Biometric feature recognition adaptive optimization method, device, equipment and storage medium

Abstract

The application provides a biometric feature adaptive optimization method, device, equipment and storage medium. The method comprises the following steps: extracting biometric feature sample data and corresponding registered user data from a historical database after triggering updating; performing feature drift and incremental analysis on the biometric feature sample data according to time sequence to obtain feature incremental data; compensating the registered user data according to collected environmental data and the feature incremental data to obtain verification feature data; obtaining latest collected legal feature data from the historical database to perform similarity analysis and confidence conversion on the legal feature data and the verification feature data to obtain a confidence score value; and updating the registered user data according to the verification feature data when the confidence score value is greater than a feature updating threshold. The application triggers dynamic updating based on feature drift and incremental analysis of historical samples, nonlinear environmental compensation and time sequence evolution, and is real-time adaptive to biometric feature and environmental changes.

Description

Technical Field

[0001] This application relates to the field of feature recognition technology, and in particular to an adaptive optimization method, apparatus, device and storage medium for biometric recognition. Background Technology

[0002] During the identity verification process, a user's biometric characteristics may drift slightly over time due to factors such as health status and environmental conditions. If the registration template remains unchanged for a long period of time, it will be difficult to accurately reflect the current characteristics, which may lead to false rejection or false acceptance.

[0003] Currently, common adaptive optimization methods in identity recognition systems typically consist of simple threshold adjustments, sliding window updates, or incremental feature fusion. For example, they may linearly weight the template by calculating the average feature vector of the most recent authentications, or update the template after removing low-quality samples based on quality assessments. These methods often incorporate normalization and principal component analysis to reduce noise and improve the system's tolerance to feature shifts. While these methods can enhance the model's adaptability to static drift to some extent, they lack a comprehensive characterization of environmental changes and temporal evolution. Summary of the Invention

[0004] In view of this, this application provides an adaptive optimization method, apparatus, device and storage medium for biometric identification, which improves the accuracy of identification templates.

[0005] On one hand, embodiments of this application provide an adaptive optimization method for biometric identification, the method comprising: When the number of valid authentications reaches the preset update trigger threshold, biometric sample data and corresponding registered user data are extracted from the preset historical database. The biometric sample data is subjected to feature drift analysis according to time sequence to obtain drift data of various biometrics, and the drift data is subjected to incremental analysis to obtain feature incremental data. The registered user data is nonlinearly compensated based on the collected environmental data and the incremental feature data to obtain verification feature data. The latest collected legal feature data is obtained from the historical database, and similarity analysis and confidence conversion are performed on the legal feature data and the verification feature data to obtain a confidence score. When the confidence score is greater than the preset feature update threshold, the registered user data is updated according to the verification feature data.

[0006] In an optional implementation, the step of performing feature drift analysis on the biometric sample data according to time sequence to obtain drift data for various biometrics includes: The sample feature set is classified and sorted according to the time sequence and the preset biometric type to obtain the time sequence feature data sequence of each biometric. Based on the time-series characteristic data sequence and the collected environmental parameter set, a correlation analysis between biological characteristics and environmental parameters is performed to obtain the environmental change response coefficient of each biological characteristic; The trend analysis of each biomarker is performed based on the time-series feature data sequence to obtain trend data for each biomarker.

[0007] In an optional implementation, the incremental analysis of the drift data to obtain feature incremental data includes: Evolution analysis and fluctuation monitoring are performed on the trend data to obtain the long-term evolution component and short-term fluctuation component of each biological characteristic. The linear rate of change of the long-term evolution component is calculated based on the time span of the collection of the biometric sample data to obtain the basic increment of each biometric. Environmental attribution analysis is performed on the short-term fluctuation components based on the environmental change response coefficient to obtain environmental compensation parameters for each biological characteristic.

[0008] In an optional implementation, the step of performing nonlinear compensation on the registered user data based on the collected environmental data and the feature increment data to obtain verification feature data includes: Real-time environmental dynamic adjustment analysis is performed based on the collected environmental data and the environmental compensation parameters to obtain a set of environmental dynamic adjustment parameters; The registered user data is nonlinearly compensated based on the environmental dynamic adjustment parameter set and the basic increment to obtain verification feature data.

[0009] In an optional implementation, the step of performing nonlinear compensation on the registered user data using the dynamic feature compensation model to obtain verification feature data includes: The registered user data is dynamically compensated by the environmental compensation module to obtain environmental compensation feature data. The environmental compensation feature data is calibrated by performing feature evolution on the time-series evolution module to obtain verification feature data.

[0010] In an optional implementation, the step of performing similarity analysis and confidence conversion on the legitimate feature data and the verification feature data to obtain a confidence score includes: Based on the biometric type and the preset difference assessment method, calculate the similarity of the legitimate feature data and the verification feature data on various biometrics; The similarity is weighted and fused according to the preset fusion weight coefficient to obtain a comprehensive confidence level; Based on a preset environmental interference attenuation method, the overall confidence level is corrected and normalized to obtain a confidence score.

[0011] In an optional implementation, updating the registered user data based on the verification feature data when the confidence score is greater than a preset feature update threshold includes: When the confidence score is greater than the preset feature update threshold, feature change trend analysis is performed based on the verification feature data and the registered user data to obtain a feature evolution dataset. The verification feature data is verified based on the environmental data and the feature evolution dataset by using preset rules for changes in the relationship between the environment and physiology. When the verification feature data conforms to the rules governing changes in the relationship between environment and physiology, the verification feature data and the registration feature data in the registered user data are weighted and fused according to a preset feature fusion coefficient set to update the registered user data.

