extreme frequencies with the stored reference material frequencies, and extracting the frequency feature point offset to characterize the polarization resistance state of the smart sensor electrode interface; performing phase space reconstruction on the transient surge envelope signal and constructing a recursive matrix, extracting quantitative descriptive indicators characterizing signal determinism, laminarity, and information entropy from the recursive matrix as quantitative analysis features of the recursive graph to characterize the wetting dynamics of the biochemical reaction layer in the microchannel; and performing weighted fusion of the frequency feature point offset and the quantitative analysis features of the recursive graph to generate the coupling feature vector characterizing the physical properties and hydrodynamic stability of the blood glucose test strip electrode.

[0011] Preferably, the process of inputting the coupling feature vector into a preset matrix library includes: projecting the coupling feature vector onto the feature space where the preset matrix library is located, and performing a similarity measurement with a pre-stored category center vector to locate the physicochemical attribute subspace to which the blood glucose test strip belongs; extracting the corresponding electrochemical reference coordinate values ​​from the located attribute subspace to identify the redox potential range of the reaction mediator; calculating the projection modulus value of the coupling feature vector in the attribute subspace, and comparing the modulus value with a preset loading level curve to map and obtain the effective loading abundance of the enzyme layer; retrieving interpolation nodes associated with the hydrodynamic stability features in the attribute subspace, and obtaining the nonlinear coefficient of the hematocrit interference by performing weighted interpolation operations between nodes.

[0012] Preferably, the process of synthesizing the target excitation waveform sequence in the preset matrix library includes: extracting the potential step slope and plateau potential peak associated with the redox potential range of the reaction mediator, and performing time-domain linear interpolation to construct the step-growing pre-polarized voltage segment; retrieving the sampling start offset and sampling integration window width corresponding to the effective load abundance of the enzyme layer, and performing time-domain feature mapping on the reference level to generate the constant potential acquisition segment modulated by the load abundance; performing phase connection between the pre-polarized voltage segment and the constant potential acquisition segment, and outputting the target excitation waveform sequence executed by the intelligent sensor hardware.

[0013] Preferably, the process of acquiring the steady-state response current signal includes: matching the identified nonlinear coefficient of the hematocrit interference with the spatial dimension coordinates in the multidimensional feature control matrix library to retrieve the dynamic weight matrix containing the diffusion gain factor and the baseline offset vector; loading the target excitation waveform sequence into the drive register of the smart sensor, and triggering the smart sensor to smoothly switch to the constant potential acquisition segment output state after the step-growing pre-polarization voltage segment is completed; during the duration of the constant potential acquisition segment, synchronously capturing the Faraday current signal of the smart sensor sensing interface according to a preset sampling frequency, and performing mean filtering processing to extract the steady-state response current signal after recognizing that the Faraday current signal has entered the stable diffusion range.

[0014] Preferably, the process of outputting the detection result includes: extracting the diffusion gain operator from the dynamic weight matrix and multiplying it with the steady-state response current signal to complete the diffusion flux gain adjustment; calling the charge layer thickness correction operator from the dynamic weight matrix and performing baseline differential processing on the adjusted steady-state response current signal to achieve the space charge layer thickness correction; mapping the corrected steady-state response current signal to the linearized calibration interval corresponding to the target test strip model, performing nonlinear mapping solution through the concentration conversion logic to obtain detection data characterizing the glucose content of the blood sample, and outputting it as the detection result.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. By introducing recursive graph quantitative analysis technology and dynamic weight allocation logic based on discriminant power, feature extraction of the depth fingerprint of transient injection signals was achieved. The deterministic and laminar flow indices in the recursive graph can effectively characterize the nonlinear dynamics of the blood wetting enzyme layer process, capturing microchannel design differences and enzyme layer wettability fluctuations that are difficult to detect with traditional time-domain analysis. Combined with an automatic weighting mechanism based on inter-class and intra-class variance analysis, an objective data foundation is laid for subsequent accurate classification.

[0016] 2. By performing feature projection within the attribute subspace and introducing an inverse distance weighted interpolation algorithm, the problem of continuity compensation in heterogeneous attribute identification is solved. Subspace projection can decouple high-dimensional features into a clear mediator system and load abundance coordinates, achieving accurate positioning of the physicochemical properties of the test strip. At the same time, by using the inverse distance weight calculation of the four nearest neighbor nodes, the nonlinear coefficient of hematocrit interference is finely calculated. This interpolation logic based on spatial geometric relationships effectively avoids the step error caused by the lookup table method, enabling the system to output a compensation operator with greater linearity and consistency when processing complex individual difference samples.

[0017] 3. By dynamically synthesizing the target excitation waveform sequence using the identified mediator potential range and load abundance, this application achieves active adaptation between the hardware driving protocol and the test strip reaction system. The step-growing pre-polarization voltage segment can perform gradient driving according to the potential requirements of specific mediators, effectively suppressing the unsteady double-layer charging current caused by voltage jumps; while the constant potential acquisition segment modulated by load abundance ensures that the current capture is always in the steady-state range of diffusion control by dynamically adjusting the sampling blind zone and integration window. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the structure of a multi-type blood glucose test strip recognition system according to the present invention; Figure 2 This is a schematic diagram of a multi-type blood glucose test strip recognition system according to the present invention; Figure 3 This is a schematic diagram of the process of inputting coupled feature vectors into a preset matrix library according to the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Please see Figures 1 to 3 This invention provides a multi-type blood glucose test strip recognition system, the technical solution of which is as follows:

