A non-invasive hemoglobin detection system and method based on multi-wavelength array dynamic decoupling

Abstract

The present application relates to the technical field of non-invasive physiological parameter detection, and particularly relates to a non-invasive hemoglobin detection system and method based on multi-wavelength array dynamic decoupling, which can effectively solve the problems of channel quality stratification, common optical disturbance correction and current window result output gating in the multi-wavelength finger transmission collection process, and comprises a detection main body, a flip clamping piece, a finger supporting part, a multi-wavelength light source assembly and a main control collection circuit provided with a photodetector and an algorithm processing unit. The main control collection circuit sequentially performs dark field collection, low-power sweep and formal power pre-scan before formal collection, so as to establish the access state and reference baseline of each wavelength channel, and drives the multi-wavelength light source assembly to collect multi-channel photoelectric signals according to the access state. The algorithm processing unit determines a reference channel set according to the access state and channel quality information in the formal collection, and the reference channel set contains at least two stable reference channels.

Description

Technical Field

[0001] This invention relates to the field of non-invasive physiological parameter detection technology, specifically to a non-invasive hemoglobin detection system and method based on dynamic decoupling of a multi-wavelength array. Background Technology

[0002] The detection of hemoglobin-related parameters typically requires obtaining the response information of the tested tissue to light of different wavelengths. When using a finger detection structure for non-invasive detection, a multi-wavelength light source assembly can sequentially provide detection light of different center wavelengths to the finger detection area. Photodetectors positioned on opposite sides receive the light signals transmitted through the tissue, and the main control acquisition circuit obtains photoelectric signals from multiple wavelength channels accordingly. This acquisition process is influenced by the placement of the finger, the degree of local pressure, tissue thickness, skin condition, the coupling state between the light source and the detector, ambient light, and differences in the response of each channel.

[0003] During a single acquisition, some wavelength channels may experience DC baseline drift, saturation, AC component attenuation, waveform breakpoints, phase shifts, or a decrease in signal-to-noise ratio. If the current state of these channels is not distinguished and all channel data are directly used as the estimation input for the same batch, local optical coupling changes may be mixed with target characteristics. The presence of a pulse waveform in a single channel alone cannot determine whether that channel is suitable as a reference for common changes within the current acquisition window; similarly, obtaining changes from only a stable channel makes it difficult to determine whether the change is indeed an optical coupling change experienced by multiple channels.

[0004] Existing multi-wavelength acquisition schemes can obtain certain acquisition quality information through frequency sweeping, dark field subtraction, or channel quality evaluation. However, it is still necessary to clarify: when can changes in the reference channel be used for target channel correction, when should such changes be rejected, and how to handle the current acquisition window when reference conditions are not met. If the number of reference channels is insufficient, the changes of each reference channel relative to the baseline are not common, or saturation, discontinuities, and dark field anomalies appear in the current window, continuing to use the reference information of that window in the estimation may introduce unverified common coupling changes into the results.

[0005] Therefore, there is a need for a non-invasive hemoglobin detection system and method that can link reference frequency sweep, channel-by-channel admission, commonality determination of multiple stable reference channels, common perturbation formation, target wavelength feature correction, and prohibited output paths of the current window, so that reference information participates in the feature correction of the current window only when the commonality condition is met, thereby dynamically decoupling the common optical coupling perturbation from the target wavelength feature. Summary of the Invention

[0006] To address the shortcomings of existing technologies, the present invention aims to provide a non-invasive hemoglobin detection system and method based on dynamic decoupling of multi-wavelength arrays, which can effectively handle the problems of channel quality stratification, common optical disturbance correction, and current window result output gating during multi-wavelength finger transmission acquisition.

[0007] The above-mentioned technical objective of this invention is achieved through the following technical solution: a non-invasive hemoglobin detection system based on dynamic decoupling of a multi-wavelength array, comprising a detection body, a flip-top clamping component, a finger support, a multi-wavelength light source assembly, and a main control acquisition circuit equipped with a photodetector and an algorithm processing unit. The flip-top clamping component cooperates with the detection body to form a detection area for accommodating the finger to be tested. The finger support is connected to the detection body, and the center line of the finger support and the detection body passes through the detection area. The multi-wavelength light source assembly and the photodetector are respectively disposed on opposite sides of the detection area. Before formal acquisition, the main control acquisition circuit sequentially performs dark field acquisition, low-power frequency sweep, and formal power pre-sweep to establish the admission state and reference baseline of each wavelength channel, and drives the multi-wavelength light source assembly to acquire multi-channel photoelectric signals according to the admission state in a time-division manner. The algorithm processing unit, based on the admission state and the channel quality during formal acquisition... The quantity information determines a reference channel set, which includes at least two stable reference channels. A common optical coupling perturbation index is generated only when the common variation of each stable reference channel relative to its respective reference baseline satisfies the commonality condition. The target wavelength feature is then corrected based on the common optical coupling perturbation index and the correction coefficient corresponding to the target wavelength to decouple the common optical coupling perturbation from the target wavelength feature. The corrected target wavelength feature is provided to the hemoglobin estimation module of the algorithm processing unit to generate the hemoglobin concentration result for the current acquisition window only when the common optical coupling perturbation index has been generated and the corrected target wavelength feature is available. If the number of reference channels is less than two, the commonality condition is not met, or the corrected target wavelength feature is unavailable, the generation of the hemoglobin concentration result for the current acquisition window is prohibited, and the main control acquisition circuit is controlled to re-execute the reference frequency sweep or re-acquire.

[0008] Preferably, a flexible positioning pad is provided on the detection body between the detection area and the finger support. The flexible positioning pad cooperates with the finger support to limit the placement direction and height of the finger to be tested, and reduce the lateral displacement of the finger to be tested relative to the multi-wavelength light source assembly and the photodetector.

[0009] Preferably, the multi-wavelength light source assembly includes a light source circuit board and nine narrow-band light-emitting diodes disposed on the light source circuit board. The center wavelengths of the nine narrow-band light-emitting diodes are 620 nm, 630 nm, 660 nm, 690 nm, 720 nm, 760 nm, 810 nm, 850 nm and 940 nm, respectively. The photodetector is disposed on the side opposite to the light source circuit board, so that the detection light from different wavelength channels passes through the same detection area before reaching the photodetector.

