Biometric user authentication method, user visual training device and system
By combining biometric identification methods with multimodal biometrics and data encryption technology, the security and personalization issues of identity authentication in visual training devices are solved, achieving highly secure and efficient visual training management.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- ZHENGZHOU HUNUOKANG VISION TECHNOLOGY CO LTD
- Filing Date
- 2026-03-12
- Publication Date
- 2026-06-12
AI Technical Summary
Existing user authentication methods for visual training equipment suffer from insufficient security, susceptibility to loss and theft, shared use by multiple users, lack of personalized adaptation and data protection, resulting in cumbersome operation, low efficiency and easy data leakage.
The system employs biometric identification methods, reads user identity identifiers via NFC, collects facial and iris features using a multimodal biometric module, dynamically fuses feature vectors for authentication, adaptively adjusts training parameters, implements tiered fatigue warnings, and encrypts and transmits training data back to the system.
It achieves high security, accuracy, and personalized adaptation of user authentication, ensures the security and scientific nature of training data, and improves the consistency and effectiveness of visual training.
Smart Images
Figure CN122196993A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to intelligent eye vision training equipment and the field of biometrics, specifically to a biometric identification user authentication method, user vision training equipment, and system. Background Technology
[0002] As an important tool for improving visual function and correcting visual impairment, visual training equipment has been widely used in ophthalmic clinical rehabilitation, daily visual ability improvement and other scenarios. When using the equipment, users need to match a personalized training plan based on their personal visual basic indicators, training progress and physiological limitations. At the same time, the visual physiological data and training effect data generated during the training process are all sensitive personal information, which puts forward high requirements for the accuracy of identity authentication, personalized adaptation of training plan and data security. Existing visual training equipment still commonly uses traditional methods for user authentication and training management, such as manual information input, universal account password login, or simple proximity card recognition. While these methods can achieve basic user differentiation and training program invocation, they have gradually revealed many problems in practical applications due to limitations in technical design. They can no longer meet the core requirements for security, personalization, and data protection in visual training scenarios. The specific defects are as follows: The authentication methods of existing vision training devices have many drawbacks. They rely on non-biometric identifiers such as passwords and universal proximity cards, which are easily lost, stolen, or shared by multiple people, leading to ineffective identity verification and potential security risks of impersonation. Furthermore, there is no accurate automatic identity verification and file association mechanism, requiring manual retrieval of files and parameter settings, which cannot be adapted to personalized training programs. The cumbersome operation affects the continuity and effectiveness of training. At the same time, the lack of an identity authentication encryption mechanism means that sensitive physiological data transmission and storage are not bound to accurate identities, making them prone to misuse, tampering, and privacy leaks. Manual operation also makes the use of equipment and data management inefficient, increases operating and maintenance costs, and is prone to data statistical errors.
[0003] To address the aforementioned shortcomings, a technical solution is provided. Summary of the Invention
[0004] The purpose of this invention is to address the problems of insufficient user authentication security in visual training devices, lack of personalized training schemes, and risk of data leakage, and proposes a biometric identification user authentication method, user visual training device, and system.
[0005] Biometric user authentication methods include: S1. Identity Acquisition and Wake-up Trigger: The user identification information in the user identification medium is acquired through the NFC reader, and multimodal biometric identification is triggered based on the identification validity verification result; S2. Biometrics Acquisition and Evaluation: The multimodal biometrics module sequentially collects the user's facial and iris feature data, performs liveness detection and quality evaluation on the collected data, and generates a feature acquisition quality index. S3. Identity Authentication Decision: Based on a dynamic weight fusion strategy, the facial feature vector and iris feature vector are fused and matched with the pre-stored user biometric template for verification. The authentication decision result is generated by combining multi-dimensional security indicators. S4. Data Loading and Device Execution: After authentication, the personalized training file bound to the user's identification information is retrieved from the database, and the visual training device parameters are automatically configured based on the user's historical training status and visual function evaluation parameters. S5. Data Recording and Encrypted Backhaul: After training, the training data is hierarchically encrypted and associated with user identification information, then backhauled to the database to update the individual training profile, and subsequent training plans are intelligently optimized based on the accumulated data.