[0021] Example 1: A multi-type blood glucose test strip recognition system, comprising: The feature perception module acquires the high-frequency impedance phase sequence, charge-discharge relaxation time constant, and transient surge envelope signal during the sample introduction stage collected by the intelligent sensor; it extracts the frequency feature point offset in the high-frequency impedance phase sequence and calculates the recursive graph of the transient surge envelope signal for quantitative analysis of features, generating a coupled feature vector. The intelligent identification module inputs the coupling feature vector into a preset matrix library to identify the redox potential range of the reaction mediator, the effective loading abundance of the enzyme layer, and the nonlinear coefficient of hematocrit interference of the current test paper. Based on the identified redox potential range of the reaction mediator and the effective loading abundance of the enzyme layer, the module synthesizes the target excitation waveform sequence in the preset matrix library, which includes a prepolarized voltage segment with step-like growth and a constant potential acquisition segment modulated by the loading abundance. The excitation execution module, based on the identified nonlinear coefficients, calls a weight matrix from a preset matrix library to control the intelligent sensor to drive the biochemical reaction process according to the target excitation waveform sequence and acquire the steady-state response current signal; The compensation solution module uses a weight matrix to correct the space charge layer thickness and adjust the diffusion flux gain of the steady-state response current signal, performs concentration conversion, and outputs the detection results.

[0022] Furthermore, the process of data acquisition by the intelligent sensor includes: applying a multi-frequency AC excitation voltage covering a preset frequency range, and simultaneously acquiring the phase shift value of the current relative to the voltage in the sensing circuit to construct the high-frequency impedance phase sequence; applying a transient step pulse to the sensing electrode and capturing the charge response curve of the double layer at the electrode interface, extracting the characteristic time required for the current signal to decay from the peak to a preset steady-state threshold, and determining the charge-discharge relaxation time constant; at the sample injection trigger moment when the blood sample enters the microchannel, recording the original current peak value generated at the sensing interface and the subsequent morphological evolution trajectory through high-frequency sampling to generate a transient surge envelope signal.

[0023] The system control unit generates a set of sinusoidal AC excitation voltages covering a preset frequency range of 100Hz to 100kHz through an internal signal generator. After the excitation voltage is applied to the sensing electrode of the smart sensor, the system synchronously captures the current signal in the loop through a high-speed transimpedance amplifier circuit. The bandwidth of the high-speed transimpedance amplifier circuit needs to be greater than 5 times the upper limit of the preset frequency range to reduce the interference of the circuit's own phase shift on the signal. Subsequently, using an orthogonal phase-sensitive detector or fast Fourier transform logic, the zero-crossing time difference between the excitation voltage waveform and the response current waveform is compared to calculate the phase shift value at each frequency point. During this process, the system needs to pre-acquire open-circuit and short-circuit reference signals when the sensor is not connected. The system's inherent phase shift deviation is subtracted from the phase shift value through a software algorithm to achieve phase calibration. Finally, each frequency point and its corresponding phase shift value are arranged in ascending order of frequency to construct a high-frequency impedance phase sequence that reflects the polarization characteristics of the electrode interface.

[0024] During a preset detection interval, the system applies a transient step pulse voltage with a fixed potential to the sensing electrode of the smart sensor. The amplitude of the fixed potential is set in the range of 30 mV to 300 mV to ensure that the electrode interface is in an unrestricted diffusion polarization state rather than an overpotential reaction state. Under the action of this voltage, the electrical double-layer capacitance generated at the interface between the electrode and the blood sample begins to charge. The system records the transient decay current signal generated thereafter in real time. The algorithm unit starts a high-precision timer, taking the moment when the current signal reaches the initial peak as the starting point, and monitors the decay trajectory of the current value over time. The sampling frequency is not less than 100 kHz to ensure that the nonlinear decay process within the first 500 microseconds after the initial peak can be completely captured. When the current value drops to a preset steady-state current threshold (such as 5% of the peak value or a specific current reference), the timer stops. The duration required for the current to decay from the peak to the threshold is recorded as the charge-discharge relaxation time constant, which is used to quantify the charge transfer rate on the electrode surface.

[0025] The system is in standby monitoring mode, monitoring the background current of the intelligent sensor interface in real time with microampere-level sampling accuracy. When a blood sample is drawn into the microchannel and contacts the electrode, the loop current experiences a sudden surge due to the activation of the electrochemical circuit. The system uses the moment when the current change rate exceeds a preset threshold as the sample injection trigger moment. This current change rate is obtained through differential calculation of three consecutive sampling points. A valid sample injection trigger is determined when the change rate exceeds a preset slope threshold twice consecutively to filter out random noise interference and immediately switch to high-frequency sampling mode. Simultaneously, the system reads and backtracks the current data within 10 milliseconds before the trigger moment through a preset circular buffer, combining the baseline data before triggering with the dynamic data after triggering. Next, the system acquires raw sampling points containing the complete starting rising edge. In this mode, the system continuously records raw current sampling points within a preset time (e.g., 50 milliseconds) after the sample injection trigger, fully preserving the amplitude of the current peak and the subsequent morphological evolution trajectory caused by the sample wetting the enzyme layer. Finally, smoothing and feature extraction are performed on these high-frequency sampling points to generate a transient surge envelope signal characterizing the hydrodynamic process. The smoothing process uses a five-point cubic smoothing algorithm or a Savitzky-Golay filter to remove high-frequency glitches while preserving the curvature characteristics of the current peak.

[0026] By simultaneously acquiring high-frequency impedance phase, relaxation time constant, and transient surge signals, this system can construct a deep physical fingerprint of the test strip from three dimensions: electrode polarization characteristics, charge transfer rate, and hydrodynamics. This multi-parameter coupled detection method can effectively capture the microscopic differences in material properties and flow channel design among different test strip models, providing a solid data foundation for subsequent attribute identification and nonlinear compensation. This helps improve the system's compatibility accuracy with heterogeneous consumables and the stability of detection results in complex and ever-changing mobile monitoring environments.

