Intelligent management and control system and method for AI glasses

Abstract

The invention discloses an intelligent management and control system and method for AI glasses, and the system comprises an environment sensing layer which integrates a millimeter-wave radar, an infrared sensor and a UWB positioning module, and senses the information of the environment where the AI glasses are located; the biological characteristic layer is used for authenticating the identity of the user by adopting an iris recognition and gait analysis composite authentication mode; the strategy engine layer comprises a dynamic permission matrix, a compliance rule base and a risk prediction model, and formulates and executes a corresponding control strategy according to environment perception information and a biological characteristic authentication result; the data protection layer is composed of a trusted execution environment and a federated learning framework and is used for guaranteeing the data security in the AI glasses system; and the man-machine interaction layer is provided with a tactile feedback and bone conduction alarm device, interacts with a user and provides alarm information. Through cooperative work of the layers, comprehensive, intelligent and safe management and control of the AI glasses are realized, so that the AI glasses can stably operate in various complex environments, and more reliable services are provided for users.

Description

Technical Field

[0001] The present invention belongs to the technical field of augmented reality device control, and relates to an intelligent management and control system and method for AI glasses. Background Art

[0002] With the rapid development of artificial intelligence technology, AI glasses, as intelligent wearable devices integrating multiple technologies such as computer vision, speech recognition, and augmented reality, are gradually entering people's daily lives. AI glasses can not only realize functions such as real-time information acquisition, voice interaction, photographing and video recording, but also show great application potential in fields such as industrial inspection, medical assistance, and education and training. However, the existing AI glasses systems have exposed many defects in actual applications. In some places with extremely high requirements for privacy protection, such as the intensive care unit of a hospital, the core R & D area of an enterprise, and the confidential meeting room of a government department, photographing and video recording behaviors are usually explicitly prohibited. Currently, the existing AI glasses are unable to automatically identify these prohibited shooting environments, and users can still casually activate the video recording function in these areas, which will undoubtedly pose a serious threat to the privacy and confidential information of others.

[0003] Most current AI glasses systems adopt a single user identity authentication method, which can only verify the user's identity when logging in and lack the ability to dynamically adjust permissions. In actual usage scenarios, different application scenarios and operations require different levels of permissions. For example, in an enterprise environment, ordinary employees may only have the permission to view basic information and use basic functions, while managers need higher permissions for data management and system settings. However, the existing AI glasses systems cannot dynamically adjust permissions according to the user's location and actual needs. Once the user passes the identity authentication, they can access all functions, increasing the risk of information leakage and abuse. Data security is also one of the important issues faced by AI glasses systems. The existing AI glasses systems do not establish an end-to-end encryption channel, and during data transmission, sensitive information such as biometric data is easily stolen and tampered with. Biometric data, such as fingerprints and facial features, is unique and unchangeable. Once leaked, it will bring serious security risks to users.

[0004] In summary, the existing AI glasses systems have many defects in aspects such as privacy protection, permission management, and data security, which restrict the further development and application of AI glasses technology. Therefore, it is necessary to develop an AI glasses system that can effectively solve the above problems to improve its security, reliability, and user experience. Summary of the Invention

[0005] The purpose of the present invention is to solve the problem that the existing AI glasses systems have many defects in aspects such as privacy protection, permission management, and data security, and to provide an intelligent management and control system and method for AI glasses.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] An intelligent management and control system for AI glasses includes an environmental perception layer, a biometric layer, a strategy engine layer, a data protection layer, and a human-computer interaction layer. The specific structures of each layer are as follows:

[0008] Environmental perception layer: Integrates millimeter-wave radar, infrared sensor, and UWB positioning module to perceive the environment in which the AI ​​glasses are located;

[0009] Biometric layer: uses iris recognition and gait analysis to authenticate user identities;

[0010] Policy engine layer: includes dynamic permission matrix, compliance rule base, and risk prediction model, which are used to formulate and execute corresponding control strategies based on environmental perception information and biometric authentication results;

[0011] Data protection layer: It consists of a trusted execution environment and a federated learning framework to ensure data security in the AI ​​glasses system;

[0012] Human-computer interaction layer: equipped with tactile feedback and bone conduction alarm devices to interact with users and provide alarm information.

[0013] The environmental perception layer is specifically:

[0014] The millimeter wave radar uses electromagnetic waves in the millimeter wave frequency band to detect targets, specifically including: distance detection, by transmitting millimeter wave signals and receiving their echoes, accurately measuring the distance between the AI ​​glasses and surrounding objects based on the time difference of the signal round trip; angle detection, with angle resolution capability, determining the azimuth angle of surrounding objects relative to the AI ​​glasses through signal phase difference analysis of multiple receiving antennas, including horizontal and vertical angles, to provide comprehensive environmental information for the system to plan the user's action path;

[0015] The infrared sensor senses the distribution of heat sources in the environment based on the infrared radiation emitted by the object, specifically including: sensing the surrounding environment by detecting the infrared radiation emitted by the object; identifying and tracking the location of people by detecting the infrared signals emitted by the human body; detecting abnormal heat sources in the environment and issuing an alarm in time when an abnormality is found;

[0016] The UWB positioning module uses ultra-wideband pulse signals for high-precision positioning to provide accurate location information for AI glasses, including: achieving centimeter-level high-precision positioning; being able to update the location information of AI glasses in real time to ensure that the system always knows the user's accurate location; by pre-setting the boundaries of prohibited areas or special operations areas, real-time monitoring of whether AI glasses enter these areas, and issuing an alarm and taking corresponding control measures when entering prohibited areas.