[0006] Furthermore, the specific operation steps of S1 are as follows: The system establishes near-field communication with the user identification medium via an NFC reader, reads the encrypted user identification information stored in the medium, and obtains an identification validity index by combining time validity, user permission level, and user historical usage credibility score. When the identification validity index is greater than a preset threshold, the system wakes up the multimodal biometric module and starts the high-definition camera autofocus and infrared iris collector light source preheating program. Simultaneously, based on the user permission level matching verification mode, including iris priority mode, face priority mode and face and iris dual-modal parallel mode, the timestamp, device identifier and geographical location information of this identity trigger event are recorded at the same time to generate access log entries.
[0007] Furthermore, the specific operation steps of S2 include: The system captures user facial image sequences using a high-definition camera, performs multi-dimensional liveness detection on the image sequences, detects natural blinking behavior and calculates blinking frequency, analyzes the amplitude of facial micro-expression changes using optical flow, reconstructs a three-dimensional facial depth map based on structured light projection imaging and calculates the depth consistency index, and obtains the face liveness index by weighted fusion of skin texture authenticity score. If the face liveness index exceeds the preset threshold, the frame corresponding to the maximum face liveness index is selected as the optimal frame. Facial feature points are extracted through a deep convolutional neural network and a multidimensional face feature vector is generated. At the same time, the optimal frame is evaluated to obtain the face acquisition quality index. Simultaneously, an infrared iris scanner captures iris image sequences under near-infrared light illumination, and the annular iris region is segmented by pupil localization and iris boundary separation. Based on the Daugman rubber sheet model, the circular iris region is geometrically normalized and unfolded into a fixed-size rectangular texture image. Then, the iris texture features are extracted through a two-dimensional Gabor filter to generate iris feature codes. At the same time, the quality of the image frames with generated feature codes is evaluated to obtain the iris acquisition quality index.
[0008] Furthermore, the specific operation steps of S3 include: The fusion weights are dynamically calculated based on the face acquisition quality index and the iris acquisition quality index, and the fusion feature vector is generated by concatenating the face feature vector and the iris feature code. Retrieve pre-stored biometric templates from a secure encrypted database based on the user's unique identifier (UID), calculate the facial feature similarity and the Hamming distance of the iris feature, and then weight the facial and iris consistency verification scores to obtain a comprehensive matching score. Simultaneously, a comprehensive security assessment index is obtained by combining the face liveness index, iris capture quality index, user historical usage credibility score, and environmental security score. A primary authentication threshold and a security threshold are preset. If the comprehensive matching score is greater than the primary authentication threshold and the comprehensive security assessment index is greater than the security threshold, the authentication is successful, an authorization token is generated, and the authentication success log is recorded. If the comprehensive matching score is between the security threshold and the primary authentication threshold, no more than 3 supplementary verifications are triggered. If the comprehensive matching score is less than the security threshold or the supplementary verification limit is exceeded, the authentication is rejected, the process is terminated, and the failure information is recorded.
[0009] Furthermore, the specific operation steps of S4 are as follows: After authentication, based on the authorization token and the user's unique identification code, the personalized training file is retrieved from the cloud database or local encrypted cache. Multiple training records are extracted to calculate the user's training status index, and visual status parameters are extracted. Combined with the training progress percentage and visual function improvement rate, a personalized adjustment factor is obtained. After adaptively adjusting the basic parameters of the recommended training module for the day, the configuration parameters are encapsulated into a control instruction set and sent to the actuator to complete operations such as training mode switching.
[0010] Simultaneously, various thresholds and markers from the physiological safety restriction parameter set are loaded. The eye-tracking module collects user eye status data and calculates the gaze stability index and real-time fatigue index. A three-level fatigue warning mechanism is implemented. If a special medical restriction marker is detected to be valid, the enhanced monitoring mode is immediately activated, the eye status data is uploaded to the medical monitoring terminal, and a reminder notification is sent to the preset contact person when an abnormality occurs.
[0011] Furthermore, the specific operation steps of S5 are as follows: After the visual training session ends, complete data of the training is collected, including actual training duration, training parameter records at each stage, training completion rate, eye state change curve during training, and user subjective feedback scores. The training data is then structured and encapsulated to generate a training data package, and training timestamps, device identifiers, software version numbers, and data integrity check codes are added to the data package as metadata. A tiered encryption strategy is implemented for training data based on data sensitivity levels. The original eye state data is encrypted using the AES256GCM algorithm, while the training statistics data is encrypted using the national cryptographic SM4 algorithm. The encrypted data packets are transmitted back to the cloud database via the TLS1.3 secure communication protocol. At the same time, two-way certificate verification and data integrity verification are performed. After decryption, the cloud database updates the user's personalized training profile and analyzes the trend of visual function changes based on the accumulated training data through a machine learning model. The personalized training plan for the next stage is optimized and synchronized to the user profile for automatic loading during the next authentication login.