[0027] Further, the process of generating the coupling feature vector includes: identifying the phase extreme frequencies of the high-frequency impedance phase sequence within a preset frequency sweep interval, comparing the extreme frequencies with the stored reference material frequencies, and extracting the frequency feature point offset to characterize the polarization resistance state of the smart sensor electrode interface; performing phase space reconstruction on the transient surge envelope signal and constructing a recursive matrix, extracting quantitative descriptive indicators characterizing signal determinism, laminarity, and information entropy from the recursive matrix as quantitative analysis features of the recursive graph to characterize the wetting dynamics of the biochemical reaction layer in the microchannel; and performing weighted fusion of the frequency feature point offset and the quantitative analysis features of the recursive graph to generate the coupling feature vector characterizing the physical properties and hydrodynamic stability of the blood glucose test strip electrode.

[0028] The acquired high-frequency impedance phase sequence undergoes an extreme value step search. By comparing the phase values ​​of adjacent frequency points, the specific frequency corresponding to the local maximum or minimum phase angle is identified, i.e., the phase extreme frequency. Specifically, the extreme value step search employs a sliding window mean smoothing process, with the window width set to 3 to 7 sampling frequency points to filter out high-frequency random noise. Simultaneously, the criterion for determining a local maximum or minimum requires that the difference between the phase value at that point and the two step points before and after it exceeds a preset significance threshold (e.g., 0.1 degrees), ensuring that the identified phase extreme frequency has physical representational significance. Subsequently, the system retrieves the relevant information from non-volatile memory. A reference material frequency constant matched to the temperature is used. The extreme frequency identified in real time is subtracted from this reference frequency, and the difference is the frequency characteristic point offset. This offset is directly related to the electric double-layer capacitance effect and polarization resistance of the electrode interface. The non-volatile memory stores a pre-calibrated temperature-frequency mapping table. The system reads the real-time value of the built-in temperature sensor, uses linear interpolation to retrieve and calculate the reference material frequency at the current temperature from the mapping table, and compensates for the polarization frequency drift caused by temperature fluctuations. This reflects the micro-roughness of the electrode surface or the degree of material aging, thereby completing the physical characterization of the polarization state of the sensor electrode interface.

[0029] For the transient surge envelope signal, the embedding dimension and delay time parameters are first determined. The one-dimensional current amplitude sequence is mapped to a multi-dimensional phase space to perform phase space reconstruction. The embedding dimension is set to a range of 3 to 6, and the delay time is determined based on the time interval when the autocorrelation function first drops to the original value 1 / e, to ensure that the phase space trajectory can be fully expanded in the topological structure without overlap. Simultaneously, the preset neighborhood threshold is set to 10% to 20% of the maximum diameter of the reconstructed trajectory in the phase space to filter out subtle digital quantization noise and retain macroscopic recursive features. In the reconstructed spatial trajectory, the system calculates the Euclidean distance between any two time-point vectors and compares it with the preset neighborhood threshold. If the distance is less than the threshold, the corresponding coordinate point in the recursion matrix is ​​marked as a recursive state. Subsequently, the phase space is reconstructed by scanning... Quantitative analysis is performed on the geometric structure in the recursive matrix: the distribution of straight line segments parallel to the main diagonal is statistically analyzed, and the proportion of diagonal points to the total number of recursive points is calculated to obtain signal deterministic indicators. During quantitative analysis, only continuous straight line segments with a length exceeding a preset minimum line segment threshold (e.g., 2 or 3 pixels) are statistically analyzed to exclude false recursive points caused by the coincidence of phase trajectory tangents, ensuring that the extracted deterministic and laminar flow indicators truly reflect the fluid steady-state characteristics of the biochemical reaction layer. The length of straight line segments in the vertical direction is statistically analyzed to obtain laminar flow indicators characterizing the persistence of the signal state. Finally, the Shannon entropy algorithm is applied to calculate the information content of the distribution frequency of diagonal lengths to obtain information entropy indicators reflecting the complexity of the signal. These indicators are integrated into the quantitative analysis features of the recursive graph to characterize the nonlinear dynamic state of the blood sample wetting enzyme layer process.

[0030] The system first performs dimensional normalization on the extracted frequency feature point offsets and the quantitative analysis features of the recursive graph, including determinism, laminar flow, and information entropy, compressing their values ​​to a uniform floating-point range (e.g., between 0 and 1). Then, according to a preset feature weight configuration file, weight coefficients are assigned to feature components of different dimensions. The processed feature components are then combined in a preset topological order using a vector concatenation operator to form a multi-dimensional one-dimensional array. This array is the coupled feature vector, which deeply integrates the static polarization properties of the electrode with the dynamic wetting characteristics of the biochemical reaction in its data structure, serving as a standardized data source for subsequent input to the attribute identification module.

[0031] The allocation of weighting coefficients for feature components of different dimensions is specifically determined by performing dynamic gain analysis based on feature discriminative power to objectively determine the weighting coefficients for each dimension, rather than using manually preset values. Specifically, the system utilizes a pre-stored dataset of known test strip samples to calculate the inter-class variance between different brands and models of test strips, and the intra-class variance between different batches and aging states of the same model of test strips, for each dimension feature (such as frequency offset, information entropy, etc.). Subsequently, the ratio of the inter-class variance to the intra-class variance of each dimension feature is defined as the discriminative sensitivity index of that feature, used to quantify the contribution of that feature to distinguishing the heterogeneous properties of the test strips. The system performs normalization mapping on the discriminative sensitivity indices of all dimensions, that is, calculates the percentage value of each dimension index relative to the sum of all dimension indices, and automatically determines the proportion of each index in the overall sensitivity distribution as the weighting coefficient of the corresponding dimension. This automatic weighting mechanism based on the statistical distribution law of the samples ensures that the construction of the coupled feature vector is driven by the objective differences in the physicochemical properties of the test strips, eliminating the impact of human intervention on recognition accuracy.