[0012] On one hand, embodiments of this application provide an adaptive optimization device for biometric identification, the device comprising: The data extraction module is used to extract biometric sample data and corresponding registered user data from a preset historical database when the number of legitimate authentications reaches a preset update trigger threshold. The transformation analysis module is used to perform feature drift analysis on the biometric sample data according to the time sequence to obtain drift data of various biometrics, and to perform incremental analysis on the drift data to obtain feature increment data. The feature compensation module is used to perform nonlinear compensation on the registered user data based on the collected environmental data and the feature increment data to obtain verification feature data. The similarity detection module is used to obtain the latest collected legal feature data from the historical database, perform similarity analysis and confidence conversion on the legal feature data and the verification feature data, and obtain a confidence score. The update decision module is used to update the registered user data based on the verification feature data when the confidence score value is greater than the preset feature update threshold.

[0013] In summary, this application includes at least the following beneficial technical effects: 1. By performing feature drift and incremental analysis on historical authentication samples, it is possible to dynamically capture changes in biometrics over time and in the environment, thereby enhancing the timeliness and matching accuracy of registration matching templates.

[0014] 2. By introducing nonlinear environmental compensation and temporal evolution models, the impact of external interferences such as light and temperature on recognition performance is effectively suppressed, thereby improving the robustness of the system.

[0015] 3. Based on similarity-weighted fusion and confidence assessment, safe screening and dynamic updating of updated features are achieved, reducing false acceptance and false rejection rates. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a flowchart of an adaptive optimization method for biometric identification provided in an embodiment of this application; Figure 2 This is a functional block diagram of an adaptive optimization device for biometric identification provided in an embodiment of this application; Figure 3 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0018] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0019] It should be noted that the message processing solution provided in this application requires special explanation of the following two points: 1. The relevant data involved in the message processing process in this application (such as registered user data, etc.). When the above embodiments of this application are applied to specific products or technologies, permission or consent from the target audience is required, and the collection, use, and processing of related data must comply with the relevant laws, regulations, and standards of the region, conforming to the principles of legality, legitimacy, and necessity, and not involving the acquisition of data types prohibited or restricted by laws and regulations. In some optional embodiments, the relevant data involved in the embodiments of this application is obtained after separate authorization from the target audience. In addition, when obtaining separate authorization from the target audience, the purpose of the relevant data is explained to the target audience.

[0020] 2. It is understood that in this application, the term "at least one" refers to one or more, and "multiple" means two or more; for example, "at least one notification method" means one, two, or more notification methods. The terms "first," "second," etc., are used to distinguish identical or similar items with essentially the same function. It should be understood that there is no logical or temporal dependency between "first," "second," and "nth," nor is there any limitation on the quantity or execution order.

[0021] like Figure 1 The diagram shows a flowchart of an adaptive optimization method for biometric recognition provided in an embodiment of this application. The adaptive optimization method for biometric recognition provided in this embodiment includes the following steps.

[0022] Step S1: When the number of valid authentications reaches the preset update trigger threshold, extract biometric sample data and corresponding registered user data from the preset historical database.

[0023] It should be understood that the number of successful authentication attempts refers to the number of times a user has successfully authenticated their identity through the smart lock. The update trigger threshold is set to a fixed value, such as fifty authentication attempts. When the system detects that a specific user's accumulated successful authentication records have reached this value, it automatically activates a deep scan program in the historical database. The historical database uses a distributed architecture for storage and contains three core sub-databases: a biometric database that records the original biometric images and extracted feature vectors for each unlock; an environmental parameter database that associates and stores real-time values ​​collected by temperature, humidity, and light sensors; and an identity tag database that maintains the binding relationship between the user's unique identifier and the device.

[0024] The extraction process first locates the target user through identity tags, then filters authentication records within the last six months using a time range filter. The biometric database returns a raw biometric set consisting of vein infrared images, fingerprint texture maps, and near-infrared and visible light images of the face; the environmental parameter database simultaneously outputs corresponding temperature values, humidity percentages, and light intensity values. Registration information refers to registered user data—this data structure contains a baseline biometric template generated during user registration or the previous update cycle, consisting of vein bifurcation point coordinate vectors, fingerprint ridge feature codes, and facial depth feature vectors, while also binding initial environmental calibration parameters. Registered user data is essentially the core reference benchmark for the system to determine user identity; its function is similar to a biometric ID card. When a user authenticates, real-time feature collection needs to be matched with this data to calculate the degree of matching.

[0025] The data cleaning module then initiates a four-stage filtering process on the extracted samples: temporal continuity detection removes redundant samples with consecutive authentications within one minute; environmental mutation point analysis eliminates abnormal data when temperature changes exceed ten degrees Celsius; feature integrity verification discards records where feature extraction for any modality fails; and liveness detection log review excludes spoofing samples. The cleaned samples are sorted by timestamp to generate biometric sample data containing vein feature sequences, fingerprint feature sequences, facial feature sequences, and environmental parameter sequences. The biometric sample data is a temporal feature set extracted from a historical database, structured as a multidimensional matrix of {timestamp, vein feature vector, fingerprint feature code, facial feature vector, temperature, humidity, illumination}. Essentially, it records the drift trajectory of biometric features in the real environment, providing training data for the compensation model. In this embodiment, the historical database employs a hybrid architecture of a time-series database (e.g., InfluxDB) and object storage (e.g., MinIO). Biometric images are stored in encrypted binary format, and feature vectors are converted to JSON format and indexed by time. Environmental parameters are synchronized in real-time via an IoT gateway, forming a traceable environment-feature mapping chain.