[0010] Preferably, when the generation of hemoglobin concentration results for the current acquisition window is prohibited, the algorithm processing unit controls the main control acquisition circuit to re-execute the dark field acquisition, low-power frequency sweep and formal power pre-sweep to establish a new admission state and reference baseline, or to re-perform time-division acquisition after the finger is repositioned; the reference baseline, reference channel set, common optical coupling disturbance index and correction characteristics of the failed acquisition window are not used in the new acquisition batch after re-sweeping or re-acquisition.

[0011] Preferably, the hemoglobin estimation module first determines the hemoglobin level category based on the corrected target wavelength features or fusion features, and then the regression model combines the hemoglobin level category and the features to form the hemoglobin concentration result; or the hemoglobin estimation module directly uses a pre-configured regression or parameter estimation module.

[0012] The above-mentioned technical objective of this invention is achieved through the following technical solution: a non-invasive hemoglobin detection method based on dynamic decoupling of a multi-wavelength array, applied to a finger detection device having a multi-wavelength light source assembly and a photodetector with relative arrangement, comprising the following steps: before formal acquisition, performing dark field acquisition, low-power frequency sweep, and formal power pre-sweep sequentially to establish the admission state and reference baseline of each wavelength channel; acquiring multi-channel photoelectric signals based on the admission state in a time-division manner, and forming target wavelength characteristics and channel quality information for each wavelength channel; determining an effective wavelength subset and dynamic feature weights that change with the current acquisition state based on the channel quality information; and determining a package from the effective wavelength subset that satisfies the admission state and stability conditions. A reference channel set containing at least two stable reference channels is used. When the common variation of each stable reference channel relative to its respective reference baseline satisfies the commonality condition, a common optical coupling perturbation index is generated. The target wavelength feature is corrected based on the common optical coupling perturbation index and the correction coefficient corresponding to the target wavelength to decouple the common optical coupling perturbation from the target wavelength feature. The corrected target wavelength feature is provided to hemoglobin estimation to generate the hemoglobin concentration result for the current acquisition window. When the number of reference channels is less than two, the commonality condition is not met, or the corrected target wavelength feature is unavailable, the generation of the hemoglobin concentration result for the current acquisition window is prohibited, and a reference frequency sweep or re-acquisition is performed.

[0013] Preferably, the original sampled values ​​of each wavelength channel are subjected to dark field correction, and the corresponding DC and AC components are extracted to form target wavelength features; the dark field acquisition obtains the dark field value or ambient light baseline, the low-power frequency sweep obtains the channel response state, and the formal power pre-sweep obtains the saturation state, initial signal-to-noise ratio, and DC baseline; based on the above results, a channel-by-channel acquisition parameter table is formed, and the parameter table includes at least the channel number, drive current, illumination duration, sampling integration time, number of samplings, and a flag indicating whether formal acquisition has begun.

[0014] Preferably, a channel quality score is formed based on at least two of the channel quality information, including signal-to-noise ratio, saturation state, stability of the ratio of AC component to DC component, phase consistency, and inter-channel correlation. Wavelength channels that meet the quality admission criteria are added to the effective wavelength subset. Dynamic feature weights are determined based on the channel quality scores. The channel quality scores or dynamic feature weights are smoothed using at least one of hysteresis, shortest hold time, upper limit of weight change, or moving average.

[0015] Preferably, the commonality condition is determined based on the common changes of DC baseline variation, AC component to DC component ratio variation, and waveform phase variation of each stable reference channel relative to its reference baseline. The common changes are judged by the unidirectionality of the relative changes, the degree of dispersion of the changes, and the presence of saturation, discontinuity, dark field anomaly, or phase anomaly. The common optical coupling disturbance index is obtained by aggregating the normalized DC baseline changes of each stable reference channel through aggregation coefficients. The correction coefficient is the equipment parameter corresponding to the target wavelength channel in advance.

[0016] Preferably, at least two of the following are fused according to the dynamic feature weights: the corrected target wavelength features, the correlation features between different wavelength channels, the number of channels in the effective wavelength subset, and the common optical coupling perturbation index. Detection confidence is formed based on at least two of the following: the number of channels in the effective wavelength subset, the statistics of channel quality scores, the distribution of dynamic feature weights, and the common optical coupling perturbation index. The hemoglobin concentration result for the current acquisition window is generated only when the reference channel set is valid, the corrected target wavelength features are available, and the detection confidence meets the output conditions. Otherwise, the target wavelength features of the current acquisition window are no longer provided to hemoglobin estimation, and a reference frequency sweep or re-acquisition is performed to refresh the admission status, reference baseline, effective wavelength subset, and reference channel set.

[0017] Preferably, after prohibiting the generation of hemoglobin concentration results for the current acquisition window, the dark field acquisition, low-power frequency sweep, and formal power pre-scan are re-executed to establish a new admission state and reference baseline, or time-division acquisition is re-performed after the finger is repositioned; the reference baseline, reference channel set, common optical coupling perturbation index, and correction characteristics of the failed acquisition window are not carried over to the new acquisition batch after re-scanning or re-acquisition.

[0018] In summary, the non-invasive hemoglobin detection system and method based on dynamic decoupling of a multi-wavelength array of the present invention has the following beneficial technical effects: A reusable finger detection area is formed by the detection body, flip-top clamp, finger support, and flexible positioning pad, with the multi-wavelength light source assembly and photodetector positioned relative to each other within the detection area; through reference frequency sweeping, channel quality stratification, and effective wavelength subsets, abnormal channels are reduced in contribution or removed from the current subset before feature fusion; by selecting multiple stable reference channels and performing commonality determination, common perturbation generation uses only a validated reference set; through dynamic feature weight smoothing, frequent jumps of effective channels between adjacent windows are reduced; and by detecting confidence levels, current non-output, and rescanning or re-sampling paths, windows that do not meet the output conditions are not mistakenly used as the current result. These effects arise from the sequential relationship between entity localization, multi-wavelength acquisition, quality stratification, commonality determination, common perturbation generation, target feature correction, and output gating, rather than relying on preset model types, training data, or performance values. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the external structure of the non-invasive hemoglobin detection system of the present invention.

[0020] Figure 2 This is a schematic diagram of the internal structure of the optical components and circuit board in this invention.

[0021] Figure 3 This is a front view schematic diagram of the detection area and optical components in this invention.

[0022] Figure 4 This is a flowchart of the overall detection and result gating process of the present invention.

[0023] Figure 5 This is a flowchart of the reference frequency sweep and time-division sampling process of the present invention.

[0024] Figure 6 This is a flowchart illustrating the commonality determination, target feature correction, and failure recovery process of the present invention.