[0017] The biometric layer specifically includes:

[0018] An iris recognition module, including a high-definition iris camera built into the AI glasses frame. The high-definition iris camera uses a high-precision optical lens and an image sensor to clearly capture the fine texture of the user's iris under different lighting conditions; and an equipped infrared illumination system, which includes infrared LED lights and is used to provide uniform infrared illumination in low-light or night environments to ensure the clarity and contrast of the iris image; an iris feature extraction algorithm is used to extract unique feature points from the captured iris image using advanced image processing techniques; and an iris matching algorithm is used to compare the extracted iris feature template with the template pre-stored in the database to determine whether the identity matches by calculating the similarity;

[0019] A gait analysis module, including an accelerometer and a gyroscope built into the AI glasses, which are used to collect acceleration and angular velocity data of the user during walking in real time; a gait feature extraction algorithm is used to process and analyze the collected acceleration, angular velocity, and pressure data to extract key gait features such as stride length, walking frequency, and body swing pattern; and a gait modeling and recognition algorithm is used to establish a user gait model and compare it with the model in the database to achieve identity authentication;

[0020] Combining the two authentication methods of iris recognition and gait analysis to form a composite authentication strategy. During the authentication process, both the iris recognition and gait analysis modules are started simultaneously for identity authentication. Only when both authentication methods pass, the user's identity authentication is considered successful; otherwise, the system will deny access or trigger corresponding security alerts.

[0021] The policy engine layer specifically is:

[0022] The dynamic permission matrix is a two-dimensional table, where the rows represent different user roles or identities; the columns represent the functions or services of the AI glasses; each cell in the matrix defines a specific permission level, including allowed, prohibited, and restricted access;

[0023] The compliance rule library contains a set of compliance requirements; each rule in the rule library defines specific compliance conditions and corresponding handling measures;

[0024] The risk prediction model is a prediction model based on deep learning, which is used to analyze potential risks in environmental perception information and biometric authentication results;

[0025] The policy engine layer formulates specific control policies based on the analysis results of the dynamic permission matrix, compliance rule library, and risk prediction model, including restricting the use of certain functions, requiring users to perform additional authentication, triggering alarms, or notifying administrators.

[0026] The risk prediction model is specifically a neural network model based on deep learning, with an architecture that fuses a long short-term memory network and a convolutional neural network. The long short-term memory network is used to process sequential data, capture the changing trends of environmental perception information and biometric authentication results over time, and the convolutional neural network is used to extract spatial features from the data. The two are combined to mine complex patterns and potential risk features in the data;

[0027] Input data:

[0028] Environmental perception information: Includes real-time images and videos collected by the camera equipped with the AI glasses, covering the surrounding environmental scenes and personnel activities; the real-time geographical location information of the wearer obtained through GPS and indoor positioning technologies; and data collected by accelerometers, gyroscopes, and barometers for monitoring the wearer's movement state, posture changes, and environmental pressure;

[0029] Biometric authentication results: Include face recognition results, fingerprint recognition results, and iris recognition results;

[0030] Data preprocessing: Screen and clean the collected raw data to remove noise data, duplicate data, and error data; extract valuable features from the cleaned data, extract the shape, color, and texture features of the target object for visual data, extract the matching degree and feature vectors for biometric authentication results, and extract the movement trajectory and acceleration change rate for sensor data;

[0031] Model training: Collect historical data in normal usage scenarios and potential risk scenarios, and perform annotation to clarify the risk level corresponding to each data sample; use stochastic gradient descent to train the model and adjust the model parameters to minimize the error between the prediction result and the true label;

[0032] Risk level classification: Analyze the input data and output the risk level, including low risk, medium risk, and high risk;

[0033] Risk trend prediction: Predict whether the risk is tending to rise, fall, or remain stable by analyzing the dynamic changes of historical data and current data;

[0034] Output result: Output the risk level and risk trend prediction results to the policy engine layer in the form of structured data, including risk level labels, risk descriptions, and risk trend information.

[0035] The data protection layer is specifically:

[0036] The trusted execution environment is implemented based on ARM TrustZone and Intel SGX technologies, providing an isolated and secure environment for the execution of sensitive data and code; it includes a secure processor core, memory area, and encrypted key storage; it isolates sensitive operations from the regular operating system to prevent malware or unauthorized access; within the trusted execution environment, all data is encrypted during storage and transmission.

[0037] The federated learning framework is a decentralized machine learning method that allows multiple participants to jointly train a model while keeping the data local; it includes a central coordination server and multiple local training nodes, and exchanges and updates model parameters through an encrypted communication protocol.