[0026] In this step, registered user data is copied as an independent unit and packaged together with biometric sample data into a data package. For example, after a child user goes through a growth and development cycle, the distance between the fingerprint ridges may increase by 5%, and the diameter of the veins may increase by 0.2 mm. At this time, samples extracted from the historical database over the past six months will clearly show this trend, while the initial fingerprint template in the registered user data still retains the state from six months ago. The combination of the two provides key input for subsequent drift analysis: biometric sample data reveals the trajectory of feature changes over time and environment, while registered user data serves as the benchmark anchor for change calculation. This step balances computational efficiency and the real-time nature of feature updates through a threshold triggering mechanism, avoiding system load surges caused by frequent operations, while ensuring that the dataset used for model training covers the entire seasonal cycle and typical environmental scenarios.

[0027] Step S2: Perform feature drift analysis on the biometric sample data according to the time sequence to obtain drift data of various biometrics, and perform incremental analysis on the drift data to obtain feature incremental data.

[0028] The biometric sample data comprises four core elements: vein feature vectors, fingerprint ridge codes, facial depth features, and environmental parameter sets. Vein feature vectors represent the mathematical expression of the coordinates of blood vessel bifurcation points; fingerprint ridge codes are binary feature sequences generated at ridge intersections; facial depth features are 128-dimensional codes of the three-dimensional contour; and the environmental parameter set records real-time values ​​of temperature, humidity, and illumination. After this data is input into the drift analysis engine, modality separation is first performed—the mixed data stream is split into independent channels according to biometric type, thus forming a temporal feature data sequence for each biometric. The temporal feature data sequence, arranged along the time axis, is a sequence of biometric values, such as [Day 1 vein value 0.35, Day 2 0.34, ..., Day 30 0.28], forming a winter vein curve. Essentially, this visualizes the evolution trajectory of biometrics over time. The vein feature sequence forms a curve of blood vessel diameter change, the fingerprint feature sequence constructs a curve of ridge clarity fluctuation, and the facial feature sequence generates a curve of skin texture evolution. Each curve is accompanied by a timestamp and environmental parameters to form a time-series characteristic data sequence. For example, the vein curve collected over 30 consecutive days in winter shows an overall downward trend, while the fingerprint curve drops sharply when the humidity is below 30%.

[0029] The correlation analysis module was then activated, using the Pearson correlation coefficient to calculate the correlation strength between each curve and environmental parameters. The correlation coefficient between vein features and temperature reached -0.83, indicating that for every degree Celsius decrease in temperature, the vein feature value decreased by 0.83 units; the correlation coefficient between fingerprint features and humidity was -0.91, confirming that dry environments cause ridge blurring; the correlation coefficient between facial features and light intensity was -0.95, revealing a decrease in skin texture recognition rate under strong light. The environmental change response coefficient is a structured output of these correlation coefficients, essentially quantifying the strength of biometric responses to environmental changes. Trend analysis, through linear regression fitting of curve slopes, showed that a 5% annual growth rate for the vein curve reflects child growth and development, while a 0.7% monthly decay rate for the facial curve indicates the natural aging process. These long-term patterns of change were extracted as trend data.

[0030] The obtained drift data includes environmental change response coefficients and trend data, which are then transferred to the incremental analysis stage. The evolution analysis in the incremental analysis stage employs a seasonal trend decomposition algorithm to separate the venous trend data into an annual growth component and a winter contraction component: the annual component represents the natural expansion of blood vessel diameter, while the winter component reflects the periodic contraction caused by low temperatures. Fluctuation monitoring captures sudden deviations in fingerprint data, such as an abnormal drop in feature values ​​caused by a sudden drop in humidity to 20% on a given day. The long-term evolution component is input into a rate of change calculator; if the collection time span is 180 days and the annual growth of the venous component is 5%, then the base increment is determined to be a daily increase of 0.076%. The environmental attribution analysis in the incremental analysis stage matches the fingerprint fluctuation component with the environmental change response coefficient. When a fluctuation is detected at a humidity value of 25%, combined with the fingerprint humidity sensitivity coefficient of -0.91, the environmental compensation parameter function is derived: the feature compensation amount equals 0.91 multiplied by (40% standard humidity minus the measured humidity).

[0031] The incremental feature data obtained from incremental analysis includes two types of outputs: basic increments and environmental compensation parameters. The basic increment is a rate calibration of the natural evolution of biological characteristics; for example, the average daily venous dilation in children is quantified as 0.1 mm. The environmental compensation parameters are mathematical rules for feature repair; for example, the venous compensation formula triggered by a winter environment of -5 degrees Celsius involves relaxing the validation threshold by 0.15 units.

[0032] By deconstructing the natural attributes and environmental interference of biometric changes, we provide a precise mathematical basis for dynamic compensation models, ensuring that smart locks automatically relax vein recognition requirements in extremely cold environments, adjust fingerprint matching algorithms in dry seasons, and continuously track the evolution of biometrics caused by user growth and development.

[0033] Step S3: Perform nonlinear compensation on the registered user data based on the collected environmental data and the feature increment data to obtain verification feature data.

[0034] Registered user data serves as the benchmark anchor for biometric verification, comprising three core templates: vein bifurcation point topological vectors, fingerprint ridge binary encoding, and facial 3D contour depth features, as well as temperature, humidity, and illumination calibration parameters under the initial registration environment. This data structure is essentially a mathematical representation of the user's biometrics, and features collected in real-time during the authentication process must be matched for similarity. When the environmental sensor transmits data indicating a current temperature of -5 degrees Celsius, humidity of 25%, and illumination of 300 lumens, the environmental compensation parameter set is simultaneously activated. In this embodiment, the environmental compensation parameter set includes: a vein temperature sensitivity coefficient of -0.83 indicating vasoconstriction; a fingerprint humidity sensitivity coefficient of -0.91 reflecting ridge blurring caused by dryness; and a facial illumination sensitivity coefficient of -0.95 characterizing texture recognition attenuation under strong light.