[0032] By coupling high-frequency impedance phase shift with recursive graph dynamics, this application achieves in-depth characterization of the physical properties of the test strip and the wetting characteristics of the biochemical reaction layer. This multi-dimensional feature extraction method effectively mines heterogeneous fingerprints hidden in transient signals. Combined with discriminative weight allocation logic, it lays a solid data foundation for the subsequent adaptive synthesis of the excitation protocol and accurate compensation of detection results.

[0033] Further, the process of inputting the coupling feature vector into a preset matrix library includes: projecting the coupling feature vector onto the feature space where the preset matrix library is located, and performing a similarity measurement with the pre-stored category center vector to locate the physicochemical attribute subspace to which the blood glucose test strip belongs; extracting the corresponding electrochemical reference coordinate value from the located attribute subspace to identify the redox potential range of the reaction mediator; calculating the projection modulus value of the coupling feature vector in the attribute subspace, and comparing the modulus value with a preset loading level curve to map and obtain the effective loading abundance of the enzyme layer; retrieving interpolation nodes associated with the hydrodynamic stability feature in the attribute subspace, and obtaining the nonlinear coefficient of the hematocrit interference by performing weighted interpolation operations between nodes.

[0034] First, a spatial transformation matrix is ​​retrieved from a preset matrix library. The input coupled feature vector is then multiplied with this matrix, projecting it into a multidimensional high-order feature space. Specifically, the spatial transformation matrix is ​​a feature vector projection matrix obtained by performing principal component analysis or linear discriminant analysis on the feature vector set of known model samples. This matrix maps the original high-dimensional feature vectors to an orthogonal feature subspace of a preset dimension (e.g., 3D or 5D), achieving data dimensionality reduction and decoupling of key attribute features. Within the projection space, the algorithm calculates the Euclidean distance between the projected vector and the center vectors of each known test strip category stored in the library to perform a similarity measurement. The center vectors of the categories are the arithmetic mean of the feature vectors extracted from multiple samples of each brand and model of test strip under standard experimental conditions. The system presets a threshold; if the calculated minimum Euclidean distance exceeds this threshold, the current test strip is determined to be an unknown model or an illegal consumable, thus ensuring the exclusivity of the positioning result. The system identifies the category center vector with the smallest distance and locks the current test paper in the corresponding physicochemical property subspace accordingly, thereby completing the coarse-grained positioning of the test paper's underlying material system.

[0035] After locating a specific attribute subspace, the algorithm further retrieves pre-stored electrochemical reference coordinate values ​​from the metadata of that subspace. These reference coordinate values ​​specifically represent the characteristic components in the attribute subspace related to the standard redox potential of the reaction mediator, characterizing the critical voltage required for electron transfer in a specific mediator system. These reference coordinate values ​​represent the characteristic potential characteristics of the specific material system under standard conditions. The system matches these coordinate values ​​with a preset potential range mapping table to identify the redox potential range corresponding to the current reaction mediator (e.g., identifying it as a low-potential region or a high-potential region). This identification result serves as the basic level reference for the waveform sequence synthesized by the subsequent excitation execution module, ensuring that the excitation voltage accurately covers the characteristic reaction window of the mediator.

[0036] To quantify the active substance density of the enzyme layer, the system calculates the scalar projection length of the coupled feature vector along the principal axis of the attribute subspace, i.e., the projection modulus value. Subsequently, the algorithm calls the load level response curve dedicated to this subspace. This load level response curve is constructed using a third-order polynomial fitting, with the horizontal axis representing the normalized projection modulus and the vertical axis representing enzyme activity units. By substituting the real-time projection modulus into the corresponding polynomial equation and solving it, the current continuous load abundance value is obtained. This curve predefines the nonlinear relationship between signal intensity and enzyme layer load. By substituting the calculated projection modulus value into this response curve and performing point-to-point comparison and interpolation, the system maps and outputs the current effective enzyme layer load abundance level. This parameter reflects the load difference of the test strip during production or the degree of aging after storage, and is used to guide the modulation depth of the voltage in subsequent acquisition segments.

[0037] To address the hematocrit interference caused by blood sample viscosity, the system retrieves discrete interpolation nodes associated with hydrodynamic stability characteristics (i.e., quantitative indicators of the recursive graph) within the attribute subspace. Each interpolation node has a preset HCT interference correction benchmark under a specific physical stability. The algorithm unit locates several neighboring interpolation nodes surrounding the current feature point and calculates inverse weights based on the spatial distance between the feature point and each node. By performing weighted interpolation operations between nodes, the system calculates the nonlinear coefficient of hematocrit interference for the current sample. This coefficient characterizes the specific impact of blood viscosity on diffusion flux and provides a correction operator for the final concentration compensation. In specific execution, the system uses an inverse distance weighted interpolation algorithm. First, it searches for the four discrete interpolation nodes closest to the current feature point. Based on the Euclidean distance between the feature point and each node in the attribute subspace, it calculates the normalized weight coefficient of each node and performs a weighted summation on the preset correction benchmark of each node to obtain a high-precision nonlinear coefficient.

[0038] By performing feature projection and similarity measurement in the attribute subspace, accurate identification of the physical properties of the test strip is achieved. The projection modulus is used to map the abundance of the test strip, and an interpolation algorithm based on inverse distance weighting is introduced to solve the nonlinear coefficients. This effectively solves the coupling problem between enzyme activity differences and blood viscosity interference in different models and aging states, ensuring the targeted nature of the excitation protocol synthesis.