[0025] Reference numerals: 1. Detection body; 2. Flip-top clamp; 3. Finger support; 4. Multi-wavelength light source assembly; 5. Multi-wavelength LED array; 6. Photodetector; 7. Detection area; 8. Light source circuit board; 9. Flexible positioning pad; 10. Main control acquisition circuit; 301. Dark field acquisition; 302. Low-power frequency sweep; 303. Formal power pre-scan; 304. Parameter table generation; 305. Time-division sampling; 306. Channel quality assessment; 307. Reference channel selection; 308. Commonality determination; 309. Common disturbance generation; 310. Target feature correction; 311. Hemoglobin estimation; 312. No output at present; 313. Rescan or resample. Detailed Implementation

[0026] The invention will be further described below with reference to the accompanying drawings. Figure 1 As shown, the system employs a finger-transmitting acquisition structure. The detection body 1 serves as the supporting structure for the detection components and acquisition circuit. The flip-top clamp 2 cooperates with the detection body 1 to form a detection area 7 for accommodating the finger to be tested. The finger support 3 is located within the detection area 7 and supports the lower side of the finger. A flexible positioning pad 9 can be provided within the detection area 7. The flexible positioning pad 9 cooperates with the finger support 3 to limit the placement direction and height of the finger and buffer the local pressure on the finger after the flip-top clamp 2 is closed, thereby reducing the lateral displacement of the finger relative to the optical components.

[0027] like Figure 2 and Figure 3 As shown, the multi-wavelength light source assembly 4 includes a light source circuit board 8 and a multi-wavelength LED array 5 disposed on the light source circuit board 8. A photodetector 6 is disposed on the other side of the detection area 7 and opposite to the multi-wavelength LED array 5. The light source circuit board 8 carries the multi-wavelength LED array 5 and is connected to the main control acquisition circuit. The main control acquisition circuit illuminates the corresponding light-emitting diodes in the multi-wavelength LED array 5 in a time-division manner. The photodetector 6 receives the light signal after passing through the finger tissue in the detection area 7. The housing, plates, and optical components shown in the supplementary drawings are used to illustrate the relative arrangement and load-bearing relationship. The housing dimensions, plate thicknesses, connector types, or fastening methods not shown in the figures are not considered necessary limitations of the present invention.

[0028] The main control acquisition circuit is connected to the multi-wavelength light source component 4 and the photodetector 6. The algorithm processing unit receives the sampling data and channel status data output by the main control acquisition circuit. The entire system uses transmitted light signals as the acquisition object of this technical solution and does not use reflective light paths as the acquisition path in this implementation.

[0029] like Figure 4As shown, the system first enters dark-field acquisition 301 at the start of a detection. Dark-field acquisition 301 is controlled by the main control acquisition circuit, and reads the baseline response of the photodetector when the detection light of the corresponding channel is not used for formal sampling. Dark-field acquisition 301 does not indiscriminately replace all subsequent data, but instead retains a corresponding baseline for each wavelength channel that can be used to subtract the influence of dark field or ambient light. The main control acquisition circuit associates and stores this baseline with the subsequent original photoelectric sampling values ​​of the same channel, enabling the algorithm processing unit to identify which wavelength channel and corresponding baseline batch the sampling value belongs to before forming the DC component, AC component, and target wavelength characteristics.

[0030] Following dark-field acquisition 301, a low-power frequency sweep 302 is executed. The low-power frequency sweep 302 sequentially illuminates at least a portion of the wavelength channels according to a predetermined channel order, and the photodetector outputs the channel response within the corresponding illumination time. The low-power frequency sweep 302 is used to observe whether each channel has a usable response and whether there are any obvious abnormal states, rather than directly generating the current hemoglobin concentration result at this stage. For channels with insufficient response, obvious saturation, or dark-field anomalies, the main control acquisition circuit can record them as not entering formal acquisition or awaiting verification; the specific judgment conditions are determined by the device parameter configuration and the frequency sweep data, and are not limited by filling in a fixed threshold in the instruction manual.

[0031] Following the low-power frequency sweep 302, the main control acquisition circuit performs the formal power pre-sweep 303. The formal power pre-sweep 303 acquires the saturation state, initial signal-to-noise ratio, and DC baseline of each channel according to the conditions of the channel to be acquired in the formal acquisition. Unlike the baseline obtained in the dark field acquisition 301, the DC baseline of the formal power pre-sweep 303 is used to characterize the optical response state of the corresponding channel before it enters the formal acquisition in this batch of tests. Because different fingers, different placement positions, and different channel transmission conditions may differ, the formal power pre-sweep 303 provides each channel with its own reference baseline for subsequent commonality determination, rather than copying the baseline of one channel to other channels.

[0032] like Figure 5 As shown, after the dark field acquisition 301, low-power frequency sweep 302, and formal power pre-sweep 303 are completed sequentially, the parameter table generation 304 reads the acquisition results of the above three stages and forms a channel-by-channel acquisition parameter table. The parameter table generation 304 associates at least the channel number, LED drive current, illumination duration, sampling integration time, sampling count, and whether it has entered formal acquisition with the corresponding wavelength channel. The parameter table can be stored in a storage area accessible to the main control acquisition circuit, or it can be generated by the algorithm processing unit and written back to the main control acquisition circuit. The function of parameter table generation 304 is to convert the channel status obtained from this frequency sweep into control information that can be executed for time-division sampling, thereby avoiding continuous acquisition of channels determined not to enter formal acquisition under the same sampling conditions.

[0033] Time-division sampling 305 operates according to the channel parameter table output by parameter table generation 304. Within one sampling cycle, the main control acquisition circuit sequentially drives LEDs with different center wavelengths and reads the photodetector output during each corresponding lighting period. For the implementation with nine narrowband LEDs, the center wavelengths can be 620 nm, 630 nm, 660 nm, 690 nm, 720 nm, 760 nm, 810 nm, 850 nm, and 940 nm, respectively. Time-division sampling 305 enables the sampling results of different wavelength channels to be distinguished and merged into their respective channels according to the lighting period, and correlated with the corresponding baselines formed by dark-field acquisition 301 and formal power pre-scan 303. The sampling sequence, drive current, integration time, and number of samples can be jointly determined by the device parameter configuration and the current frequency sweep result. The specific values ​​vary depending on the device implementation and do not affect the processing order in this invention where reference information must first undergo commonality determination.