[0038] The federated learning framework is a decentralized machine learning method that allows multiple participants to jointly train a model while keeping the data local. The specific process includes:

[0039] In the initialization stage, the central coordination server is responsible for coordinating and managing the entire federated learning process, initializing the model parameters, and formulating the encrypted communication protocol; multiple local training nodes join the federated learning process, and each node has local data, and this data does not leave the device.

[0040] In the model training stage, each local training node trains the model using the local data according to the initial model parameters sent by the central coordination server, and updates the model parameters to better adapt to the local data; after training is completed, the local training node uploads the updated model parameters to the central coordination server through the encrypted communication protocol, and the central coordination server receives and aggregates the model parameters of all nodes to update the global model; the central coordination server updates the global model according to the aggregated model parameters and sends the updated model parameters to all local training nodes, and the nodes continue the next round of local model training after receiving the new parameters.

[0041] In the iterative optimization stage: The federated learning process includes multiple rounds of iterative training. In each round of iteration, the local training nodes perform local training according to the latest model parameters sent by the central coordination server and upload the updated parameters, and the central coordination server aggregates and updates the global model until the model performance reaches the optimal level.

[0042] In the application and deployment stage: When the global model training is completed, the central coordination server deploys the model to each local training node, and the nodes use the deployed model for real-time prediction or decision-making.

[0043] The tactile feedback device is integrated into the frame or temple part of the AI glasses, and a micro vibration motor or piezoelectric material is used as the tactile feedback element; it is connected to the main control chip of the AI glasses through a control circuit, receives instructions from the policy engine layer, and generates corresponding tactile feedback; it can generate vibrations with different frequencies, intensities, and durations to distinguish different types of warning messages or interaction instructions.

[0044] The bone conduction warning device is located at the frame of the AI glasses, and uses bone conduction technology to directly transmit sound signals to the inner ear of the user; it includes a bone conduction transducer, an audio processing circuit, and a power module part, and is connected to the main control chip of the AI glasses.

[0045] An intelligent control method for AI glasses includes the following steps:

[0046] Environmental perception step: Using the millimeter-wave radar, infrared sensor, and UWB positioning module integrated on the AI glasses, perceive the environmental information of the AI glasses; transmit the perceived environmental information to the policy engine layer for the formulation of subsequent control policies;

[0047] Biometric authentication step: Adopt a composite authentication method of iris recognition and gait analysis to authenticate the user's identity; transmit the biometric authentication result to the policy engine layer as the basis for formulating control policies;

[0048] Control policy formulation and execution step: The policy engine layer receives the environmental perception information and biometric authentication result, combines the dynamic permission matrix, compliance rule library, and risk prediction model, and formulates corresponding control policies; according to the formulated control policies, dynamically adjusts and executes the functions and data access permissions of the AI glasses;

[0049] Data security guarantee step: The data protection layer uses a trusted execution environment to provide an isolated and secure environment for the execution of sensitive data and code, ensuring the confidentiality and integrity of data during storage and transmission; through the federated learning framework, allowing multiple participants to jointly train models while maintaining data localization, improving the accuracy and generalization ability of the models, and protecting user privacy and data security at the same time;

[0050] Human-computer interaction step: The human-computer interaction layer is equipped with a tactile feedback and bone conduction warning device to interact with the user; according to the control policies formulated by the policy engine layer or the system state, provide necessary warning information or operation confirmation prompts to the user through the tactile feedback and bone conduction warning device.

[0051] Compared with the prior art, the present invention has the following beneficial effects:

[0052] An intelligent control system for AI glasses proposed by the present invention realizes comprehensive, intelligent and secure management and control of AI glasses by integrating an environmental perception layer, a biometric layer, a policy engine layer, a data protection layer and a human-computer interaction layer. Its beneficial effects are mainly reflected in:

[0053] Comprehensive and accurate environmental perception ability: The environmental perception layer can real-time and accurately sense the environmental information where the AI glasses are located through integrating advanced technologies such as millimeter-wave radar, infrared sensors, UWB positioning modules, etc., including key parameters such as the positions, speeds, temperatures of surrounding objects, etc. It provides rich data support for subsequent policy formulation and execution, enabling the AI glasses to automatically adjust their working modes according to different environmental scenarios, improving the user experience and security.

[0054] Efficient and reliable user identity authentication: The biometric layer adopts a composite authentication method of iris recognition and gait analysis to double-verify the user identity, greatly improving the accuracy and reliability of identity authentication. It can not only effectively prevent unauthorized access, but also ensure that the personalized services of the AI glasses can be accurately provided to legitimate users while protecting user privacy.

[0055] Intelligent and flexible control strategy formulation and execution: The policy engine layer includes core components such as a dynamic permission matrix, a compliance rule library, a risk prediction model, etc. It can intelligently formulate and execute corresponding control strategies according to the environmental perception information and biometric authentication results. It can dynamically adjust the function permissions of the AI glasses to ensure their compliance and security in different scenarios, while reducing potential risks.

[0056] Powerful data security guarantee: The data protection layer consists of a trusted execution environment and a federated learning framework, providing comprehensive security guarantee for the data in the AI glasses system. The trusted execution environment ensures the execution of sensitive data and code in an isolated and secure environment through hardware-level security technology, preventing data leakage and malicious attacks. The federated learning framework allows multiple participants to jointly train models while keeping the data local, improving the accuracy and generalization ability of the models, while protecting user privacy and data security.