[0039] Furthermore, the process of synthesizing the target excitation waveform sequence in the preset matrix library includes: extracting the potential step slope and plateau potential peak associated with the redox potential range of the reaction mediator, and performing time-domain linear interpolation to construct the step-like increasing pre-polarized voltage segment; retrieving the sampling start offset and sampling integration window width corresponding to the effective load abundance of the enzyme layer, and performing time-domain feature mapping on the reference level to generate the constant potential acquisition segment modulated by the load abundance; performing phase connection between the pre-polarized voltage segment and the constant potential acquisition segment, and outputting the target excitation waveform sequence executed by the intelligent sensor hardware.

[0040] Based on the redox potential range of the reaction mediator identified in the previous step, the corresponding potential step slope (in volts per second) and plateau potential peak (in millivolts) are extracted from the parameter index table of the preset matrix library. Taking the sample injection trigger signal as the zero point of time, the voltage increment required for each centrifugal step is calculated according to the update frequency of the digital-to-analog converter. By performing time-domain linear interpolation between the initial potential and the plateau potential peak, the system generates a series of digital pulse commands with consistent step heights that increase linearly with time. These command sequences combine to form a step-like pre-polarization voltage segment, the slope of which and the endpoint potential are precisely matched with the electrochemical polarization characteristics of the specific mediator of the test strip. Specifically, the initial potential is set to the real-time open-circuit potential measured by the smart sensor or a preset zero-level reference. At the same time, the voltage increment of the centrifugal step is limited to between 5mV and 20mV to ensure that the current response is always in a quasi-steady-state cycle during the pre-polarization process, avoiding nonlinear polarization caused by large potential abrupt changes.

[0041] For the identified enzyme layer effective load abundance level, the system retrieves the associated sampling time-domain control parameters from a preset matrix library, including the sampling start offset and the sampling integration window width. The system locks the acquisition level at a preset constant reaction potential and constructs a time axis mapping based on this level. According to the sampling start offset, a blind zone time is set after the potential enters the constant plateau, during which no data acquisition is performed. The duration of this blind zone time is directly or inversely proportional to the enzyme load. In this embodiment, the blind zone time is directly proportional to the enzyme layer effective load abundance, that is, the higher the identified load abundance, the longer the set blind zone time (e.g., adjusted within the range of 0.5 to 1.5 seconds), to ensure that before sampling is initiated, The initial unsteady Faraday current generated by high-concentration enzyme catalysis has decayed to the diffusion control region that conforms to the Cottrell equation. Subsequently, the duration of continuous sampling is set according to the sampling integration window width. Within the sampling integration window width, the system initiates high-frequency sampling at a frequency of 1kHz to 10kHz to obtain a series of raw current discrete points. Then, it performs arithmetic mean or median filtering to convert the charge integral value within the window into a single-point steady-state response current value, so as to filter out the 50Hz / 60Hz power frequency interference commonly found in mobile medical devices. Through this time-domain feature mapping, the system generates a constant potential acquisition segment whose position and length on the time axis are dynamically modulated by the load abundance.

[0042] After completing the parameter calculations for the two voltage signal segments, the system performs phase transition processing between the pre-polarized voltage segment and the constant potential acquisition segment. The algorithm unit, through synchronous clock control, ensures seamless time-domain alignment between the voltage at the last step point of the pre-polarized segment and the starting level of the acquisition segment, preventing transient voltage jumps during potential switching. Specifically, the termination potential of the pre-polarized voltage segment is set within ±10mV of the reference level of the constant potential acquisition segment. If the calculated theoretical difference exceeds this range, the system automatically inserts a set of transition step points at the transition point, utilizing a small voltage gradient within 10 milliseconds to achieve a smooth transition, thereby suppressing baseline drift caused by double-layer recharging. The synthesized complete waveform data, in the form of discrete voltage value points, is sequentially loaded into the hardware driver register of the smart sensor or directly stored in the buffer memory of the digital-to-analog converter. Finally, according to the preset hardware timer interrupt frequency, the hardware timer interrupt is strictly synchronized with the write clock of the digital-to-analog converter (DAC). Whenever the timer triggers an interrupt, the system pushes the next voltage value into the DAC buffer via direct memory access, ensuring that the stepping accuracy of the excitation voltage waveform in the time domain reaches the microsecond level, guaranteeing the high repeatability of the biochemical reaction driving process. These digital sequences are converted into analog electrical signals in real time and output to the sensing electrodes, thereby completing the controlled driving of the biochemical reaction process.

[0043] By dynamically synthesizing excitation waveforms based on identified reaction mediators and loading abundance, precise potential matching and time-domain sampling optimization for heterogeneous test strips were achieved. The stepped pre-polarization section effectively suppressed transient impacts of the charging current, while the modulated sampling window precisely avoided unsteady-state interference in the early stages of the reaction. This adaptive driving mechanism improved the signal-to-noise ratio and waveform stability of the sensor signal.

[0044] Further, the process of acquiring the steady-state response current signal includes: matching the nonlinear coefficient of the identified hematocrit interference with the spatial dimension coordinates in the multidimensional feature control matrix library to retrieve the dynamic weight matrix containing the diffusion gain factor and the baseline offset vector; loading the target excitation waveform sequence into the drive register of the smart sensor, and triggering the smart sensor to smoothly switch to the output state of the constant potential acquisition segment after the prepolarization voltage segment with step-like growth is completed; during the duration of the constant potential acquisition segment, synchronously capturing the Faraday current signal of the sensing interface of the smart sensor according to a preset sampling frequency, and performing mean filtering processing to extract the steady-state response current signal after recognizing that the Faraday current signal has entered the stable diffusion range.