[0035] In the real-time environmental dynamic adjustment analysis, sensor data is first input into the compensation function: the vein compensation is equal to 0.83 multiplied by (current temperature minus standard temperature of 25 degrees Celsius), generating a vein environment compensation factor of -2.49; the fingerprint compensation is equal to 0.91 multiplied by (standard humidity of 40% minus measured humidity of 25%), yielding a fingerprint humidity compensation factor of 1.365; the face compensation is equal to 0.95 multiplied by (standard illumination of 500 lumens minus measured illumination of 300 lumens), generating a face illumination compensation factor of 0.19. These factors constitute the environmental dynamic adjustment parameter set, which essentially quantifies physical environmental interference into biometric correction values. The base increment is loaded synchronously as a time compensation coefficient; for example, the average daily vein expansion of 0.00041 mm for a child user is converted into a linear increase of 0.0738 mm after 180 days. Finally, a dynamic feature compensation model is constructed based on the environmental dynamic adjustment parameter set and the time compensation coefficients.

[0036] In this embodiment, the dynamic feature compensation model adopts a dual-channel architecture. One channel is the environment compensation module, which performs vector-level operations. It multiplies the vein template of the registered user data by an exponential scaling term with a compensation factor of -2.49, adds a translation term with a compensation factor of 1.365 to the fingerprint template, and applies a rotation matrix transformation with a compensation factor of 0.19 to the face template, outputting the environment-compensated feature data. The other channel is the temporal evolution module, which performs feature evolution calibration on the compensated data—adding a diameter increment of 0.0738 mm to the vein features, fusing the ridge smoothing coefficient caused by three months of natural wear and tear to the fingerprint features, and incorporating the skin texture relaxation caused by aging to the face features. The resulting dynamic feature compensation model is used for nonlinear compensation of the registered user data. This process uses a nonlinear sigmoid function to constrain the change amplitude, ensuring that the updated vein features do not exceed 20% of the original value, the fingerprint feature change rate is limited to within 5% monthly decay, and the face features retain 90% of their historical weights.

[0037] Ultimately, the generated verification feature data is a dynamically optimized version of the registration template. In this embodiment, the original vein bifurcation point coordinates are expanded by 15% after low-temperature compensation, and the overall offset is 0.1 mm after adding growth and development increments; the initial fingerprint ridge code is enhanced in edge sharpness by 30% through drying compensation, and the feature point contrast is reduced by combining natural wear adjustment; the registered facial depth features restore details in the shadow area through illumination compensation, while simultaneously adding an age-related cheekbone contour softening coefficient. This data can be directly used for subsequent authentication matching. For example, when a child unlocks the phone on a winter morning, the system automatically calls this compensated template instead of the original data registered six months ago or the feature data obtained in the previous update cycle, eliminating vein recognition failures caused by low temperatures and fingerprint rejections caused by growth and development.

[0038] Step S4: Obtain the latest collected legal feature data from the historical database, perform similarity analysis and confidence conversion on the legal feature data and the verification feature data to obtain a confidence score.

[0039] It should be understood that the latest collected legitimate feature data refers to the original biometric set from the user's most recent successful authentication, including uncompensated vein infrared image feature vectors, fingerprint ridge binary codes, face depth codes, and real-time environmental parameters. This data is retrieved in real-time from a historical database, for example, a record of unlocking in the early morning: the vein feature shrinks by 15% due to low temperature, resulting in a coordinate set shift; the fingerprint feature's ridge blurring increases by 30% due to dryness; and the face feature's three-dimensional contour is distorted under direct morning light. The verification feature data is the dynamically optimized template generated in step S3 above—the vein template has expanded its spacing by 15% in low-temperature environments, the fingerprint template has enhanced edge sharpness by 30% in dry scenes, and the face template has restored shadow details in strong light environments. After both are input into the similarity analysis engine, they are processed separately according to biometric type.

[0040] It should be understood that the embodiments of this application employ different difference assessment methods to calculate the similarity between two biometric features for each type. For vein features, a modified Hausdorff distance algorithm is used to calculate similarity, measuring the spatial overlap rate of the vascular bifurcation point set before and after compensation. A -5°C environment increases the distance between matching points of the original vein feature set and the compensation template by 0.2 mm, resulting in a similarity of 0.75 after distance normalization. For fingerprint features, a difference bit rate is generated through XOR operation on the ridge code. In a dry environment, 40% of the original ridge code points are blurred, resulting in a 25% difference rate after comparison with the compensation template, leading to a similarity of 0.75. For facial features, a weighted cosine similarity is used. Under strong light, the original depth-encoded periorbital texture area is 15% missing, and the similarity reaches 0.92 after repair of the corresponding area in the compensation template. Essentially, the difference assessment method is a mathematical matching strategy adapted to the physical characteristics of biometric features: spatial distance measurement is suitable for vascular networks, bit difference rate is suitable for binary ridge codes, and local weighted vector similarity is suitable for 3D facial features.