[0012] A second aspect of the present invention provides a biometric recognition user visual training device, comprising: Identity wake-up triggering device: Equipped with an NFC reader, it reads the encrypted information of the user's identification medium and calculates the identification validity index. After triggering wake-up, it matches the biometric verification mode based on user permissions and records the access log simultaneously. Biometric data acquisition equipment: Equipped with a high-definition camera and an infrared iris scanner, it completes face and iris image acquisition and liveness detection, quality assessment, extracts feature vectors and feature codes, and outputs acquisition quality index; Authentication decision processing equipment: It has a built-in secure encrypted database and computing unit, calculates feature fusion weights and generates fusion feature vectors, compares them with pre-stored templates, and determines the authentication result based on a set threshold, triggering decision instructions; Training execution control equipment: After certification, it retrieves personalized training files, adjusts training parameters and issues instructions, monitors eye status through eye tracking, triggers fatigue warnings at different levels and executes training control; Data encryption and transmission device: Collects and encapsulates all training data, performs differentiated encryption, and transmits it back to the cloud via TLS 1.3 protocol to complete file updates and training plan optimization.
[0013] A third aspect of the present invention provides a biometric recognition user visual training system, comprising: Identity wake-up module: Equipped with an NFC near-field communication unit, it reads the encrypted information of the user's identification medium and calculates the identification validity index. Once the index is met, it wakes up the multimodal biometric identification module and records the access log simultaneously based on the permission matching verification mode. Feature acquisition and evaluation module: Integrates high-definition camera and infrared iris collector to acquire face and iris image sequences, extract feature vectors and feature codes through detection and analysis, and generate acquisition quality index; Authentication decision module: It has a built-in secure encrypted database and computing engine. It generates a fused feature vector based on feature fusion weights, compares it with a pre-stored template, and combines the comprehensive security assessment index and a set threshold to determine the authentication result. Training execution module: After authentication, the module retrieves the personalized training file, adaptively adjusts the training parameters and sends them out for execution, monitors the eye status through eye tracking, and provides graded early warning and control based on the real-time fatigue index; Data encryption and backhaul module: Collects and encapsulates all training data and encrypts it differently. It then transmits the data back to the cloud via the TLS 1.3 protocol. After updating the archive, the machine learning model analyzes visual trends and optimizes the training plan.
[0014] Compared with the prior art, the beneficial effects of the present invention are: This invention triggers identity wake-up via NFC near-field communication and matches verification modes according to permissions. It collects facial and iris features through multi-dimensional liveness detection and completes quality assessment. After dynamically fusing features and combining them with a security index, it achieves accurate authentication decisions. After successful authentication, it adaptively adjusts visual training parameters and implements a three-level fatigue warning. After training data is encrypted and transmitted back in a hierarchical manner, it optimizes personalized training plans. This significantly improves the security, accuracy, and adaptability of user authentication, realizes personalized, intelligent, and secure management of visual training, ensures the security of training data transmission and storage, and effectively avoids physiological risks in visual training through real-time eye monitoring, thereby improving the scientific nature and effectiveness of visual training. Attached Figure Description
[0015] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation
[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0017] Example: like Figure 1 As shown, the biometric identification user authentication method includes identity acquisition and wake-up triggering, biometric collection and evaluation, identity authentication decision, data loading and device execution, and data recording and encrypted transmission.
[0018] S1. Identity verification and wake-up trigger: When a user brings the identification medium close to the sensing area of the NFC reader, the NFC reader establishes a near-field communication connection with the identification medium through radio frequency signals and reads the encrypted user identification information from the identification medium. The user identification information includes the user's unique identifier (UID), registration timestamp, and permission level identifier. Through formula The identification effectiveness index is calculated, where, Indicates the current timestamp. Indicates the registration timestamp. Indicates the validity period of the user's identity identifier. Indicates the user's permission level. This represents a user's historical usage credibility score, derived from statistics on past usage records and unusual behavior. Indicates time validity. These represent the weighting factors for time validity, user permission level, and user history usage credibility score, respectively. When the identification validity index is greater than the preset threshold, a wake-up signal is generated to trigger the multimodal biometric identification module to enter the data collection standby state, and at the same time the autofocus program of the high-definition camera and the light source preheating program of the infrared iris collector are started. The priority of subsequent biometric verification modes is determined based on the user's permission level, and the verification mode selection index is calculated using the following formula: ; in, Indicates the verification mode selection index. These represent the high-privilege level threshold and the low-privilege level threshold, respectively. When the verification mode selection index is 1, it indicates that the user is a high-privilege user and the iris recognition priority mode is used. When the verification mode selection index is 2, it indicates that the user is a user with ordinary permissions and the face-first mode is used. When the verification mode selection index is 3, it indicates that the user is a low-privilege user and adopts a dual-modal parallel mode of face and iris recognition. Simultaneously, the timestamp, device identifier, and geographic location information of this identity-triggered event are recorded to generate an access log entry.