[0034] The multi-channel photoelectric signals obtained after time-division sampling 305 first enter channel quality assessment 306. Channel quality assessment 306 is executed by the algorithm processing unit, which can generate quality information for each channel, such as signal-to-noise ratio, saturation state, stability of the AC to DC component ratio, phase consistency, and inter-channel correlation. For the sampled signal after dark field correction, the algorithm processing unit extracts the DC and AC components within the current window and can generate the AC to DC component ratio, absorbance change, peak-to-trough amplitude difference, cardiac cycle area, or characteristic ratios between different wavelength channels. The results of channel quality assessment 306 are used in conjunction with the admission status obtained from parameter table generation 304: channels that have been prohibited from entering formal acquisition should not be directly entered into the reference set due to the accidental appearance of waveforms in the current window; channels that have already been admitted still need to pass the stability conditions of the current window.

[0035] To illustrate the relationship between dark field baseline, common perturbation, and target feature correction in this invention, the algorithm processing unit can process the data of the current acquisition window according to the following relationship. This relationship is used to define under what data conditions the correction feature is formed, and does not indicate that this invention has obtained any clinical performance data. For the i-th wavelength channel, the original photoelectric sampling value is subtracted from the dark field value or ambient light baseline of the same channel to obtain the corrected sampling value. The dark field subtraction relationship is as follows: Pi(t) = Ri(t) - Bi(t) (1)

[0036] Where P_i(t) represents the corrected sampled value of the i-th wavelength channel after dark field or ambient light subtraction, R_i(t) represents the original photoelectric sampled value of the i-th wavelength channel at time t, and B_i(t) represents the dark field value or ambient light baseline corresponding to this channel. The dark field value is obtained from the dark field acquisition in this detection. The main control acquisition circuit obtains this baseline during the period when the multi-wavelength light source components are not lit or have not entered the formal acquisition phase. This processing ensures that the subsequent DC component, AC component, and target wavelength characteristics are based on the corrected sampled value of the same channel, rather than directly mixing in the unsubtracted fixed bias.

[0037] The relationship between common perturbation and target feature correction is such that the algorithm processing unit can process the data of the current acquisition window according to the following relationship. This relationship is used to define under what data conditions the correction feature is formed, and does not indicate that the present invention has obtained any clinical performance data. For the i-th wavelength channel, the original photoelectric sampling value is subtracted from the dark field value or ambient light baseline of the same channel to obtain the corrected sampling value. The dark field subtraction relationship is as follows: Pi(t) = Ri(t) - Bi(t) (1)

[0038] Where P_i(t) represents the corrected sampled value of the i-th wavelength channel after dark field or ambient light subtraction, R_i(t) represents the original photoelectric sampled value of the i-th wavelength channel at time t, and B_i(t) represents the dark field value or ambient light baseline corresponding to this channel. The dark field value is obtained from the dark field acquisition in this detection. The main control acquisition circuit obtains this baseline during the period when the multi-wavelength light source components are not lit or have not entered the formal acquisition phase. This processing ensures that the subsequent DC component, AC component, and target wavelength characteristics are based on the corrected sampled value of the same channel, rather than directly mixing in the unsubtracted fixed bias.

[0039] In one quality stratification approach, the algorithm processing unit generates channel quality scores based on a combination of factors including signal-to-noise ratio, saturation state, stability of the AC to DC component ratio, phase consistency, and inter-channel correlation, according to the device configuration. Channel quality scores do not require fixed threshold values; quality admission criteria can be determined based on device batch calibration, current acquisition status, or pre-configured parameters. Channels that meet the admission criteria and quality admission conditions constitute an effective wavelength subset A, while channels that do not meet the criteria have their subsequent contributions reduced or are removed from the current effective wavelength subset A.

[0040] For channels within the effective wavelength subset A, the algorithm processing unit can assign dynamic feature weights based on the relative relationship of the quality scores of each channel. These dynamic feature weights are used for multi-wavelength feature fusion and do not directly determine whether the common perturbation index is valid. The reference channel set R still needs to select at least two stable reference channels from the effective wavelength subset A and pass a commonality determination. To avoid a channel frequently being added to or removed from the effective wavelength subset A between adjacent windows due to a single instance of noise, the algorithm processing unit can use hysteresis, minimum hold time, upper limit of change, or moving average processing for the channel quality score or dynamic feature weights. A channel is removed only if it continuously fails to meet the quality admission criteria, and is allowed to rejoin only after continuous recovery.

[0041] Dynamic feature weights and reference aggregation coefficients in common perturbation relationships are different parameters: dynamic feature weights are used to determine the contribution of effective channel features to the fused features, while reference aggregation coefficients are only used to aggregate the changes of reference channels relative to their respective baselines after the commonality condition has been met.

[0042] After the dynamic feature weights are formed, the algorithm processing unit can form a fusion feature from at least two of the following: the corrected target wavelength features, the correlation features between different wavelength channels, the number of channels in the effective wavelength subset A, and the common optical coupling perturbation index. Simultaneously, the algorithm processing unit forms a detection confidence level based on at least two of the following: the number of channels in the effective wavelength subset A, the statistics of channel quality scores, the distribution of dynamic feature weights, and the common optical coupling perturbation index. This detection confidence level only indicates whether the current window meets the output conditions and does not represent clinical accuracy; the fusion feature is only provided for hemoglobin estimation when the reference channel set R is valid, the target feature correction is complete, and the detection confidence level meets the output conditions.

[0043] like Figure 6 As shown, the reference channel selection 307 receives the quality information output by the channel quality assessment 306, the admission status of each channel, and the reference baseline established by the formal power pre-scan 303. Reference channel selection 307 first eliminates channels with saturation, breakpoints, dark field anomalies, or insufficient stability, and then determines stable reference channels from the remaining channels. The reference channel set R formed by reference channel selection 307 must contain at least two stable reference channels. This minimum requirement provides a basis for subsequent processing to compare the relative changes of multiple reference channels; a change in a single channel only indicates that the channel itself may have changed, and cannot be used as evidence that multiple channels have changed together.

[0044] The commonality determination 308 is executed after the reference channel selection 307. The commonality determination 308 does not automatically consider all changes in the reference set R as common optical coupling perturbations. Instead, it compares the direction of change, dispersion, phase, saturation, discontinuity, and dark field state of each stable reference channel relative to its respective reference baseline. When the relative changes of the stable reference channels meet the preset common change conditions and no inapplicable state occurs, the commonality determination 308 forms a valid state; when the number of reference channels in the set R is less than two, at least one key reference channel fails resulting in an insufficient set, or the changes of each channel are not in the same direction or the dispersion does not meet the common change conditions, the commonality determination 308 forms an invalid state. The conditions used for commonality determination are derived from the device parameter configuration and the current channel state; specific thresholds, fixed window lengths, or continuous heartbeat counts are not considered necessary technical limitations in this implementation.