[0041] Furthermore, the weighted fusion module loads preset fusion weight coefficients. In this embodiment, the weight for vein features is 0.2 (easily interfered with at low temperatures), the weight for fingerprints is 0.4 (sensitive to dryness), and the weight for faces is 0.4 (significantly affected by lighting). After obtaining the fusion weight coefficients and the similarities of the various biometric features mentioned above, the similarities are weighted and fused according to the fusion weight coefficients to obtain a comprehensive confidence score. In this embodiment, the comprehensive confidence score is calculated as follows: vein similarity 0.75 x 0.2 + fingerprint similarity 0.75 x 0.4 + face similarity 0.92 x 0.4, resulting in an initial value of 0.832. Subsequently, the environmental interference attenuation method is activated. When the temperature sensor returns a reading of -5 degrees Celsius, the vein attenuation compensation function is activated—the similarity is increased by 0.15 multiplied by the temperature deviation value (25 minus -5) divided by 10, outputting a vein correction factor of 0.45. The humidity sensor detects a value of 20%, triggering fingerprint attenuation compensation. The function multiplies 0.3 by (standard humidity of 40 minus measured humidity of 20) divided by 10 to generate a fingerprint correction factor of 0.6. The light sensor's 900-lumen reading activates a face correction factor of 0.15. The total environmental correction is 0.45 plus 0.6 plus 0.15 equals 1.2, and the overall confidence level is increased to 0.952 after correction.

[0042] Finally, the corrected overall confidence score is constrained to the range of 0 to 1 using the sigmoid function. Specifically, the correction value of 0.952 is input into the sigmoid function—divided by (1 plus e^(-10) multiplied by (x - 0.8))—and the final score is output as 0.987. The confidence score directly determines the subsequent update logic. For example, in this authentication scenario on a winter morning, the low temperature and dry environment resulted in a 60% risk of initial matching failure. The high score after environmental correction not only avoids falsely rejecting legitimate users but also blocks attacks that exploit environmental anomalies. The confidence score is essentially a dynamic indicator that quantifies the reliability of biometrics. Its correction mechanism accurately offsets the similarity distortion caused by environmental interference, ensuring that the system maintains a pass rate of over 95% even in blizzard conditions.

[0043] Step S5: When the confidence score is greater than the preset feature update threshold, update the registered user data according to the verification feature data.

[0044] In this embodiment of the application, when the confidence score reaches the preset feature update threshold of 0.85, the registered user data update process is triggered.

[0045] First, feature change trend analysis is performed based on the verification feature data and registered user data. During this analysis, the difference in vein feature vectors, Hamming distance of fingerprint ridge codes, and cosine similarity offset of facial depth features are extracted from the verification feature data and registered user data to generate a feature evolution dataset. For example, in the verification of vein features for children, the vessel diameter increased by 0.15 mm compared to the registered template, the fingerprint ridge spacing increased by 3%, and the facial cheekbone contour softening coefficient increased by 0.07. The feature evolution dataset is a quantitative record of biometric changes, for example, with the structure {vein diameter change: +0.15 mm, fingerprint ridge spacing change: +3%, facial contour softening: +0.07}. The purpose of the feature evolution dataset is to transform biometric drift into verifiable mathematical vectors, providing input for physiological rule verification.

[0046] Furthermore, a pre-defined rule base for environmental and physiological correlation changes is then used for verification. These rules are environmental constraints on biometric evolution, such as "when the temperature is <5℃, the rate of change in veins must be ∈ [-0.2mm, 0]", which essentially prevents attackers from using extreme environments to forge biometric evolution. The environmental and physiological correlation change rules in this embodiment include: the rule "When the vein contraction rate is >5% in a low-temperature environment, positive vein feature growth is prohibited" matches the current temperature of -5 degrees Celsius, and a vein growth of 0.15 mm triggers a rule conflict alarm; the rule "When the fingerprint ridge clarity attenuation rate is >10% in a dry environment, negative feature shift is allowed" matches the measured humidity of 20%, and a fingerprint feature attenuation of 5% passes verification; the rule "When the facial texture contrast decrease rate is <8% in a strong light environment, skin relaxation and updating are rejected" successfully matches a facial feature shift of 0.07.

[0047] During the verification of feature data using the aforementioned rules relating environmental and physiological changes, weighted fusion is initiated upon successful verification. The feature fusion coefficient set is loaded with differential rate parameters: a vein feature fusion coefficient of 0.9 undergoes a strong constraint update, where the new vein template equals the original template multiplied by 0.9 plus the verification feature multiplied by 0.1; a fingerprint feature fusion coefficient of 0.7 undergoes a moderate update, where the new fingerprint template equals the original template multiplied by 0.7 plus the verification feature multiplied by 0.3; and a face feature fusion coefficient of 0.5 undergoes a fast update, where the new face template equals the original template multiplied by 0.5 plus the verification feature multiplied by 0.5. For example, after updating a child's vein feature, the diameter increases by a net 0.015 mm, reflecting growth and development trends while avoiding mutation risks; after fingerprint feature fusion, the ridge spacing increases by 0.9%, eliminating temporary deformation records caused by dry environments; and the face feature retains 50% of historical data while incorporating current lighting optimization results.

[0048] The updated registered user data generates a timestamped version snapshot, which is stored using a blockchain-style architecture, linking the parent version hash value with the environment context. The authentication counter is synchronously reset to zero and restarted for accumulation. When the average confidence score of three out of the subsequent twenty authentications falls below 0.7, the version rollback mechanism is automatically triggered to restore the system to the state before the update. This ensures that temporary feature distortions caused by dry, cracked fingers in elderly users during winter do not contaminate the biometric template, while physiological changes during children's growth are gradually integrated into the system, maintaining the core capability of the smart lock to be lifelong and registration-free.