[0057] Intuitive and convenient human-computer interaction experience: The human-computer interaction layer is equipped with a tactile feedback and a bone conduction warning device, providing an intuitive, convenient and concealed interaction experience for users. The tactile feedback device can generate vibrations with different frequencies, intensities and durations to distinguish different types of warning information or interaction instructions; the bone conduction warning device can provide clear warning information in a noisy environment to ensure that users can accurately receive it.

[0058] Through the collaborative work among various layers, the present invention realizes the comprehensive, intelligent and secure management and control of AI glasses. The integrated design not only improves the overall performance of the system, but also enhances the stability of the system, enabling the AI glasses to operate stably in various complex environments and providing more reliable services for users. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as limiting the scope. For those of ordinary skill in the art, other related drawings can be obtained based on these drawings without creative efforts.

[0060] Figure 1 It is a flowchart of the intelligent management and control method for AI glasses of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0061] To make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are some but not all of the embodiments of the present invention. Usually, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0062] Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed present invention, but merely represents the selected embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts fall within the scope of protection of the present invention.

[0063] It should be noted that similar reference numerals and letters denote similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings.

[0064] An intelligent management and control system for AI glasses in the present invention includes an environmental perception layer, a biometric layer, a policy engine layer, a data protection layer, and a human-computer interaction layer. The specific structures of each layer are as follows:

[0065] Environmental perception layer: Integrates millimeter-wave radar, infrared sensors, and UWB positioning modules for perceiving the environmental information where the AI glasses are located.

[0066] Specifically, the environmental perception layer is:

[0067] The millimeter-wave radar detects targets using electromagnetic waves in the millimeter-wave band, specifically including: performing distance detection. By transmitting millimeter-wave signals and receiving their echoes, it accurately measures the distance between the AI glasses and surrounding objects based on the time difference of the signal round-trip, so as to detect the distance of the front obstacle in real time in a complex environment and timely remind the user to avoid collisions when the distance is too close; performing angle detection. It has angle resolution ability. By analyzing the signal phase difference of multiple receiving antennas, it determines the azimuth angle of surrounding objects relative to the AI glasses, including horizontal angle and vertical angle, providing comprehensive environmental information for the system to plan the user's movement path.

[0068] The infrared sensor senses the heat source distribution in the environment based on the infrared radiation emitted by objects, specifically including: in the night or low-light environment, being not restricted by the lighting conditions, and sensing the surrounding environment by detecting the infrared radiation emitted by objects; identifying and tracking the position of personnel by detecting the infrared signals emitted by the human body to meet the requirements for personnel monitoring in scenarios such as large event sites and factory workshops; detecting abnormal heat sources in the environment, such as fires and overheating of equipment, and issuing an alarm in time when an abnormality is found.

[0069] The UWB positioning module uses ultra-wideband pulse signals for high-precision positioning, providing accurate position information for the AI glasses, specifically including: achieving centimeter-level high-precision positioning, providing accurate navigation services for users in indoor environments to help users quickly find their destinations; being able to update the position information of the AI glasses in real time, ensuring that the system always knows the accurate position of the user, and improving the work efficiency and management level in scenarios with high requirements for position information such as logistics warehousing management and industrial inspection; by presetting the boundaries of areas where entry is prohibited or special operations are required, it monitors in real time whether the AI glasses enter these areas, and issues an alarm and takes corresponding control measures when entering the prohibited area.

[0070] Biometric layer: Adopts a composite authentication method of iris recognition and gait analysis to authenticate the user's identity.

[0071] The biometric layer specifically includes:

[0072] The iris recognition module includes a high-definition iris camera built into the AI glasses frame. The high-definition iris camera uses a high-precision optical lens and an image sensor to clearly capture the fine texture of the user's iris under different lighting conditions. And it is equipped with an infrared illumination system, which contains infrared LED lights to provide uniform infrared illumination in low-light or night environments, ensuring the clarity and contrast of the iris image. It adopts an iris feature extraction algorithm to extract unique feature points such as texture and spots from the captured iris image using advanced image processing techniques to form an iris feature template. And an iris matching algorithm is used to compare the extracted iris feature template with the template pre-stored in the database, and judge whether the identity matches by calculating the similarity. Iris recognition technology has extremely high uniqueness and stability. Each person's iris texture is unique and remains basically unchanged throughout a person's life. And encryption technology is used to protect the secure transmission and storage of iris data, preventing data leakage and illegal access.