[0045] The common optical coupling disturbance index is determined based on the combined changes of the DC baseline change, the ratio of AC component to DC component change, and the waveform phase change of each stable reference channel in the reference channel set relative to its reference baseline.

[0046] When commonality determination 308 is valid, common perturbation generation 309 reads the current DC component, reference sweep DC baseline, and aggregation coefficient of each channel in the reference channel set R. Common perturbation generation 309 forms a common optical coupling perturbation index for the current window based on the aforementioned common perturbation relationship. The aggregation coefficient is used to aggregate the changes of stable reference channels that have been verified for commonality relative to their own baselines; it is not used as the prediction weight for each target wavelength channel, nor is it used to directly replace the commonality determination with the channel quality score. If the reference channel set changes in the current window, common perturbation generation 309 re-forms the common perturbation index using the currently valid reference set; the common perturbation index formed in the previously failed window is not used in the new window after resweeping or re-acquisition.

[0047] When the reference channel set is valid, the normalized change of each stable reference channel relative to the DC baseline obtained in this reference frequency sweep can be aggregated by the aggregation coefficient into a common optical coupling perturbation index for the current window. The common perturbation relationship is as follows: D(t)=∑j∈RαjDCj(t)-DCj,0DCj,0 (2)

[0048] Where D(t) represents the common optical coupling perturbation index of the current acquisition window, R represents the reference channel set containing at least two stable reference channels, the aggregation coefficient represents the aggregation coefficient of the j-th channel in the reference channel set, DC_j(t) represents the DC component of the j-th reference channel in the current window, and DC_{j,0} represents the DC baseline obtained by the reference channel in this reference frequency sweep. The aggregation coefficient is determined by the device parameter configuration and the reference channel status and is not used as the prediction weight for the target wavelength. This common perturbation relationship is only used when the number of reference channels is at least two, each reference channel is stable, and the commonality condition is met; if any of the prerequisites are not met, D(t) is not formed.

[0049] When the reference channel set is valid, the normalized change of each stable reference channel relative to the DC baseline obtained in this reference frequency sweep can be aggregated by the aggregation coefficient into a common optical coupling perturbation index for the current window. The common perturbation relationship is as follows: D(t)=∑j∈RαjDCj(t)-DCj,0DCj,0 (2)

[0050] Where D(t) represents the common optical coupling perturbation index of the current acquisition window, R represents the reference channel set containing at least two stable reference channels, the aggregation coefficient represents the aggregation coefficient of the j-th channel in the reference channel set, DC_j(t) represents the DC component of the j-th reference channel in the current window, and DC_{j,0} represents the DC baseline obtained by the reference channel in this reference frequency sweep. The aggregation coefficient is determined by the device parameter configuration and the reference channel status and is not used as the prediction weight for the target wavelength. This common perturbation relationship is only used when the number of reference channels is at least two, each reference channel is stable, and the commonality condition is met; if any of the prerequisites are not met, D(t) is not formed.

[0051] The target feature correction 310 only receives the common optical coupling perturbation index formed by the common perturbation generation 309 in an effective state. For each target wavelength channel, the algorithm processing unit obtains the correction coefficient corresponding to that channel from the device parameters and corrects the target feature using the common perturbation index to obtain the corrected feature. The correction coefficient is not a quantity temporarily obtained in the current detection window using unverified reference channel changes, but rather a device parameter corresponding to the target wavelength channel. The target feature correction 310 does not generate the corrected feature when the common perturbation index has not been formed, the target feature is insufficient, or the correction coefficient is missing; in this case, the uncorrected current feature cannot bypass the commonality determination and directly enter the hemoglobin estimation.

[0052] The aforementioned correction coefficients are used to establish a correspondence with the target channel, so that the correction of the common optical coupling perturbation index for different target wavelength characteristics has a corresponding dimensional relationship; the present invention does not limit the specific value of the correction coefficients, the training sample size or the update rule.

[0053] When the common optical coupling perturbation index has been formed and the target features are available, the algorithm processing unit forms the correction features based on the correction coefficients corresponding to the target wavelength. The target feature correction relationship is as follows: F′i=Fi-kiD(t) (3)

[0054] Where F'_i represents the correction feature of the i-th target wavelength channel, F_i represents the original feature of the i-th target wavelength channel, k_i represents the correction coefficient corresponding to the target wavelength channel, and D(t) represents the common optical coupling perturbation index formed after commonity verification. The dimensions of the correction coefficients make the correction terms comparable to the target features; when the correction coefficients are missing, the common perturbation is not formed, or the target features are insufficient, the algorithm processing unit does not perform the current correction and enters the path of not outputting or rescanning or resampling.

[0055] When the common optical coupling perturbation index has been formed and the target features are available, the algorithm processing unit forms the correction features based on the correction coefficients corresponding to the target wavelength. The target feature correction relationship is as follows: F′i=Fi-kiD(t) (3)

[0056] Where F'_i represents the correction feature of the i-th target wavelength channel, F_i represents the original feature of the i-th target wavelength channel, k_i represents the correction coefficient corresponding to the target wavelength channel, and D(t) represents the common optical coupling perturbation index formed after commonity verification. The dimensions of the correction coefficients make the correction terms comparable to the target features; when the correction coefficients are missing, the common perturbation is not formed, or the target features are insufficient, the algorithm processing unit does not perform the current correction and enters the path of not outputting or rescanning or resampling.

[0057] Once the target features are formed and available, the algorithm processing unit generates corrected features based on correction coefficients corresponding to the target wavelength. The target feature correction relationship is as follows: F′i=Fi-kiD(t) (3)

[0058] Where F'_i represents the correction feature of the i-th target wavelength channel, F_i represents the original feature of the i-th target wavelength channel, k_i represents the correction coefficient corresponding to the target wavelength channel, and D(t) represents the common optical coupling perturbation index formed after commonity verification. The dimensions of the correction coefficients make the correction terms comparable to the target features; when the correction coefficients are missing, the common perturbation is not formed, or the target features are insufficient, the algorithm processing unit does not perform the current correction and enters the path of not outputting or rescanning or resampling. (3)

[0059] The hemoglobin estimation 311 only receives correction features output by the target feature correction 310 that belong to the current valid window. The hemoglobin estimation 311 can be a pre-configured parameter estimation module, and its specific model type is not a necessary technical feature of this invention. After receiving the correction features of the current valid window, the hemoglobin estimation 311 generates the hemoglobin concentration result for the current window while retaining the valid state of the current window. Since the target feature correction 310 has already completed the quantity and commonality judgment of the reference set, the hemoglobin estimation 311 will not use window data with invalid reference conditions as the estimation input for this result.