[0019] S2. Biometrics Collection and Assessment: A sequence of facial images of the user is captured using a high-definition camera at a preset frame rate; the sequence of facial images consists of continuously acquired images. Frame image, where, ; Multi-dimensional liveness detection processing is performed on facial image sequences, including: detecting the user's natural blinking behavior during the acquisition window, counting the number of blinks and comparing it with the acquisition window duration to obtain the blinking frequency; The trajectory of facial micro-expression changes was analyzed using optical flow to extract the amplitude characteristics of expression changes. The user's facial 3D point cloud data is acquired using structured light projection imaging to reconstruct a facial 3D depth map. The facial depth consistency index is calculated based on the curved structural features of the real human face. The specific calculation formula is as follows: ; in, This indicates the facial depth consistency index. These represent the height and width pixel values of the 3D depth map of the face, respectively. Represents the first dimension in the 3D depth map of the face. The depth value corresponding to each pixel. This represents the average depth of the entire 3D depth map of the face; Through formula The facial liveness index is calculated, where, Indicates blink frequency, This represents the baseline value for normal blinking frequency. This represents the normalized value of blink frequency. Represents the normalization function. Indicates the characteristics of the range of facial expression changes. This indicates the score for the realism of skin texture. These represent the influencing weights of the normalized blink frequency, facial expression variation amplitude, facial depth consistency index, and skin texture realism score, respectively. ; When the face liveness index is greater than the preset liveness threshold, the frame corresponding to the maximum value of the face liveness index is selected from the facial image sequence as the frame with the best quality, and facial feature points are extracted based on a deep convolutional neural network to generate a multidimensional face feature vector. Simultaneously, the image quality of the best frame is evaluated. The sharpness score of the best frame image is obtained based on the Laplacian variance method, the pose frontality score is based on the symmetry of facial feature points, and the illumination uniformity score is based on the gray-level histogram distribution of the facial region. The face acquisition quality index is obtained by combining the weighted fusion formula. Iris image sequences were captured under near-infrared light illumination using an infrared iris scanner; pupil localization and iris boundary segmentation were performed on the iris images, and the inner and outer boundary parameters of the iris were obtained using the following formula: ; in, Optimal coordinates representing the center of the pupil and optimal radius , This represents the maximize operator. Indicated by radius The Gaussian function of the variable, Indicates the radius The partial derivatives, This represents the coordinates and radius of the input candidate circle. This represents the radius of the current candidate circle. This represents the integral term of the circumference of the circle. This represents the line integral along the circumference of a circle. Indicates the iris image in coordinates grayscale value at that location Indicates radius as The circumference of the circle, The arc length infinitesimal element representing the line integral; Similarly, through the formula Obtain the center coordinates of the outer boundary of the iris and the optimal radius of the outer boundary of the iris ; The circular iris region is segmented from the iris image based on the inner and outer boundary parameters of the iris. The segmented circular iris region is geometrically normalized based on the Daugman rubber sheet model, unfolded into a rectangular texture image of a fixed size, and the iris texture features are extracted through a two-dimensional Gabor filter to generate the iris feature code. While extracting iris texture features, the quality of the iris image frames used to generate iris feature codes is evaluated using a formula. The iris acquisition quality index is calculated, where, The focus score of the iris image is obtained by performing a sharpness test on the iris image. It represents the percentage of occlusion, which is obtained based on the ratio of the pixel area of the pixel region covered by the occluder to the total area of the iris region. The occluder includes the eyelids and eyelashes. This represents a score indicating the reasonableness of the pupil diameter relative to a preset baseline value. It is calculated by real-time detection of the pupil diameter in the current iris image, comparing the detected pupil diameter to a preset baseline diameter range, and determining the reasonableness score based on the proportion of pupil diameters falling within the preset range. , These represent the weighting factors affecting focus score, occlusion percentage, and reasonableness score, respectively.