[0073] The gait analysis module includes an accelerometer and a gyroscope built into the AI glasses, which are used to collect the acceleration and angular velocity data of the user when walking in real time. And an optional pressure sensor is used to capture the pressure distribution on the soles of the user's feet when walking, further enriching the gait feature data. It adopts a gait feature extraction algorithm to process and analyze the collected acceleration, angular velocity and pressure data, and extract key gait features such as stride length, walking frequency and body swing mode. And a gait modeling and recognition algorithm is used to establish a user gait model and compare it with the model in the database to achieve identity authentication. Gait analysis technology can perform identity authentication during the natural walking process of the user without the user's active cooperation, improving the convenience of use. And gait features are affected by various factors such as personal physiological structure, muscle strength and walking habits, making it difficult to be imitated or disguised, enhancing the security of authentication. Combining the two authentication methods of iris recognition and gait analysis to form a composite authentication strategy. During the authentication process, the iris recognition and gait analysis modules are started simultaneously to perform identity authentication respectively. Only when both authentication methods pass, the user's identity authentication is considered successful, otherwise the system will reject access or trigger corresponding security alarms. According to the user's usage scenario and security requirements, the system can dynamically adjust the parameters and thresholds of the composite authentication strategy. For example, it can increase the strictness of authentication in high-security places or sensitive operations, and appropriately reduce the complexity of authentication in ordinary scenarios to improve the user experience. Exception handling: When the iris recognition or gait analysis module has an abnormality or failure, the system can automatically switch to an alternative authentication method or trigger a manual review process to ensure the continuity and reliability of identity authentication.

[0074] The policy engine layer: includes a dynamic permission matrix, a compliance rule library, and a risk prediction model, which are used to formulate and execute corresponding control policies according to the environmental perception information and biometric authentication results.

[0075] The policy engine layer specifically includes:

[0076] The dynamic permission matrix is a two-dimensional table, where rows represent different user roles or identities, such as administrators, ordinary users, visitors, etc.; columns represent the functions or services of the AI glasses, such as taking pictures, recording videos, accessing sensitive data, etc.; each cell in the matrix defines a specific permission level, including allowed, prohibited, and restricted access; the dynamic permission matrix can adjust the user's permission level in real time according to environmental perception information, such as location, time, surrounding people, etc., and biometric authentication results; it supports fine-grained control of the functions of the AI glasses, and different permission levels can be set for individual functions or services to meet the security requirements in different scenarios.

[0077] The compliance rule library contains a set of compliance requirements, including various laws, regulations, industry standards, enterprise policies, etc.; each rule in the rule library defines specific compliance conditions and corresponding handling measures; the policy engine layer can automatically check whether the usage behavior of the AI glasses complies with the requirements in the compliance rule library, and once a violation is detected, the corresponding handling measures will be triggered immediately; as laws, regulations, and industry standards change, the compliance rule library can be updated regularly to ensure that the use of the AI glasses always complies with the latest compliance requirements.

[0078] The risk prediction model is a prediction model based on deep learning, used to analyze potential risks in environmental perception information and biometric authentication results;

[0079] Specifically, the risk prediction model adopts a neural network model based on deep learning, which is an architecture that fuses a long short-term memory network and a convolutional neural network. The long short-term memory network is used to process sequential data and capture the changing trends of environmental perception information and biometric authentication results over time, and the convolutional neural network is used to extract spatial features from the data. The two are combined to mine complex patterns and potential risk features in the data;

[0080] Input data:

[0081] Environmental perception information: includes real-time images and videos collected by the cameras equipped on the AI glasses, covering the surrounding environmental scenes and personnel activities; the real-time geographical location information of the wearer obtained through GPS and indoor positioning technologies; and data collected by accelerometers, gyroscopes, and barometers for monitoring the wearer's motion state, posture changes, and environmental pressure;

[0082] Biometric authentication results: include face recognition results, fingerprint recognition results, and iris recognition results;

[0083] Data preprocessing: Screen and clean the collected raw data to remove noisy data, duplicate data, and error data; Extract valuable features from the cleaned data. For visual data, extract the shape, color, and texture features of the target object. For biometric authentication results, extract the matching degree and feature vectors. For sensor data, extract the motion trajectory and acceleration change rate;

[0084] Model training: Collect historical data in normal usage scenarios and potential risk scenarios, and perform annotation to clarify the risk level corresponding to each data sample; Use stochastic gradient descent to train the model and adjust the model parameters to minimize the error between the prediction result and the true label;

[0085] Risk level classification: Analyze the input data and output the risk level, including low risk, medium risk, and high risk;

[0086] Risk trend prediction: Predict whether the risk tends to rise, fall, or remain stable by analyzing the dynamic changes in historical data and current data;

[0087] Output result: Output the risk level and risk trend prediction result to the policy engine layer in the form of structured data, including the risk level label, risk description, and risk trend information.

[0088] The policy engine layer formulates specific control strategies based on the analysis results of the dynamic permission matrix, compliance rule library, and risk prediction model, including restricting the use of certain functions, requiring users to perform additional identity verification, triggering an alarm, or notifying the administrator.

[0089] Data protection layer: Consists of a trusted execution environment and a federated learning framework, which is used to ensure the data security in the AI glasses system.

[0090] The data protection layer specifically is:

[0091] The trusted execution environment is implemented based on ARM TrustZone and Intel SGX technologies, providing an isolated and secure environment for the execution of sensitive data and code; including a secure processor core, memory area, and encrypted key storage; Isolate sensitive operations from the conventional operating system to prevent malware or unauthorized access; Within the trusted execution environment, all data is encrypted during storage and transmission;

[0092] The federated learning framework is a decentralized machine learning method that allows multiple participants to jointly train a model while keeping the data local; including a central coordination server and multiple local training nodes, and exchanging and updating model parameters through an encrypted communication protocol.