[0060] In one optional estimation method, hemoglobin estimation 311 includes a category determination stage and a regression stage. The category determination stage outputs the hemoglobin level category based on the fusion features, and the regression stage combines the hemoglobin level category and the fusion features to form the hemoglobin concentration result. The category determination stage and the regression stage can be implemented by a pre-configured statistical model, machine learning model, or parametric model; the two-stage structure described is only used to illustrate optional estimation carriers and is not intended to limit the invention. Regardless of whether a two-stage structure or a direct regression structure is used, the following prerequisites apply: the reference channel set R is valid, the common perturbation index has been formed, the correction features are available, and the detection confidence meets the output conditions.

[0061] When commonality determination 308 becomes invalid, correction features are unavailable, or detection confidence does not meet the output conditions, "Do Not Output Now" 312 is triggered. "Do Not Output Now" 312 means prohibiting the generation of hemoglobin concentration results for the current acquisition window, rather than marking the window as low quality and still outputting a current result. "Do Not Output Now" 312 prevents the common perturbation index, correction features, or fusion features of the current window from being provided to hemoglobin estimation 311; if the device interface retains previously obtained records, these records should only be displayed as historical records and should not be represented as the result of the current acquisition window. "Do Not Output Now" 312 can simultaneously prompt the system to reposition the finger, re-execute the reference sweep, or re-acquire data to obtain applicable reference sweep data again.

[0062] Rescanning or resampling 313 is executed after the current output 312 is stopped. Rescanning or resampling 313 can first control the main control acquisition circuit to re-perform dark field acquisition 301, low-power frequency sweep 302, and formal power pre-scan 303, and then generate a new channel parameter table from the parameter table generation 304; alternatively, it can restart time-division sampling 305 after the finger is repositioned, and redetermine the reference channel set R in the new acquisition batch. Rescanning or resampling 313 does not continue to use the reference baseline, reference channel set, common disturbance index, or correction characteristics of the failed window. In this way, the validity of the reference conditions is directly correlated with whether the current window result allows generation, preventing invalid windows from being interpreted as current results by retaining old reference data.

[0063] The method of this invention can be implemented as follows: S1. The main control acquisition circuit sequentially performs dark field acquisition, low-power frequency sweep, and formal power pre-sweep to establish a channel-by-channel admission state, reference baseline, and parameter table. S2. The main control acquisition circuit drives the multi-wavelength light source component in a time-division manner according to the parameter table, and the algorithm processing unit performs dark field correction on the sampled signal to form target features and channel quality information. S3. The algorithm processing unit forms a channel quality score, an effective wavelength subset A, and dynamic feature weights based on the channel quality information, and smooths the quality score or dynamic feature weights. S4. The algorithm processing unit determines a reference channel set R from the effective wavelength subset A, which includes no less than two stable reference channels, and makes a commonality judgment on the changes of each stable reference channel relative to its reference baseline. S5. Only when the reference channel set R is valid, the algorithm processing unit forms a common optical coupling perturbation index and forms a correction feature with corresponding correction coefficients. S6. The algorithm processing unit forms a fusion feature based on the dynamic feature weights, and combines at least two of the following: the number of effective channels, quality score, weight distribution, and common optical coupling perturbation index to form a detection confidence level. S7. When the reference channel set is valid, the correction feature is available, and the detection confidence meets the output conditions, the hemoglobin estimation 311 generates the current window result; otherwise, the current result is not output 312 and the current result is prohibited. The main control acquisition circuit re-establishes the reference data required for this detection according to the rescan or re-sampling 313.

[0064] like Figure 4 and Figure 6As shown, in a complete detection batch, the main control acquisition circuit first uses dark field acquisition 301 to obtain the dark field or ambient light baseline of each channel, and then uses low-power frequency sweep 302 and formal power pre-sweep 303 to sequentially obtain the channel response state, saturation state, initial signal-to-noise ratio, and DC baseline. Parameter table generation 304 writes these current batch data into a channel-by-channel parameter table, and time-division sampling 305 lights up the corresponding wavelength LEDs according to the parameter table and acquires the photoelectric response of the same detection area. Channel quality evaluation 306 extracts the DC component, AC component, phase, and inter-channel relationship for judgment from the signal after dark field correction, while retaining the correspondence between each data item and its wavelength channel, sampling period, and reference baseline. The algorithm processing unit forms a channel quality score, effective wavelength subset A, and dynamic feature weights based on the current quality information, and smooths the quality score or dynamic feature weights; reference channel selection 307 does not select a seemingly better channel from all channels, but establishes a reference channel set R from the channels that have entered formal acquisition and are currently stable, and ensures that the set has at least two stable reference channels. After the reference channel selection 307 forms a set, the commonality determination 308 compares the direction and dispersion of change of each channel relative to its respective reference baseline. Only when these changes meet the common change conditions, and there are no saturation, breakpoints, dark fields, or phase anomalies in each channel that would render the current determination inapplicable, will the commonality determination 308 output a valid state. The common perturbation generation 309 only receives the reference set and corresponding baseline under this valid state to form a common perturbation index to characterize the common optical coupling changes in the current window. The target feature correction 310 then combines this index with the correction coefficient corresponding to the target wavelength to obtain the correction feature. The algorithm processing unit further forms fusion features and detection confidence based on dynamic feature weights. The hemoglobin estimation 311 only consumes the correction features of the above valid states. Therefore, single-channel anomalies or reference condition failures during the acquisition process will not bypass the commonality determination through ordinary channel screening. If the commonality determination 308 finds that the reference channel set R is less than two, the common changes are not in the same direction, the dispersion does not meet the conditions, or the subsequent correction lacks necessary data, or the detection confidence does not meet the output conditions, then the current output 312 immediately prevents this window from entering hemoglobin estimation 311, and at the same time, the main control acquisition circuit re-executes dark field acquisition 301, low power frequency sweep 302, and formal power pre-scan 303 by rescanning or re-sampling 313. After restarting, parameter table generation 304, time-division sampling 305, channel quality assessment 306, reference channel selection 307, and commonality determination 308 are all run for the new acquisition batch; the baseline, reference set, and common disturbance index of the failed batch will not be used as input for the new batch. Therefore, Figure 4 The acquisition, judgment, correction, estimation, and recovery paths shown are sequential within the same batch. Figure 6The invalid branch shown can also clearly return to the new reference frequency sweep, rather than continuing to generate the results of the current window under invalid conditions. In the case where commonality determination 308 has reached a valid state but target feature correction 310 has not yet been completed, the algorithm processing unit should still retain the correspondence between the reference set, channel baseline, and target features of that window; once the correction coefficient is missing, the target feature is incomplete, or the reference channel is found to be invalid during the correction process, the current output 312 will cover the subsequent estimation path of that window, and rescanning or resampling 313 will isolate the data to be processed in that window from the new frequency sweep batch. In the case where commonality determination 308 remains valid and target feature correction 310 is completed, the data received by hemoglobin estimation 311 comes from the same batch of time-division sampling 305 and common perturbation generation 309, therefore the reference baselines on which sampling, determination, and correction are based correspond to each other. For acquisition restarted after finger repositioning, dark field acquisition 301 re-acquires the dark field or ambient light baseline, low-power frequency sweep 302 and formal power pre-scan 303 reconfirm the channel status, and parameter table generation 304 updates the channels that can enter formal acquisition accordingly, avoiding the direct transfer of the channel admission status before repositioning to the window after repositioning. Through this sequence, whether the result of the current window can be generated is jointly limited by the commonality determination 308 and the branch of not outputting 312. The signal transmission between time-division sampling 305, common perturbation generation 309 and target feature correction 310 also has a traceable sequential relationship.