[0045] The system acquires the hematocrit interference nonlinear coefficients identified in the previous steps and inputs them as index vectors into a multidimensional feature control matrix library. This library employs a multidimensional grid storage structure, with each dimension's coordinates corresponding to different nonlinear coefficient ranges, test strip types, and ambient temperatures. By performing nearest neighbor search or multidimensional linear mapping, the system locates the target coordinate point in the matrix library space and extracts the dynamic weight matrix stored at that point. This matrix contains two sets of core parameters: a baseline offset vector for correcting electromagnetic interference and background noise, and a diffusion gain factor for compensating for diffusion rate deviations caused by blood viscosity, providing a compensation benchmark for subsequent fine-tuning of the signal. The multidimensional grid's step size in the temperature dimension is set to 1°C to 5°C, and the step size in the nonlinear coefficient dimension is divided into 10 to 50 equal parts based on the full range of the nonlinear coefficients. When the index vector falls between grid points, the system executes a multidimensional linear interpolation algorithm, using the weight parameters of four or eight adjacent grid points to calculate the current real-time dynamic weight matrix, ensuring the continuous and smooth evolution of compensation parameters with environmental and sample attributes.

[0046] The synthesized target excitation waveform sequence is written into the driver register of the smart sensor in the form of a numerical array. During the drive execution phase, the hardware state machine sequentially triggers the digital-to-analog converter (DAC) to output a pre-polarization voltage that increases in a stepped manner according to a preset clock frequency. When the timer counter of the pre-polarization voltage segment reaches a preset termination value, the hardware level immediately triggers a switch from step mode to constant potential output mode. To achieve a smooth switch, the system maintains the voltage continuity of the DAC output buffer to ensure that the rate of change of the potential difference at the moment of switching is within a preset slope threshold. The slope threshold is limited to a voltage change of no more than 1mV per microsecond. Specifically, this is achieved by inserting a set of micro-stepping instruction sequences into the driver register. That is, within a 1-millisecond transition period after the end of the pre-polarization segment, the potential is gradually adjusted to the reference level of the constant potential acquisition segment in steps of no more than 0.5mV. This allows the DAC output buffer to maintain charge balance in the analog domain, suppressing transient glitches with amplitudes exceeding 100nA, thereby effectively suppressing non-steady double-layer charging current glitches caused by voltage steps.

[0047] After entering the constant potential acquisition segment, the system starts the analog-to-digital converter (ADC) to synchronously capture the Faraday current signal of the sensing interface at a preset sampling frequency (e.g., 1kHz). It monitors the absolute value of the slope of the current signal changing with time in real time. When the absolute value of the slope is lower than the preset convergence threshold for multiple consecutive sampling points, it is determined that the Faraday current has entered the stable diffusion region. The convergence threshold is set to the absolute value of the rate of change of the current with time being less than 10nA / s. The sliding window first-order difference operation is performed. When the rate of change of 50 consecutive sampling points (corresponding to 50 milliseconds) is within the range of this threshold, the logic determination of entering the stable diffusion region is triggered. If convergence is not detected within the preset maximum waiting time (e.g., 3 seconds), the system automatically selects the mean value of the last 200 milliseconds of the constant potential acquisition segment as the baseline observation value and outputs an abnormal diffusion status marker. Subsequently, the system selects a sampling window of fixed length (e.g., the last 500 milliseconds) within the stable diffusion interval and performs mean filtering on all discrete sampling points within the window. The current amplitude obtained after filtering out random high-frequency noise is extracted as the steady-state response current signal reflecting the core characteristics of the sample concentration. The sampling window length of the mean filtering (e.g., 500 milliseconds) is set to an integer multiple of the local power frequency period (20 milliseconds) to achieve coherent cancellation of power frequency electromagnetic interference. At the same time, before performing the mean calculation, abnormal sampling points that deviate from the median by more than 3 times the standard deviation within the window are removed to eliminate random pulse interference caused by human body micro-movements and ensure that the signal-to-noise ratio of the steady-state response current signal is better than 40dB.

[0048] By matching nonlinear coefficients to retrieve the dynamic weight matrix and combining smooth switching of the excitation waveform with steady-state current capture, effective suppression of noise and blood viscosity interference is achieved. The smooth driving mechanism eliminates transient current interference caused by potential abrupt changes, while the slope-converged sampling logic ensures the objectivity of signal extraction.

[0049] Furthermore, the process of outputting the detection result includes: extracting the diffusion gain operator from the dynamic weight matrix and multiplying it with the steady-state response current signal to complete the diffusion flux gain adjustment; calling the charge layer thickness correction operator from the dynamic weight matrix and performing baseline differential processing on the adjusted steady-state response current signal to achieve the space charge layer thickness correction; mapping the corrected steady-state response current signal to the linearized calibration interval corresponding to the target test strip model, performing nonlinear mapping solution through the concentration conversion logic to obtain detection data characterizing the glucose content of the blood sample, and outputting it as the detection result.

[0050] The diffusion gain operator is extracted from the determined dynamic weight matrix. This operator is a compensation coefficient calculated for the current level of hematocrit interference and is used to correct the deviation in molecular diffusion rate caused by differences in blood viscosity. The algorithm unit takes the captured steady-state response current signal as input and performs arithmetic multiplication in the floating-point unit of the microprocessor. That is, it uses the diffusion gain operator to scale the original current amplitude. Specifically, the diffusion gain operator is a dimensionless normalization coefficient, the value of which is obtained by retrieving the current hematocrit nonlinear coefficient from a preset gain mapping table. When the hematocrit is higher than the reference value (e.g., 45%), resulting in increased viscosity, the operator is greater than 1, and the weakened diffusion current is compensated by multiplication. Conversely, it is less than 1 to ensure that the response current under different viscosities can be equivalently restored to the reference diffusion level under the standard viscosity. Through this step, the system standardizes the current signal under different viscosity environments to the reference diffusion level, thereby eliminating the systematic error caused by differences in the physical properties of the samples.