[0093] The federated learning framework is a decentralized machine learning method that allows multiple participants to jointly train a model while keeping the data local. The specific process includes:

[0094] In the initialization stage, the central coordination server is responsible for coordinating and managing the entire federated learning process, initializing the model parameters, and formulating an encrypted communication protocol; multiple local training nodes join the federated learning process, each node has local data, and this data does not leave the device;

[0095] In the model training stage, each local training node uses the local data to train the model according to the initial model parameters issued by the central coordination server, and updates the model parameters to better adapt to the local data; after the training is completed, the local training node uploads the updated model parameters to the central coordination server through the encrypted communication protocol, and the central coordination server receives and aggregates the model parameters of all nodes to update the global model; the central coordination server updates the global model according to the aggregated model parameters and issues the updated model parameters to all local training nodes, and the nodes continue the next round of local model training after receiving the new parameters;

[0096] Iterative optimization stage: The federated learning process includes multiple rounds of iterative training. In each round of iteration, the local training node performs local training according to the latest model parameters issued by the central coordination server and uploads the updated parameters, and the central coordination server aggregates and updates the global model until the model performance reaches the optimal level;

[0097] Application and deployment stage: When the global model training is completed, the central coordination server deploys the model to each local training node, and the nodes use the deployed model for real-time prediction or decision-making.

[0098] Human-computer interaction layer: Equipped with a tactile feedback and bone conduction warning device for interacting with the user and providing warning information.

[0099] The tactile feedback device is integrated in the frame or temple part of the AI glasses, and uses a micro vibration motor or piezoelectric material as the tactile feedback element; it is connected to the main control chip of the AI glasses through a control circuit, receives instructions from the policy engine layer and generates corresponding tactile feedback; it can generate vibrations of different frequencies, intensities and durations to distinguish different types of warning information or interaction instructions.

[0100] The bone conduction warning device is located at the frame of the AI glasses, and uses bone conduction technology to directly transmit the sound signal to the inner ear of the user; it includes a bone conduction transducer, an audio processing circuit and a power module part, and is connected to the main control chip of the AI glasses.

[0101] See Figure 1, which is a flowchart of an intelligent control method for an AI glasses in the present invention, includes the following steps:

[0102] Environmental perception step: Use the millimeter-wave radar, infrared sensor, and UWB positioning module integrated on the AI glasses to sense the environmental information where the AI glasses are located; transmit the sensed environmental information to the policy engine layer for formulating subsequent control policies;

[0103] Biometric authentication step: Adopt a composite authentication method of iris recognition and gait analysis to authenticate the user's identity; transmit the biometric authentication result to the policy engine layer as the basis for formulating control policies;

[0104] Control policy formulation and execution step: The policy engine layer receives the environmental perception information and biometric authentication result, combines the dynamic permission matrix, compliance rule library, and risk prediction model to formulate corresponding control policies; dynamically adjust and execute the functions and data access permissions of the AI glasses according to the formulated control policies;

[0105] Data security guarantee step: The data protection layer uses a trusted execution environment to provide an isolated and secure environment for the execution of sensitive data and code, ensuring the confidentiality and integrity of data during storage and transmission; through the federated learning framework, allowing multiple participants to jointly train the model while keeping the data local, improving the accuracy and generalization ability of the model, and at the same time protecting user privacy and data security;

[0106] Human-computer interaction step: The human-computer interaction layer is equipped with a tactile feedback and bone conduction warning device to interact with the user; provide the user with necessary warning information or operation confirmation prompts through the tactile feedback and bone conduction warning device according to the control policies formulated by the policy engine layer or the system status.

[0107] The present invention utilizes a millimeter-wave radar, an infrared sensor, and a UWB positioning module integrated on an AI glasses to obtain multi-dimensional environmental information, enabling a comprehensive and accurate perception of the surrounding environment. Combined with a risk prediction model, potential risks can be anticipated in advance; an iris recognition and gait analysis composite authentication method is adopted to greatly improve the accuracy and reliability of identity authentication, ensuring data security and privacy; the policy engine layer receives environmental perception information and biometric authentication results, formulates and executes intelligent control policies based on a dynamic permission matrix, a compliance rule library, and a risk prediction model, can dynamically adjust the functions of the AI glasses and data access permissions, while ensuring compliance; the data protection layer uses a trusted execution environment to provide an isolated and secure environment for the execution of sensitive data and code, ensuring data confidentiality and integrity, and jointly trains a model through a federated learning framework while keeping the data local, improving the accuracy and generalization ability of the model while protecting user privacy and data security; the human-computer interaction layer is equipped with a tactile feedback and bone conduction warning device, which can provide warning information or operation confirmation prompts to the user in a timely manner according to the control policy or system status, avoiding misoperations and not affecting the normal visual and auditory experiences, effectively improving the security, reliability, and user experience of the AI glasses, and having broad application prospects and market value.