[0065] In one alternative device implementation, the multi-wavelength light source assembly can be a multi-wavelength light-emitting diode array mounted on the same light source circuit board, with the photodetector positioned on the side opposite the light source circuit board. A finger support and a flexible positioning pad within the detection body jointly define the finger position, and the flip-top clamping mechanism, when closed, holds the finger to be tested between the light source assembly and the photodetector. This device enables the detection light from each wavelength channel to pass through the same finger detection area before reaching the photodetector; its purpose is to provide a stable transmission acquisition basis for the reference frequency sweep, time-division sampling, and commonality determination of this invention, without considering the size, material, or specific connection method of the clamping structure as a necessary condition for the generation of common disturbances.

[0066] In another optional processing method, if the channel quality assessment 306 finds that a selected reference channel exhibits saturation, breakpoints, dark field anomalies, phase anomalies, or a significantly unstable ratio of AC to DC components within the current window, the channel can be removed from the candidate reference channels, and the remaining channels can be reassessed to determine whether they still meet the condition of having at least two stable reference channels. If the remaining channels still meet the condition of commonality determination 308, common disturbance generation 309 continues to form common disturbance indices based on the redefined set of reference channels; if there are insufficient remaining channels or the commonality condition is not met, the process follows the path of no output 312 and rescan or resample 313. This optional processing maintains the reference channel set corresponding to the actual state of the current window, rather than relying on a reference channel list that remains unchanged after being fixed during the frequency sweep phase.

[0067] In another optional processing method, target feature correction 310 can call the corresponding correction coefficients for different target wavelength channels and adopt parameter configurations adapted to different types of target features. Regardless of whether the target feature uses the ratio of AC to DC components, absorbance change, peak-to-trough amplitude difference, cardiac cycle area, or feature ratio between different wavelengths, it is based on the premise that commonality determination 308 has formed an effective state and common disturbance generation 309 has formed the common disturbance index of the current window. If these premises are not met, the system will still process according to the current no output 312 and rescan or resample 313, and will not interpret the switch of feature type as a replacement of the failed reference condition.

[0068] The above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. For those skilled in the art, any equivalent adjustments made to the shape of the detection subject, the arrangement of the light source components, the storage location of the parameter table, and the rescan triggering method of the current window without departing from the concept of the present invention should fall within the scope of protection of the present invention.

Claims

1. A non-invasive hemoglobin detection system based on dynamic decoupling of a multi-wavelength array, comprising a detection body (1), a flip-top clamping component (2), a finger support component (3), a multi-wavelength light source assembly (4), and a main control acquisition circuit (10) equipped with a photodetector (6) and an algorithm processing unit, wherein the flip-top clamping component (2) cooperates with the detection body (1) to form a detection area (7) for accommodating the finger to be tested, the finger support component (3) is connected to the detection body (1), and the center line of the finger support component (3) and the detection body (1) passes through the detection area (7), and the multi-wavelength light source assembly (4) and the photodetector (6) are respectively disposed on opposite sides of the detection area (7); characterized in that, Before formal acquisition, the main control acquisition circuit sequentially performs dark field acquisition (301), low power frequency sweep (302), and formal power pre-sweep (303) to establish the admission status and reference baseline of each wavelength channel, and drives the multi-wavelength light source component (4) to acquire multi-channel photoelectric signals according to the admission status in a time-division manner; the algorithm processing unit determines the reference channel set according to the admission status and channel quality information in formal acquisition, and the reference channel set contains at least two stable reference channels; A common optical coupling perturbation index is generated only when the common variation of each stable reference channel relative to its respective reference baseline meets the commonality condition. The target wavelength feature is then corrected based on the common optical coupling perturbation index and the correction coefficient corresponding to the target wavelength to decouple the common optical coupling perturbation from the target wavelength feature. The corrected target wavelength feature is provided to the hemoglobin estimation module (311) of the algorithm processing unit only when the common optical coupling perturbation index has been generated and the corrected target wavelength feature is available to generate the hemoglobin concentration result of the current acquisition window. If the number of reference channels is less than two, the commonality condition is not met, or the corrected target wavelength feature is unavailable, the generation of the hemoglobin concentration result of the current acquisition window is prohibited, and the main control acquisition circuit is controlled to re-execute the reference frequency sweep or re-acquire.

2. The non-invasive hemoglobin detection system based on dynamic decoupling of a multi-wavelength array according to claim 1, characterized in that, A flexible positioning pad (9) is provided on the detection body (1) between the detection area (7) and the finger support part (3). The flexible positioning pad (9) cooperates with the finger support part (3) to limit the placement direction and height of the finger to be tested, and reduce the lateral offset of the finger to be tested relative to the multi-wavelength light source component (4) and the photodetector (6).