[0049] This application is applied to the field of feature recognition technology. After an update is triggered, biometric sample data and corresponding registered user data are extracted from a historical database. Feature drift and incremental analysis are performed on the biometric sample data according to time sequence to obtain incremental feature data. Registered user data is compensated based on collected environmental data and incremental feature data to obtain verification feature data. The latest collected legal feature data is retrieved from the historical database. Similarity analysis and confidence conversion are performed between the legal feature data and the verification feature data to obtain a confidence score. When the confidence score is greater than the feature update threshold, the registered user data is updated based on the verification feature data. This application, based on feature drift and incremental analysis of historical samples, nonlinear environmental and temporal evolution compensation, similarity-weighted fusion, and confidence assessment to trigger dynamic updates, adapts in real time to changes in biometrics and the environment, achieving high accuracy of the recognition template, adaptive updates, and improved system robustness.

[0050] The following describes the relevant apparatus for the adaptive optimization scheme of biometric recognition provided in the embodiments of this application.

[0051] It should be noted that, in the embodiments of this application, the terms "module" or "unit" refer to a computer program or part of a computer program with a predetermined function, which works together with other related parts to achieve a predetermined goal, and can be implemented wholly or partially using software, hardware (such as processing circuitry or memory), or a combination thereof. Similarly, a processor (or multiple processors or memory) can be used to implement one or more modules or units. Furthermore, each module or unit can be part of an overall module or unit that includes the functionality of that module or unit.

[0052] Please see Figure 2 This is a functional block diagram of an adaptive optimization device for biometric recognition provided in an embodiment of this application. This adaptive optimization device 2 for biometric recognition can be used to perform the functions described in this application. Figure 1 The corresponding steps in the adaptive optimization method for biometric recognition provided in the embodiment. Specifically, the adaptive optimization device 2 for biometric recognition may include: Data extraction module 21 is used to extract biometric sample data and corresponding registered user data from a preset historical database when the number of legitimate authentications reaches a preset update trigger threshold. The transformation analysis module 22 is used to perform feature drift analysis on the biometric sample data according to the time sequence to obtain drift data of various biometrics, and to perform incremental analysis on the drift data to obtain feature incremental data. Feature compensation module 23 is used to perform nonlinear compensation on the registered user data based on the collected environmental data and the feature increment data to obtain verification feature data; Similarity detection module 24 is used to obtain the latest collected legal feature data from the historical database, perform similarity analysis and confidence conversion on the legal feature data and the verification feature data, and obtain a confidence score value. The update decision module 25 is used to update the registered user data based on the verification feature data when the confidence score value is greater than the preset feature update threshold.

[0053] In one possible implementation, the transformation analysis module 22 is also used to perform the following operations: The sample feature set is classified and sorted according to the time sequence and the preset biometric type to obtain the time sequence feature data sequence of each biometric. Based on the time-series characteristic data sequence and the collected environmental parameter set, a correlation analysis between biological characteristics and environmental parameters is performed to obtain the environmental change response coefficient of each biological characteristic; The trend analysis of each biomarker is performed based on the time-series feature data sequence to obtain trend data for each biomarker.

[0054] In one possible implementation, the transformation analysis module 22 is also used to perform the following operations: Evolution analysis and fluctuation monitoring are performed on the trend data to obtain the long-term evolution component and short-term fluctuation component of each biological characteristic. The linear rate of change of the long-term evolution component is calculated based on the time span of the collection of the biometric sample data to obtain the basic increment of each biometric. Environmental attribution analysis is performed on the short-term fluctuation components based on the environmental change response coefficient to obtain environmental compensation parameters for each biological characteristic.

[0055] In one possible implementation, the feature compensation module 23 is also used to perform the following operations: Real-time environmental dynamic adjustment analysis is performed based on the collected environmental data and the environmental compensation parameters to obtain a set of environmental dynamic adjustment parameters; The registered user data is nonlinearly compensated based on the environmental dynamic adjustment parameter set and the basic increment to obtain verification feature data.

[0056] In one possible implementation, the feature compensation module 23 is also used to perform the following operations: The registered user data is dynamically compensated by the environmental compensation module to obtain environmental compensation feature data. The environmental compensation feature data is calibrated by performing feature evolution on the time-series evolution module to obtain verification feature data.

[0057] In one possible implementation, the similarity detection module 24 is also used to perform the following operations: Based on the biometric type and the preset difference assessment method, calculate the similarity of the legitimate feature data and the verification feature data on various biometrics; The similarity is weighted and fused according to the preset fusion weight coefficient to obtain a comprehensive confidence level; Based on a preset environmental interference attenuation method, the overall confidence level is corrected and normalized to obtain a confidence score.

[0058] In one possible implementation, the update decision module 25 is also used to perform the following operations: When the confidence score is greater than the preset feature update threshold, feature change trend analysis is performed based on the verification feature data and the registered user data to obtain a feature evolution dataset. The verification feature data is verified based on the environmental data and the feature evolution dataset by using preset rules for changes in the relationship between the environment and physiology. When the verification feature data conforms to the rules governing changes in the relationship between environment and physiology, the verification feature data and the registration feature data in the registered user data are weighted and fused according to a preset feature fusion coefficient set to update the registered user data.

[0059] It should be understood that the various variations and specific embodiments of the methods provided in the above embodiments are also applicable to the adaptive optimization device for biometric recognition in this embodiment. Through the foregoing detailed description of the adaptive optimization method for biometric recognition, those skilled in the art can clearly understand the implementation method of the adaptive optimization device for biometric recognition in this embodiment. For the sake of brevity, it will not be described in detail here.