[0051] After gain adjustment, the system invokes the charge layer thickness correction operator in the dynamic weight matrix. This operator characterizes the non-Radidatic background current caused by the micro-morphology of the electrode surface or material aging, i.e., the crowding-out effect of the space charge layer on the effective electrode area. The algorithm unit performs differential processing on the gain-adjusted current signal and the correction operator, i.e., subtracts the corresponding baseline offset vector from the signal. The baseline offset vector is stored in the bias column of the dynamic weight matrix, and its value is finely adjusted based on the charging current decay rate measured by the test strip during the pre-polarization stage. During differential processing, the system extracts the corresponding bias component from the vector sequence and subtracts it based on the current actual sampling time point, thereby dynamically eliminating the non-Radidatic current background caused by changes in the active area of ​​the electrode surface, so that the corrected current signal purely characterizes the electron transfer rate of enzyme-catalyzed oxidation-reduction. This process can accurately remove the interference components generated by static polarization at the electrode interface, extract the pure effective electrochemical response signal generated by the enzyme-catalyzed reaction, and realize real-time compensation for the underlying physical defects of the sensor.

[0052] Based on the blood glucose test strip model identified by the preceding module, the system retrieves and loads the corresponding linearization calibration interval parameters (such as slope and intercept) from non-volatile memory. The final current signal, after diffusion gain and charge layer correction, is mapped to the coordinate system of this calibration interval. The linearization calibration interval contains multiple sets of linear regression coefficients for different concentration ranges (such as high, medium, and low blood glucose ranges). The system first compares the corrected current signal with a preset range switching threshold and automatically selects the slope and intercept parameters that match the current current intensity to avoid nonlinear distortion caused by a single linear model over a wide concentration range. Subsequently, the concentration conversion logic performs polynomial fitting calculations. During the calculation process, data verification is performed simultaneously: if the calculated raw concentration value exceeds the preset medical effective threshold range (e.g., 0.6 to 33.3 mmol / L), the system will trigger a recalculation mechanism or output an over-range error code to prevent clinical misdiagnosis due to occasional sensor hardware failures. This ensures that the final output detection data has medical-grade safety and reference value, converting the physical domain current value into the physiological domain glucose concentration value (e.g., in mmol / L or mg / dL). Finally, the system formats the calculated detection data and outputs the final detection result through the display interface or communication module, completing the closed-loop conversion from the underlying physical signal to medical-grade diagnostic data.

[0053] By extracting the diffusion gain and charge layer correction operator from the dynamic weight matrix, this stage achieves deep decoupling between blood viscosity interference and electrode background current. Diffusion gain adjustment effectively compensates for rate deviations caused by hematocrit, while baseline differential processing eliminates interference components caused by material aging, ensuring the Faraday purity of the response signal. Combined with multi-segment linearization calibration and abnormal threshold verification, the system transforms heterogeneous physical signals into highly reliable clinical data.

[0054] By deeply fusing high-frequency impedance phase sequences and recursive graph dynamics, this approach achieves precise characterization of the physical properties and biochemical reaction states of test strips, effectively overcoming accuracy bottlenecks caused by brand differences, enzyme layer aging, and hematocrit interference in traditional testing. Combined with adaptive stepped excitation waveforms and nonlinear correction based on dynamic weight matrices, this solution significantly improves the system's compatibility with various types and complex test strip states, as well as the repeatability of test results. Ultimately, this collaborative architecture, from low-level signal-driven to high-level algorithm compensation, provides robust technical support for the clinical-grade accuracy of mobile medical devices in diverse consumable environments.

[0055] Example 2: When the user inserts the test strip and drips blood, the smart sensor first initiates a multi-frequency sweep sequence. The feature sensing module captures the unique high-frequency impedance phase characteristics of the gold electrode interface. Due to the significant difference in polarization resistance between the gold electrode and the common carbon electrode, the system records a clear shift in frequency characteristic points. Simultaneously, considering the slow spreading of high-viscosity blood in the microchannel, the system records a transient surge envelope signal with a specific attenuation curvature and extracts "deterministic" and "laminar" indicators characterizing this fluid dynamic process through recursive graph analysis.

[0056] The system inputs the generated coupling feature vector into a pre-defined matrix library. Through similarity measurement, the algorithm accurately locates the test strip within the attribute subspace of "noble metal electrode-ruthenium mediator system." Within this subspace, the system further identifies the hematocrit interference nonlinearity coefficient of the current blood sample based on recursive graph features and finds that due to differences in test strip manufacturing processes, its enzyme layer loading abundance is at a moderate level. At this point, the system has completed a deep modeling of the consumable's identity and the sample's interference state.

[0057] Based on the identified ruthenium mediator redox range, the excitation execution module retrieves targeted potential step parameters from the matrix library and synthesizes a pre-polarization voltage segment with a gentle slope to accommodate the sensitive charge transfer characteristics of the gold electrode. For samples with moderate enzyme loading and high viscosity, the system automatically extends the "sampling blind zone time" before the constant potential acquisition segment to allow the biochemical reaction to completely overcome the diffusion limitations of high-viscosity blood. Subsequently, the hardware drive unit drives the sensor according to this dynamically generated waveform sequence to ensure that the Faraday current captured in the acquisition segment is the one that has truly entered the steady-state diffusion period.

[0058] After acquiring the steady-state response current signal, the compensation calculation module extracts the diffusion gain operator from the corresponding dynamic weight matrix. This operator performs gain amplification adjustment on the current signal to address the slowed diffusion rate caused by high HCT. At the same time, the charge layer thickness correction operator is used to eliminate residual charge interference generated on the gold electrode surface during the pre-polarization stage. Finally, the corrected signal is mapped to the linearization range specific to this test strip model to calculate an accurate blood glucose concentration value. Even under extreme conditions such as high blood viscosity and heterogeneous electrode materials, the system still outputs clinical-grade test results after eliminating physical interference.