3. The non-invasive hemoglobin detection system based on dynamic decoupling of a multi-wavelength array according to claim 1, characterized in that, The multi-wavelength light source assembly (4) includes a light source circuit board (8) and nine narrow-band light-emitting diodes disposed on the light source circuit board (8). The center wavelengths of the nine narrow-band light-emitting diodes are 620 nm, 630 nm, 660 nm, 690 nm, 720 nm, 760 nm, 810 nm, 850 nm and 940 nm, respectively. The photodetector (6) is disposed on the side opposite to the light source circuit board (8) so that the detection light of different wavelength channels passes through the same detection area (7) and then reaches the photodetector (6).

4. The non-invasive hemoglobin detection system based on dynamic decoupling of a multi-wavelength array according to claim 1, characterized in that, When the generation of hemoglobin concentration results for the current acquisition window is prohibited, the algorithm processing unit controls the main control acquisition circuit (10) to re-execute the dark field acquisition (301), low power frequency sweep (302) and formal power pre-sweep (303) to establish a new admission state and reference baseline, or to re-perform time-division acquisition after the finger is repositioned; the reference baseline, reference channel set, common optical coupling disturbance index and correction characteristics of the failed acquisition window are not used in the new acquisition batch after re-sweep or re-acquisition.

5. The non-invasive hemoglobin detection system based on dynamic decoupling of a multi-wavelength array according to claim 1, characterized in that, The hemoglobin estimation module (311) first determines the hemoglobin level category based on the corrected target wavelength features or fusion features, and then the regression model combines the hemoglobin level category and the features to form the hemoglobin concentration result; or the hemoglobin estimation module (311) directly uses the pre-configured regression or parameter estimation module.

6. A non-invasive hemoglobin detection method based on dynamic decoupling of a multi-wavelength array, applied to a finger detection device having a multi-wavelength light source assembly (4) and a photodetector (6) arranged in a relative manner, characterized in that, The process includes the following steps: before formal acquisition, perform dark field acquisition (301), low-power frequency sweep (302), and formal power pre-sweep (303) in sequence to establish the admission status and reference baseline of each wavelength channel; acquire multi-channel photoelectric signals based on the admission status in a time-division manner, and form the target wavelength characteristics and channel quality information of each wavelength channel; determine the effective wavelength subset and dynamic feature weights that change with the current acquisition status based on the channel quality information; and determine a set of reference channels containing at least two stable reference channels from the effective wavelength subset that satisfies the admission status and stability conditions. When the common variation of each stable reference channel relative to its respective reference baseline satisfies the commonality condition, a common optical coupling perturbation index is generated, and the target wavelength feature is corrected based on the common optical coupling perturbation index and the correction coefficient corresponding to the target wavelength, so as to decouple the common optical coupling perturbation from the target wavelength feature; the corrected target wavelength feature is provided to hemoglobin estimation (311) to generate the hemoglobin concentration result of the current acquisition window; when the number of reference channels is less than two, the commonality condition is not met, or the corrected target wavelength feature is unavailable, the generation of the hemoglobin concentration result of the current acquisition window is prohibited, and the reference frequency sweep or re-acquisition is performed.

7. The non-invasive hemoglobin detection method based on dynamic decoupling of a multi-wavelength array according to claim 6, characterized in that, Dark field correction is performed on the original sampled values ​​of each wavelength channel, and the corresponding DC and AC components are extracted to form the target wavelength features; The dark field acquisition (301) obtains the dark field value or ambient light baseline, the low power frequency sweep (302) obtains the channel response state, and the formal power pre-sweep (303) obtains the saturation state, initial signal-to-noise ratio and DC baseline; based on the above results, a channel-by-channel acquisition parameter table is formed, and the parameter table includes at least the channel number, drive current, lighting duration, sampling integration time, number of samplings and whether formal acquisition has begun.

8. The non-invasive hemoglobin detection method based on dynamic decoupling of a multi-wavelength array according to claim 7, characterized in that, Channel quality scores are formed based on at least two of the following channel quality information: signal-to-noise ratio, saturation state, stability of the ratio of AC component to DC component, phase consistency, and inter-channel correlation. Wavelength channels that meet the quality admission criteria are added to the effective wavelength subset. Dynamic feature weights are determined based on the channel quality scores. The channel quality scores or dynamic feature weights are smoothed using at least one of the following: hysteresis, shortest hold time, upper limit of weight change, or moving average.

9. The non-invasive hemoglobin detection method based on dynamic decoupling of a multi-wavelength array according to claim 8, characterized in that, The commonality condition is determined based on the common changes of DC baseline variation, AC component to DC component ratio variation, and waveform phase variation of each stable reference channel relative to its reference baseline. The common changes are judged by the unidirectionality of the relative changes, the degree of dispersion of the changes, and the presence of saturation, discontinuity, dark field anomaly, or phase anomaly. The common optical coupling disturbance index is obtained by aggregating the normalized DC baseline changes of each stable reference channel through aggregation coefficients. The correction coefficient is the equipment parameter corresponding to the target wavelength channel in advance.

10. The non-invasive hemoglobin detection method based on dynamic decoupling of a multi-wavelength array according to claim 11, characterized in that, According to the dynamic feature weights, at least two of the corrected target wavelength features, correlation features between different wavelength channels, number of channels in the effective wavelength subset, and common optical coupling perturbation index are fused, and a detection confidence level is formed according to at least two of the number of channels in the effective wavelength subset, statistics of channel quality scores, distribution of dynamic feature weights, and common optical coupling perturbation index. The hemoglobin concentration result of the current acquisition window is generated only when the reference channel set is valid, the corrected target wavelength features are available, and the detection confidence level meets the output conditions. Otherwise, the target wavelength features of the current acquisition window are stopped from being provided to the hemoglobin estimation (311), and the reference frequency sweep or re-acquisition is re-executed to refresh the admission status, reference baseline, effective wavelength subset, and reference channel set.

11. The non-invasive hemoglobin detection method based on dynamic decoupling of a multi-wavelength array according to claim 11, characterized in that, After the generation of hemoglobin concentration results for the current acquisition window is prohibited, the dark field acquisition (301), low power frequency sweep (302), and formal power pre-scan (303) are re-executed to establish a new admission state and reference baseline, or time-division acquisition is re-performed after the finger is repositioned; the reference baseline, reference channel set, common optical coupling perturbation index, and correction characteristics of the failed acquisition window are not used in the new acquisition batch after re-scanning or re-acquisition.

Images