[0020] S3, Identity Authentication Decision: Based on the S2 face capture quality index and iris capture quality index, the fusion weight is dynamically calculated using the following formula: , ; in, This indicates the quality sensitivity adjustment parameter. This indicates the quality index of face capture. Through formula Obtain the fused feature vector ,in, Indicates feature splicing, This represents the principal component dimensionality reduction function. Represents a facial feature vector. Represents the characteristic normalization function, Indicates iris feature code, These represent the dynamic fusion weights of the facial feature vector and the iris feature code, respectively. Retrieve pre-stored biometric templates from a secure, encrypted database based on the user's unique identifier (UID); Through formula The facial feature similarity is calculated, where, This represents the facial feature vector corresponding to the biometric template. Through formula The Hamming distance of the iris feature is calculated, where, This represents the total number of bits in the iris signature. This represents the first iris feature to be identified. Bit, This represents the first iris feature code corresponding to the biometric template. Bit, Represents the XOR operation; A comprehensive matching score is calculated based on facial feature similarity, Hamming distance of iris features, and face-iris consistency verification score, combined with a weighted fusion formula; wherein, the face-iris consistency verification score is obtained by weighted fusion of facial feature similarity and iris feature Hamming distance. A comprehensive security assessment index is obtained by combining the face liveness index, iris acquisition quality index, user history credibility score, and environmental security score with a weighted formula. The environmental security score is obtained by combining scene lighting, scene background, device environment, and abnormal signals. Set the primary authentication threshold and security threshold; When the overall matching score is greater than the main authentication threshold and the overall security assessment index is greater than the security threshold, the authentication is deemed successful, an authorization token is generated, and an authentication success log is recorded. When the overall matching score is between the security threshold and the main authentication threshold, a supplementary verification procedure is triggered, prompting the user to perform the specified cooperation action and then re-collect the data. The number of supplementary verifications shall not exceed 3. If the overall matching score is less than the security threshold, or if the number of supplementary verification attempts exceeds the limit and the authentication pass condition is still not met, the authentication will be rejected, the authentication process will be terminated, the failure log, failure reason code and collected image sample will be recorded, and the duration of subsequent verification requests for the current user identification information will be locked.
[0021] S4. Data Loading and Device Execution: After authentication, the personalized training file is retrieved from the cloud database or local encrypted cache based on the authorization token and the user's unique identification code. The personalized training file includes historical training datasets, visual function evaluation parameter sets, personalized training plans, and physiological safety restriction parameter sets. Extract the training records of the most recent K training sessions from the historical training dataset, including training duration sequence, training intensity sequence, training completion sequence, and training interval sequence, and obtain the user training state index through the exponential weighted moving average algorithm; Extract the user's current visual state parameters from the visual function assessment parameter set, including accommodation amplitude, convergence near point, stereoscopic acuity, and visual fatigue baseline; Based on the user training state index and visual function evaluation parameters, using the formula The personalized adjustment factor is calculated, where, This represents the user's training state index. Indicates the percentage of training progress. Indicates the rate of improvement in visual function. These represent the weighting factors for the user's training status index, training progress percentage, and visual function improvement rate, respectively. The recommended training modules for the day are retrieved from the personalized training, and the basic parameters are adaptively adjusted based on the personalized adjustment factor. The adjusted configuration parameters are assembled into a set of device control instructions and sent to the actuator of the vision training device through an encrypted communication channel, which automatically completes the switching of training modes, configuration of visual target parameters, and personalized presentation of the training interface. After the equipment parameters are configured, safety monitoring thresholds are loaded from the physiological safety restriction parameter set, including the maximum duration of a single training session, the cumulative daily training duration limit, the eye fatigue warning threshold, and special medical restriction indicators. The eye-tracking module integrated into the vision training device collects real-time data on the user's eye status, including real-time blink frequency, pupil diameter, fixation point coordinate sequence, and saccade speed. Through formula The gaze stability index is calculated, where, This represents the standard deviation of the gaze point coordinate sequence in the horizontal and vertical directions. Represents the horizontal and vertical coordinate values in the gaze point coordinate sequence. The radius of the effective gaze area indicates the tolerance range for allowed gaze points; The real-time fatigue index is calculated based on real-time blink frequency, pupil diameter, saccade speed, and fixation stability index, combined with a weighted formula. When the real-time fatigue index exceeds 60% of the preset fatigue threshold, a Level 1 warning is issued, a rest prompt is displayed on the training interface, and the training intensity parameter is reduced to 80% of the current value. When the real-time fatigue index exceeds 80% of the preset fatigue threshold, it is judged as a level 2 warning, the current training task is suspended, the rest countdown interface is forcibly displayed, and the rest time must not be less than the preset minimum rest time. When the real-time fatigue index exceeds the preset fatigue threshold or the cumulative training time reaches the preset maximum training time, a level 3 warning is triggered, the current training session ends, and the device is locked until the next day or until the preset recovery interval is reached before training can be restarted. When a special medical restriction marker is detected to be valid, the enhanced monitoring mode is activated. Throughout the training process, eye status data is synchronously uploaded to the associated medical monitoring terminal at a frequency of no less than 1Hz, and a reminder notification is sent to the preset contact person when abnormal indicators are triggered.