[0060] Please see Figure 3This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. The computer device 3 is used to execute the steps performed by the computer device in the aforementioned method embodiments. The computer device 3 may include one or more devices (e.g., a server, node, terminal device, etc.) or internal components (e.g., a chip, software module, or hardware module). The computer device may include at least one processor 31 and a communication interface 32. Further optionally, the computer device may also include at least one memory 33 and a bus 34. Additionally, the processor 31, communication interface 32, and memory 33 are connected via the bus 34. Wherein: (1) The processor 31 is a module that performs arithmetic and / or logical operations. Specifically, it may be one or a combination of processing modules such as a central processing unit (CPU), a graphics processing unit (GPU), a microprocessor unit (MPU), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), a complex programmable logic device (CPLD), a coprocessor (to assist the central processing unit in completing corresponding processing and applications), and a micro controller unit (MCU).

[0061] (2) The communication interface 32 can be used to provide information input or output to at least one processor 31. And / or, the communication interface 32 can be used to receive data sent from outside and / or send data to outside, and can be a wired link interface including such as an Ethernet cable, or a wireless link interface (Wi-Fi, Bluetooth, general wireless transmission, vehicle short-range communication technology and other short-range wireless communication technologies, etc.). The communication interface 32 can serve as a network interface.

[0062] (3) The memory 33 is used to provide storage space, in which data such as the operating system and computer programs (including program instructions) can be stored. The memory 33 can be one or a combination of random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), or compact disc read-only memory (CD-ROM), etc.

[0063] In specific implementation, processor 31 executes the following steps by running the computer program stored in memory 33: When the number of valid authentications reaches the preset update trigger threshold, biometric sample data and corresponding registered user data are extracted from the preset historical database. The biometric sample data is subjected to feature drift analysis according to time sequence to obtain drift data of various biometrics, and the drift data is subjected to incremental analysis to obtain feature incremental data. The registered user data is nonlinearly compensated based on the collected environmental data and the incremental feature data to obtain verification feature data. The latest collected legal feature data is obtained from the historical database, and similarity analysis and confidence conversion are performed on the legal feature data and the verification feature data to obtain a confidence score. When the confidence score is greater than the preset feature update threshold, the registered user data is updated according to the verification feature data.

[0064] In one possible implementation, processor 31 is also used to perform the following operations: The sample feature set is classified and sorted according to the time sequence and the preset biometric type to obtain the time sequence feature data sequence of each biometric. Based on the time-series characteristic data sequence and the collected environmental parameter set, a correlation analysis between biological characteristics and environmental parameters is performed to obtain the environmental change response coefficient of each biological characteristic; The trend analysis of each biomarker is performed based on the time-series feature data sequence to obtain trend data for each biomarker.

[0065] In one possible implementation, processor 31 is also used to perform the following operations: Evolution analysis and fluctuation monitoring are performed on the trend data to obtain the long-term evolution component and short-term fluctuation component of each biological characteristic. The linear rate of change of the long-term evolution component is calculated based on the time span of the collection of the biometric sample data to obtain the basic increment of each biometric. Environmental attribution analysis is performed on the short-term fluctuation components based on the environmental change response coefficient to obtain environmental compensation parameters for each biological characteristic.

[0066] In one possible implementation, processor 31 is also used to perform the following operations: Real-time environmental dynamic adjustment analysis is performed based on the collected environmental data and the environmental compensation parameters to obtain a set of environmental dynamic adjustment parameters; The registered user data is nonlinearly compensated based on the environmental dynamic adjustment parameter set and the basic increment to obtain verification feature data.

[0067] In one possible implementation, processor 31 is also used to perform the following operations: The registered user data is dynamically compensated by the environmental compensation module to obtain environmental compensation feature data. The environmental compensation feature data is calibrated by performing feature evolution on the time-series evolution module to obtain verification feature data.

[0068] In one possible implementation, processor 31 is also used to perform the following operations: Based on the biometric type and the preset difference assessment method, calculate the similarity of the legitimate feature data and the verification feature data on various biometrics; The similarity is weighted and fused according to the preset fusion weight coefficient to obtain a comprehensive confidence level; Based on a preset environmental interference attenuation method, the overall confidence level is corrected and normalized to obtain a confidence score.

[0069] In one possible implementation, processor 31 is also used to perform the following operations: When the confidence score is greater than the preset feature update threshold, feature change trend analysis is performed based on the verification feature data and the registered user data to obtain a feature evolution dataset. The verification feature data is verified based on the environmental data and the feature evolution dataset by using preset rules for changes in the relationship between the environment and physiology. When the verification feature data conforms to the rules governing changes in the relationship between environment and physiology, the verification feature data and the registration feature data in the registered user data are weighted and fused according to a preset feature fusion coefficient set to update the registered user data.

[0070] In one possible implementation, processor 31 is also used to perform the following operations: The data extraction module is used to extract biometric sample data and corresponding registered user data from a preset historical database when the number of legitimate authentications reaches a preset update trigger threshold. The transformation analysis module is used to perform feature drift analysis on the biometric sample data according to the time sequence to obtain drift data of various biometrics, and to perform incremental analysis on the drift data to obtain feature increment data. The feature compensation module is used to perform nonlinear compensation on the registered user data based on the collected environmental data and the feature increment data to obtain verification feature data. The similarity detection module is used to obtain the latest collected legal feature data from the historical database, perform similarity analysis and confidence conversion on the legal feature data and the verification feature data, and obtain a confidence score. The update decision module is used to update the registered user data based on the verification feature data when the confidence score value is greater than the preset feature update threshold.

[0071] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially as a computer program product, which includes one or more computer programs. When the computer program is loaded and executed on a computer device, it generates, in whole or in part, the processes or functions described in the embodiments of this application; the computer device can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer program can be stored in or transmitted through a computer-readable storage medium; the computer program can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium accessible to the computer device or a data processing device such as a server or data center that integrates one or more available media; wherein, the available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state disk (SSD)).