[0108] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. An intelligent control system for AI glasses, characterized in that, It includes an environmental perception layer, a biometric layer, a policy engine layer, a data protection layer, and a human-computer interaction layer. The specific structures of each layer are as follows: Environmental perception layer: Integrates millimeter-wave radar, infrared sensors, and UWB positioning modules to sense the environmental information of the AI glasses; Biometric layer: Adopts a composite authentication method of iris recognition and gait analysis to authenticate the user's identity; Policy engine layer: Includes a dynamic permission matrix, a compliance rule library, and a risk prediction model to formulate and execute corresponding control policies based on the environmental perception information and biometric authentication results; Data protection layer: Consists of a trusted execution environment and a federated learning framework to ensure the data security in the AI glasses system; Human-computer interaction layer: Equipped with tactile feedback and bone conduction warning devices to interact with users and provide warning information.

2. The intelligent control system for AI glasses according to claim 1, characterized in that, Specifically, the environmental perception layer is as follows: The millimeter-wave radar uses electromagnetic waves in the millimeter-wave band to detect targets, specifically including: performing distance detection by transmitting millimeter-wave signals and receiving their echoes, and accurately measuring the distance between the AI glasses and surrounding objects based on the time difference of the signal round-trip; performing angle detection with angle resolution ability, and determining the azimuth angle of surrounding objects relative to the AI glasses through the signal phase difference analysis of multiple receiving antennas, including horizontal angle and vertical angle, to provide comprehensive environmental information for the system to plan the user's action path; The infrared sensor senses the heat source distribution in the environment based on the infrared radiation emitted by objects, specifically including: sensing the surrounding environment by detecting the infrared radiation emitted by objects; identifying and tracking the position of personnel by detecting the infrared signals emitted by the human body; detecting abnormal heat sources in the environment and sending out alarms in a timely manner when abnormalities are found; The UWB positioning module uses ultra-wideband pulse signals for high-precision positioning to provide accurate position information for the AI glasses, specifically including: achieving centimeter-level high-precision positioning; being able to update the position information of the AI glasses in real time to ensure that the system always knows the accurate position of the user; by presetting the boundaries of areas where entry is prohibited or special operations are required, monitoring in real time whether the AI glasses enter these areas, and sending out alarms and taking corresponding control measures when entering the prohibited area.

3. The intelligent control system for AI glasses according to claim 1, characterized in that, Specifically, the biometric layer includes: Iris recognition module, including a high-definition iris camera built into the AI glasses frame. The high-definition iris camera uses a high-precision optical lens and an image sensor to clearly capture the fine texture of the user's iris under different lighting conditions; and is equipped with an infrared illumination system, which includes infrared LED lights to provide uniform infrared illumination in low-light or nighttime environments to ensure the clarity and contrast of the iris image; adopts an iris feature extraction algorithm to extract unique feature points from the captured iris image using advanced image processing techniques; and an iris matching algorithm to compare the extracted iris feature template with the template pre-stored in the database and judge whether the identity matches by calculating the similarity; The gait analysis module includes an accelerometer and a gyroscope built into the AI glasses, which are used to collect acceleration and angular velocity data of the user in real time while walking; a gait feature extraction algorithm is adopted to process and analyze the collected acceleration, angular velocity and pressure data, and extract key gait features such as stride length, walking frequency and body swing pattern; and a gait modeling and recognition algorithm is used to establish a user gait model and compare it with the models in the database to achieve identity authentication. Combine the two authentication methods of iris recognition and gait analysis to form a composite authentication strategy. During the authentication process, start the iris recognition and gait analysis modules simultaneously to conduct identity authentication respectively. Only when both authentication methods pass, the user's identity authentication is considered successful. Otherwise, the system will reject access or trigger corresponding security alarms.

4. The intelligent control system for AI glasses according to claim 1, characterized in that, The policy engine layer is specifically as follows: The dynamic permission matrix is a two-dimensional table, where the rows represent different user roles or identities; the columns represent the functions or services of the AI glasses; each cell in the matrix defines a specific permission level, including allowed, prohibited and restricted access. The compliance rule library contains a set of compliance requirements; each rule in the rule library defines specific compliance conditions and corresponding handling measures. The risk prediction model is a prediction model based on deep learning, which is used to analyze potential risks in environmental perception information and biometric authentication results. The policy engine layer formulates specific control strategies according to the analysis results of the dynamic permission matrix, the compliance rule library and the risk prediction model, including restricting the use of certain functions, requiring users to perform additional identity verification, triggering alarms or notifying administrators.