[0022] S5. Data recording and encrypted transmission: After the training is completed, collect complete data for this training, including actual training duration, training parameter records for each stage, training completion rate, eye state change curve during the training process, and user subjective feedback scores. The training data is structured and encapsulated to generate a training data package, and metadata including training timestamp, device identifier, software version number and data integrity check code is attached. A tiered encryption strategy is implemented based on data sensitivity levels: the raw eye state data is encrypted using the AES-256-GCM algorithm, and the training statistics data is encrypted using the national cryptographic SM4 algorithm. Encrypted data packets are sent back to the cloud database via the TLS 1.3 secure communication protocol to perform two-way certificate verification and data integrity verification. After receiving and decrypting the training data, the cloud database updates the user's personalized training profile, including appending historical training records, updating training progress indicators, and refreshing visual function trend data. Based on accumulated training data, the machine learning model is used to analyze the changing trends of users' visual functions; based on the analysis results and the current training, a personalized training plan for the next stage is optimized. The optimized training plan will be synchronized and updated to the user's personalized training profile, and will be automatically loaded and displayed when the user logs in for the next time.
[0023] Biometric recognition user visual training equipment, including: Identity wake-up trigger device: Equipped with an NFC reader, it reads the encrypted information in the user identification medium, calculates the identification validity index, triggers the device to wake up, matches the biometric verification mode based on the user's permission level, and records the access log simultaneously. Biometric data acquisition equipment: Equipped with a high-definition camera and an infrared iris scanner, it completes the acquisition and liveness detection of face image sequences and iris image sequences, quality assessment, extracts face feature vectors and iris feature codes, and outputs face and iris acquisition quality indices to provide feature data for authentication decisions; Authentication decision processing equipment: It has a built-in secure encrypted database and computing unit, dynamically calculates feature fusion weights, generates fused feature vectors and compares them with pre-stored templates, calculates comprehensive matching scores and security assessment indices, determines the authentication result through set thresholds, and triggers pass, supplementary verification or rejection instructions. Training execution control equipment: After authentication, it retrieves the user's personalized training file, calculates the user's training status index and personalized adjustment factor, adaptively adjusts training parameters and issues control commands, is equipped with an eye-tracking module to monitor the eye status in real time, calculates the real-time fatigue index and triggers warnings in different levels, and performs training mode switching, pause or termination operations. Data encryption and transmission device: After training, the device collects all training data and encapsulates it in a structured manner. It implements differentiated encryption based on sensitivity and transmits the data back to the cloud database via the TLS 1.3 protocol to complete file updates and training plan optimization.
[0024] A biometric recognition user visual training system, including: Identity wake-up module: Equipped with an NFC near-field communication unit, it reads the encrypted information in the user identification medium and calculates the identification validity index. When the index reaches the target, it wakes up the multimodal biometric module and matches the iris priority, face priority or dual-modal parallel verification mode based on the permission level, and records access log information simultaneously. Feature acquisition and evaluation module: Integrates high-definition camera and infrared iris collector to complete the acquisition of face image and iris image sequence. After liveness detection and 3D depth analysis, it extracts face feature vector and iris feature code to generate face and iris acquisition quality index. Authentication decision module: It has a built-in secure encrypted database and computing engine, dynamically calculates feature fusion weights and generates fusion feature vectors, compares them with pre-stored templates to obtain a comprehensive matching score, and combines the comprehensive security assessment index to determine the authentication result through a threshold, triggering pass, supplementary verification or rejection instructions; Training execution module: After authentication, the module retrieves the user's personalized training file, adaptively adjusts the training parameters based on the user's training status index and personalized adjustment factors, and sends them to the execution end. The module monitors the eye status in real time through the eye tracking module and performs graded early warning and control based on the real-time fatigue index.