[0072] The above-disclosed embodiments are merely preferred embodiments of this application and should not be construed as limiting the scope of this application. Therefore, any equivalent variations made in accordance with the claims of this application shall still fall within the scope of this application.

Claims

1. An adaptive optimization method for biometric identification, characterized in that, The method includes: When the number of valid authentications reaches the preset update trigger threshold, biometric sample data and corresponding registered user data are extracted from the preset historical database. The biometric sample data is subjected to feature drift analysis according to time sequence to obtain drift data of various biometrics, and the drift data is subjected to incremental analysis to obtain feature incremental data. The registered user data is nonlinearly compensated based on the collected environmental data and the incremental feature data to obtain verification feature data. The latest collected legal feature data is obtained from the historical database, and similarity analysis and confidence conversion are performed on the legal feature data and the verification feature data to obtain a confidence score. When the confidence score is greater than the preset feature update threshold, the registered user data is updated according to the verification feature data.

2. The adaptive optimization method for biometric recognition according to claim 1, characterized in that, The biometric sample data includes a sample feature set and a set of environmental parameters for each feature; the drift data includes environmental change response coefficients and trend data; the step of performing feature drift analysis on the biometric sample data according to time sequence to obtain drift data for various biometrics includes: The sample feature set is classified and sorted according to the time sequence and the preset biometric type to obtain the time sequence feature data sequence of each biometric. Based on the time-series characteristic data sequence and the collected environmental parameter set, a correlation analysis between biological characteristics and environmental parameters is performed to obtain the environmental change response coefficient of each biological characteristic; The trend analysis of each biomarker is performed based on the time-series feature data sequence to obtain trend data for each biomarker.

3. The adaptive optimization method for biometric recognition according to claim 1, characterized in that, The drift data includes environmental change response coefficients and trend data; The incremental feature data includes basic increments and environmental compensation parameters; the incremental analysis of the drift data to obtain the incremental feature data includes: Evolution analysis and fluctuation monitoring are performed on the trend data to obtain the long-term evolution component and short-term fluctuation component of each biological characteristic. The linear rate of change of the long-term evolution component is calculated based on the time span of the collection of the biometric sample data to obtain the basic increment of each biometric. Environmental attribution analysis is performed on the short-term fluctuation components based on the environmental change response coefficient to obtain environmental compensation parameters for each biological characteristic.

4. The adaptive optimization method for biometric recognition according to claim 1, characterized in that, The incremental feature data includes a base increment and environmental compensation parameters; the nonlinear compensation of the registered user data based on the collected environmental data and the incremental feature data to obtain verification feature data includes: Real-time environmental dynamic adjustment analysis is performed based on the collected environmental data and the environmental compensation parameters to obtain a set of environmental dynamic adjustment parameters; The registered user data is nonlinearly compensated based on the environmental dynamic adjustment parameter set and the basic increment to obtain verification feature data.

5. The adaptive optimization method for biometric recognition according to claim 4, characterized in that, The dynamic feature compensation model includes an environmental compensation module and a temporal evolution module; the nonlinear compensation of the registered user data using the dynamic feature compensation model to obtain verification feature data includes: The registered user data is dynamically compensated by the environmental compensation module to obtain environmental compensation feature data. The environmental compensation feature data is calibrated by performing feature evolution on the time-series evolution module to obtain verification feature data.

6. The adaptive optimization method for biometric recognition according to claim 2, characterized in that, The step of performing similarity analysis and confidence conversion between the legitimate feature data and the verification feature data to obtain a confidence score includes: Based on the biometric type and the preset difference assessment method, calculate the similarity of the legitimate feature data and the verification feature data on various biometrics; The similarity is weighted and fused according to the preset fusion weight coefficient to obtain a comprehensive confidence level; Based on a preset environmental interference attenuation method, the overall confidence level is corrected and normalized to obtain a confidence score.

7. The adaptive optimization method for biometric recognition according to claim 1, characterized in that, The step of updating the registered user data based on the verification feature data when the confidence score value is greater than the preset feature update threshold includes: When the confidence score is greater than the preset feature update threshold, feature change trend analysis is performed based on the verification feature data and the registered user data to obtain a feature evolution dataset. The verification feature data is verified based on the environmental data and the feature evolution dataset by using preset rules for changes in the relationship between the environment and physiology. When the verification feature data conforms to the rules governing changes in the relationship between environment and physiology, the verification feature data and the registration feature data in the registered user data are weighted and fused according to a preset feature fusion coefficient set to update the registered user data.

8. An adaptive optimization device for biometric identification, applied to the adaptive optimization method for biometric identification as described in claim 1, characterized in that, The device includes: The data extraction module is used to extract biometric sample data and corresponding registered user data from a preset historical database when the number of legitimate authentications reaches a preset update trigger threshold. The transformation analysis module is used to perform feature drift analysis on the biometric sample data according to the time sequence to obtain drift data of various biometrics, and to perform incremental analysis on the drift data to obtain feature increment data. The feature compensation module is used to perform nonlinear compensation on the registered user data based on the collected environmental data and the feature increment data to obtain verification feature data. The similarity detection module is used to obtain the latest collected legal feature data from the historical database, perform similarity analysis and confidence conversion on the legal feature data and the verification feature data, and obtain a confidence score. The update decision module is used to update the registered user data based on the verification feature data when the confidence score value is greater than the preset feature update threshold.

9. A computer device, characterized in that, include: Memory and processor: A memory, wherein one or more computer programs are stored; A processor for loading one or more computer programs to implement the adaptive optimization method for biometric identification as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the adaptive optimization method for biometric identification according to any one of claims 1 to 7.

Images