5. The intelligent control system for AI glasses according to claim 4, characterized in that, The risk prediction model is specifically a neural network model based on deep learning, which is an architecture that fuses long short-term memory network and convolutional neural network. The long short-term memory network is used to process sequential data and capture the changing trends of environmental perception information and biometric authentication results over time. The convolutional neural network is used to extract spatial features from the data. The two are combined to mine complex patterns and potential risk features in the data. Input data: Environmental perception information: includes real-time images and videos collected by the camera equipped with the AI glasses, covering the surrounding environmental scenes and personnel activities; the real-time geographical location information of the wearer obtained through GPS and indoor positioning technologies. And the data collected by accelerometers, gyroscopes and barometer sensors, which are used to monitor the wearer's motion state, posture changes and environmental pressure. Biometric authentication results: include face recognition results, fingerprint recognition results and iris recognition results. Data preprocessing: screen and clean the collected raw data to remove noise data, duplicate data and error data; extract valuable features from the cleaned data. For visual data, extract the shape, color and texture features of the target object. For biometric authentication results, extract the matching degree and feature vectors. For sensor data, extract the motion trajectory and acceleration change rate. Model Training: Collect historical data in normal usage scenarios and potential risk scenarios, and perform annotation to clarify the risk level corresponding to each data sample; Use stochastic gradient descent to train the model and adjust the model parameters to minimize the error between the prediction result and the true label; Risk Level Classification: Analyze the input data and output the risk level, including low risk, medium risk, and high risk; Risk Trend Prediction: Predict whether the risk is tending to rise, fall, or remain stable by analyzing the dynamic changes in historical data and current data; Output Result: Output the risk level and risk trend prediction result to the policy engine layer in the form of structured data, including risk level labels, risk descriptions, and risk trend information.

6. The intelligent control system for AI glasses according to claim 1, characterized in that, The specific data protection layer is as follows: The trusted execution environment is implemented based on ARM TrustZone and Intel SGX technologies, providing an isolated and secure environment for the execution of sensitive data and code; It includes a secure processor core, memory area, and encrypted key storage; Isolate sensitive operations from the conventional operating system to prevent malware or unauthorized access; Within the trusted execution environment, all data is encrypted during storage and transmission. The federated learning framework is a decentralized machine learning method that allows multiple participants to jointly train a model while keeping the data local; It includes a central coordination server and multiple local training nodes, and exchanges and updates model parameters through an encrypted communication protocol.

7. The intelligent control system for AI glasses according to claim 6, wherein, The federated learning framework is a decentralized machine learning method that allows multiple participants to jointly train a model while keeping the data local. The specific process includes: Initialization Phase: The central coordination server is responsible for coordinating and managing the entire federated learning process, initializing the model parameters, and formulating the encrypted communication protocol; Multiple local training nodes join the federated learning process, and each node has local data, and this data does not leave the device. Model Training Phase: Each local training node uses the local data to train the model according to the initial model parameters issued by the central coordination server, and updates the model parameters to better adapt to the local data; After training is completed, the local training node uploads the updated model parameters to the central coordination server through the encrypted communication protocol. The central coordination server receives and aggregates the model parameters of all nodes, and updates the global model; The central coordination server updates the global model according to the aggregated model parameters and distributes the updated model parameters to all local training nodes. After receiving the new parameters, the nodes continue the next round of local model training. Iterative Optimization Phase: The federated learning process includes multiple rounds of iterative training. In each round of iteration, the local training node performs local training according to the latest model parameters issued by the central coordination server and uploads the updated parameters. The central coordination server aggregates and updates the global model until the model performance reaches the optimal level. Application and Deployment Phase: When the global model training is completed, the central coordination server deploys the model to each local training node, and the node uses the deployed model for real-time prediction or decision-making.

8. An intelligent control system for AI glasses according to claim 1, characterized in that, The tactile feedback device is integrated into the frame or temple part of the AI glasses, and a micro vibration motor or piezoelectric material is used as the tactile feedback element; it is connected to the main control chip of the AI glasses through a control circuit, receives instructions from the policy engine layer, and generates corresponding tactile feedback; it can generate vibrations with different frequencies, intensities, and durations to distinguish different types of warning messages or interaction instructions.

9. An intelligent control system for AI glasses according to claim 1, characterized in that, The bone conduction warning device is located at the frame of the AI glasses and uses bone conduction technology to directly transmit sound signals to the inner ear of the user; it includes a bone conduction transducer, an audio processing circuit, and a power module part, and is connected to the main control chip of the AI glasses.

10. An intelligent control method for AI glasses, characterized in that, It includes the following steps: Environmental perception step: Using the millimeter wave radar, infrared sensor, and UWB positioning module integrated on the AI glasses, perceive the environmental information where the AI glasses are located; transmit the perceived environmental information to the policy engine layer for formulating subsequent control policies; Biometric authentication step: Adopt a composite authentication method of iris recognition and gait analysis to authenticate the user's identity; transmit the biometric authentication result to the policy engine layer as the basis for formulating control policies; Control policy formulation and execution step: The policy engine layer receives the environmental perception information and biometric authentication result, combines the dynamic permission matrix, compliance rule library, and risk prediction model to formulate corresponding control policies; according to the formulated control policies, dynamically adjust and execute the functions and data access permissions of the AI glasses; Data security guarantee step: The data protection layer uses a trusted execution environment to provide an isolated and secure environment for the execution of sensitive data and code, ensuring the confidentiality and integrity of data during storage and transmission; Through the federated learning framework, on the premise of keeping the data local, allow multiple participating parties to jointly train the model, improve the accuracy and generalization ability of the model, and at the same time protect user privacy and data security; Human-computer interaction step: The human-computer interaction layer is equipped with a tactile feedback and bone conduction warning device to interact with the user; according to the control policies formulated by the policy engine layer or the system status, provide necessary warning information or operation confirmation prompts to the user through the tactile feedback and bone conduction warning device.

Images