[0025] Data encryption and backhaul module: After training, the full training data is collected and structured, and differential data encryption is performed. The data is then backhauled to the cloud database via the TLS1.3 protocol. After the archive is updated, the machine learning model analyzes the visual function trends and optimizes the subsequent training plan.
[0026] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.
Claims
1. A biometric identification user authentication method, characterized in that, include: S1. Identity Acquisition and Wake-up Trigger: The user identification information in the user identification medium is acquired through the NFC reader, and multimodal biometric identification is triggered based on the identification validity verification result; S2. Biometrics Acquisition and Evaluation: The multimodal biometrics module sequentially collects the user's facial and iris feature data, performs liveness detection and quality evaluation on the collected data, and generates a feature acquisition quality index. S3. Identity Authentication Decision: Based on a dynamic weight fusion strategy, the facial feature vector and iris feature vector are fused and matched with the pre-stored user biometric template for verification. The authentication decision result is generated by combining multi-dimensional security indicators. S4. Data Loading and Device Execution: After authentication, the personalized training file bound to the user's identification information is retrieved from the database, and the visual training device parameters are automatically configured based on the user's historical training status and visual function evaluation parameters. S5. Data Recording and Encrypted Backhaul: After training, the training data is hierarchically encrypted and associated with user identification information, then backhauled to the database to update the individual training profile, and subsequent training plans are intelligently optimized based on the accumulated data.
2. The biometric identification user authentication method according to claim 1, characterized in that, The specific operation steps of S1 are as follows: The system establishes near-field communication with the user identification medium via an NFC reader, reads the encrypted user identification information stored in the medium, and obtains an identification validity index by combining time validity, user permission level, and user historical usage credibility score. When the identification validity index is greater than a preset threshold, the system wakes up the multimodal biometric module and starts the high-definition camera autofocus and infrared iris collector light source preheating program. Simultaneously, based on the user permission level matching verification mode, including iris priority mode, face priority mode and face and iris dual-modal parallel mode, the timestamp, device identifier and geographical location information of this identity trigger event are recorded at the same time to generate access log entries.
3. The biometric identification user authentication method according to claim 1, characterized in that, The specific operation steps of S2 include: The system captures user facial image sequences using a high-definition camera, performs multi-dimensional liveness detection on the image sequences, detects natural blinking behavior and calculates blinking frequency, analyzes the amplitude of facial micro-expression changes using optical flow, reconstructs a three-dimensional facial depth map based on structured light projection imaging and calculates the depth consistency index, and obtains the face liveness index by weighted fusion of skin texture authenticity score. If the face liveness index exceeds the preset threshold, the frame corresponding to the maximum face liveness index is selected as the optimal frame. Facial feature points are extracted through a deep convolutional neural network and a multidimensional face feature vector is generated. At the same time, the optimal frame is evaluated to obtain the face acquisition quality index. Simultaneously, an infrared iris scanner captures iris image sequences under near-infrared light illumination, and the annular iris region is segmented by pupil localization and iris boundary separation. Based on the Daugman rubber sheet model, the circular iris region is geometrically normalized and unfolded into a fixed-size rectangular texture image. Then, the iris texture features are extracted through a two-dimensional Gabor filter to generate iris feature codes. At the same time, the quality of the image frames with generated feature codes is evaluated to obtain the iris acquisition quality index.
4. The biometric identification user authentication method according to claim 1, characterized in that, The specific operation steps of S3 include: The fusion weights are dynamically calculated based on the face acquisition quality index and the iris acquisition quality index, and the fusion feature vector is generated by concatenating the face feature vector and the iris feature code. Retrieve pre-stored biometric templates from a secure encrypted database based on the user's unique identifier (UID), calculate the facial feature similarity and the Hamming distance of the iris feature, and then weight the facial and iris consistency verification scores to obtain a comprehensive matching score. Simultaneously, a comprehensive security assessment index is obtained by combining the face liveness index, iris capture quality index, user historical usage credibility score, and environmental security score. A primary authentication threshold and a security threshold are preset. If the comprehensive matching score is greater than the primary authentication threshold and the comprehensive security assessment index is greater than the security threshold, the authentication is successful, an authorization token is generated, and the authentication success log is recorded. If the comprehensive matching score is between the security threshold and the primary authentication threshold, no more than 3 supplementary verifications are triggered. If the comprehensive matching score is less than the security threshold or the supplementary verification limit is exceeded, the authentication is rejected, the process is terminated, and the failure information is recorded.