[0059] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A multi-type blood glucose test strip recognition system, characterized in that, include: The feature perception module acquires the high-frequency impedance phase sequence, charge-discharge relaxation time constant, and transient surge envelope signal during the sample introduction stage collected by the intelligent sensor; it extracts the frequency feature point offset in the high-frequency impedance phase sequence and calculates the recursive graph of the transient surge envelope signal for quantitative analysis of features, generating a coupled feature vector. The intelligent identification module inputs the coupling feature vector into a preset matrix library to identify the redox potential range of the reaction mediator, the effective loading abundance of the enzyme layer, and the nonlinear coefficient of hematocrit interference of the current test paper. Based on the identified redox potential range of the reaction mediator and the effective loading abundance of the enzyme layer, the module synthesizes the target excitation waveform sequence in the preset matrix library, which includes a prepolarized voltage segment with step-like growth and a constant potential acquisition segment modulated by the loading abundance. The excitation execution module, based on the identified nonlinear coefficients, calls a weight matrix from a preset matrix library to control the intelligent sensor to drive the biochemical reaction process according to the target excitation waveform sequence and acquire the steady-state response current signal; The compensation solution module uses a weight matrix to correct the space charge layer thickness and adjust the diffusion flux gain of the steady-state response current signal, performs concentration conversion, and outputs the detection results.

2. The multi-type blood glucose test strip recognition system according to claim 1, characterized in that, The process of data acquisition by the intelligent sensor includes: applying a multi-frequency AC excitation voltage covering a preset frequency range and simultaneously acquiring the phase shift value of the current relative to the voltage in the sensing circuit to construct the high-frequency impedance phase sequence; applying a transient step pulse to the sensing electrode and capturing the charge response curve of the double layer at the electrode interface, extracting the characteristic time required for the current signal to decay from the peak to a preset steady-state threshold, and determining the charge-discharge relaxation time constant; at the injection trigger moment when the blood sample enters the microchannel, recording the original current peak value generated at the sensing interface and the subsequent morphological evolution trajectory through high-frequency sampling to generate a transient surge envelope signal.

3. The multi-type blood glucose test strip recognition system according to claim 1, characterized in that, The process of generating the coupling feature vector includes: identifying the phase extreme frequencies of the high-frequency impedance phase sequence within a preset frequency sweep interval, comparing the extreme frequencies with the stored reference material frequencies, and extracting the frequency feature point offset to characterize the polarization resistance state of the smart sensor electrode interface; performing phase space reconstruction on the transient surge envelope signal and constructing a recursive matrix, extracting quantitative descriptive indicators characterizing signal determinism, laminarity, and information entropy from the recursive matrix as quantitative analysis features of the recursive graph to characterize the wetting dynamics of the biochemical reaction layer within the microchannel; and performing weighted fusion of the frequency feature point offset and the quantitative analysis features of the recursive graph to generate the coupling feature vector characterizing the physical properties and hydrodynamic stability of the blood glucose test strip electrode.

4. The multi-type blood glucose test strip recognition system according to claim 1, characterized in that, The process of inputting the coupled feature vector into a preset matrix library includes: projecting the coupled feature vector onto the feature space where the preset matrix library is located, and performing a similarity measurement with the pre-stored category center vector to locate the physicochemical attribute subspace to which the blood glucose test strip belongs; extracting the corresponding electrochemical reference coordinate value from the located attribute subspace to identify the redox potential range of the reaction mediator; calculating the projection modulus value of the coupled feature vector in the attribute subspace, and comparing the modulus value with a preset loading level curve to map and obtain the effective loading abundance of the enzyme layer; retrieving interpolation nodes associated with the hydrodynamic stability feature in the attribute subspace, and obtaining the nonlinear coefficient of the hematocrit interference by performing weighted interpolation operations between nodes.

5. The multi-type blood glucose test strip recognition system according to claim 1, characterized in that, The process of synthesizing the target excitation waveform sequence in the preset matrix library includes: extracting the potential step slope and plateau potential peak associated with the redox potential range of the reaction mediator, and performing time-domain linear interpolation to construct the step-like increasing pre-polarization voltage segment; retrieving the sampling start offset and sampling integration window width corresponding to the effective load abundance of the enzyme layer, and performing time-domain feature mapping on the reference level to generate the constant potential acquisition segment modulated by the load abundance; performing phase connection between the pre-polarization voltage segment and the constant potential acquisition segment, and outputting the target excitation waveform sequence executed by the intelligent sensor hardware.

6. The multi-type blood glucose test strip recognition system according to claim 1, characterized in that, The process of acquiring the steady-state response current signal includes: matching the nonlinear coefficient of the identified hematocrit interference with the spatial dimension coordinates in the multidimensional feature control matrix library to retrieve the dynamic weight matrix containing the diffusion gain factor and the baseline offset vector; loading the target excitation waveform sequence into the drive register of the smart sensor, and triggering the smart sensor to smoothly switch to the output state of the constant potential acquisition segment after the step-growing pre-polarization voltage segment is completed; during the duration of the constant potential acquisition segment, synchronously capturing the Faraday current signal of the smart sensor sensing interface according to a preset sampling frequency, and performing mean filtering processing to extract the steady-state response current signal after recognizing that the Faraday current signal has entered the stable diffusion range.

7. The multi-type blood glucose test strip recognition system according to claim 1, characterized in that, The process of outputting the detection result includes: extracting the diffusion gain operator from the dynamic weight matrix and multiplying it with the steady-state response current signal to complete the diffusion flux gain adjustment; calling the charge layer thickness correction operator from the dynamic weight matrix and performing baseline differential processing on the adjusted steady-state response current signal to achieve the space charge layer thickness correction; mapping the corrected steady-state response current signal to the linearized calibration interval corresponding to the target test strip model, performing nonlinear mapping solution through the concentration conversion logic to obtain detection data characterizing the glucose content of the blood sample, and outputting it as the detection result.

Images