5. The biometric identification user authentication method according to claim 1, characterized in that, The specific operation steps of S4 are as follows: After authentication, based on the authorization token and the user's unique identification code, the personalized training file is retrieved from the cloud database or local encrypted cache. Multiple training records are extracted to calculate the user's training status index, and visual status parameters are extracted. Combined with the training progress percentage and visual function improvement rate, a personalized adjustment factor is obtained. After adaptively adjusting the basic parameters of the recommended training module for the day, the configuration parameters are encapsulated into a control instruction set and sent to the actuator to complete operations such as training mode switching. Simultaneously, various thresholds and markers from the physiological safety restriction parameter set are loaded. The eye-tracking module collects user eye status data and calculates the gaze stability index and real-time fatigue index. A three-level fatigue warning mechanism is implemented. If a special medical restriction marker is detected to be valid, the enhanced monitoring mode is immediately activated, the eye status data is uploaded to the medical monitoring terminal, and a reminder notification is sent to the preset contact person when an abnormality occurs.
6. The biometric identification user authentication method according to claim 1, characterized in that, The specific operation steps of S5 are as follows: After the visual training session ends, complete data of the training is collected, including actual training duration, training parameter records at each stage, training completion rate, eye state change curve during training, and user subjective feedback scores. The training data is then structured and encapsulated to generate a training data package, and training timestamps, device identifiers, software version numbers, and data integrity check codes are added to the data package as metadata. A tiered encryption strategy is implemented for training data based on data sensitivity levels. The original eye state data is encrypted using the AES256GCM algorithm, while the training statistics data is encrypted using the national cryptographic SM4 algorithm. The encrypted data packets are transmitted back to the cloud database via the TLS1.3 secure communication protocol. At the same time, two-way certificate verification and data integrity verification are performed. After decryption, the cloud database updates the user's personalized training profile and analyzes the trend of visual function changes based on the accumulated training data through a machine learning model. The personalized training plan for the next stage is optimized and synchronized to the user profile for automatic loading during the next authentication login.
7. The biometric recognition user visual training device according to any one of claims 1-6, comprising: Identity wake-up triggering device: Equipped with an NFC reader, it reads the encrypted information of the user's identification medium and calculates the identification validity index. After triggering wake-up, it matches the biometric verification mode based on user permissions and records the access log simultaneously. Biometric data acquisition equipment: Equipped with a high-definition camera and an infrared iris scanner, it completes face and iris image acquisition and liveness detection, quality assessment, extracts feature vectors and feature codes, and outputs acquisition quality index; Authentication decision processing equipment: It has a built-in secure encrypted database and computing unit, calculates feature fusion weights and generates fusion feature vectors, compares them with pre-stored templates, and determines the authentication result based on a set threshold, triggering decision instructions; Training execution control equipment: After certification, it retrieves personalized training files, adjusts training parameters and issues instructions, monitors eye status through eye tracking, triggers fatigue warnings at different levels and executes training control; Data encryption and transmission device: Collects and encapsulates all training data, performs differentiated encryption, and transmits it back to the cloud via TLS 1.3 protocol to complete file updates and training plan optimization.
8. The biometric recognition user visual training system according to any one of claims 1-6, comprising: Identity wake-up module: Equipped with an NFC near-field communication unit, it reads the encrypted information of the user's identification medium and calculates the identification validity index. Once the index is met, it wakes up the multimodal biometric identification module and records the access log simultaneously based on the permission matching verification mode. Feature acquisition and evaluation module: Integrates high-definition camera and infrared iris collector to acquire face and iris image sequences, extract feature vectors and feature codes through detection and analysis, and generate acquisition quality index; Authentication decision module: It has a built-in secure encrypted database and computing engine. It generates a fused feature vector based on feature fusion weights, compares it with a pre-stored template, and combines the comprehensive security assessment index and a set threshold to determine the authentication result. Training execution module: After authentication, the module retrieves the personalized training file, adaptively adjusts the training parameters and sends them out for execution, monitors the eye status through eye tracking, and provides graded early warning and control based on the real-time fatigue index; Data encryption and backhaul module: Collects and encapsulates all training data and encrypts it differently. It then transmits the data back to the cloud via the TLS 1.3 protocol. After updating the archive, the machine learning model analyzes visual trends and optimizes the training plan.