A closed-loop vision rehabilitation glasses system based on ultrasonic MEMS and AI-driven frequency modulation

Abstract

This invention discloses a closed-loop vision rehabilitation glasses system based on ultrasonic MEMS and AI-driven frequency modulation, belonging to the field of vision rehabilitation equipment technology. The system includes a glasses body (100), and an ultrasonic MEMS biosensing module (200), a visual stimulation module (300), an AI closed-loop control module (400), and a wireless communication module (130) integrated into the glasses body. The ultrasonic MEMS biosensing module is used to collect the ciliary muscle activity state of the user in a non-contact real-time manner. Its sensor array (210) is installed at a distance of ≤2mm from the face, and the light-shielding baffle (140) is tilted at 30°. The AI ​​closed-loop control module includes a data acquisition unit (410), an individualized frequency feature library construction unit (420), a real-time control unit (430), a visual health state spatial modeling unit (440), and a state guidance unit (450). It automatically identifies the user's optimal intervention frequency through multi-frequency stimulation scanning (6 / 8 / 10 / 12 / 15Hz) and a weighted comprehensive scoring method (weights: contraction speed 0.5 / accommodation amplitude 0.3 / intraocular pressure fluctuation 0.2), and dynamically adjusts visual stimulation based on real-time changes in ciliary muscle activity, forming a closed-loop rehabilitation circuit. A wireless communication module (130) enables data synchronization with an external terminal. This invention deeply integrates ultrasonic MEMS sensing, AI frequency optimization, and closed-loop control, offering advantages such as precise sensing, intelligent control, and significant rehabilitation effects.

Description

Technical Field

[0001] This invention belongs to the field of vision rehabilitation equipment technology, specifically relating to a closed-loop vision rehabilitation glasses system based on the combination of ultrasonic MEMS biosensing technology and artificial intelligence-driven frequency modulation technology. Background Technology

[0002] With the widespread use of electronic display devices, vision problems, especially eye strain and decreased accommodative function caused by ciliary muscle spasm, are showing a trend of high incidence and younger age of onset. Existing vision rehabilitation equipment mainly suffers from the following technical deficiencies: 1. Low sensing accuracy, unable to monitor deep ocular physiological states in real time. Traditional vision training devices mostly use eye tracking or subjective feedback, which cannot directly measure the real-time state of accommodation structures such as the ciliary muscle and lens. Recent research attempted to integrate PMUT sensors into glasses for blink monitoring (Dual-Electrodes PMUTs on Glasses for Wearable Human Blink Monitoring, 2026), but its function is limited to detecting blinking movements and does not involve the collection of deep ocular physiological parameters or rehabilitation intervention.

[0003] 2. Lack of a closed-loop control mechanism based on physiological feedback. Most existing visual stimulation devices employ fixed frequencies or open-loop control. For example, frequency-based visual stimulation in the field of brain-computer interfaces is primarily used for command decoding, rather than dynamically adjusting according to the physiological state of the eyes. While the concept of frequency therapy has been applied (e.g., 40Hz light stimulation for cognitive health and phototherapy lamps for myopia control), these are all open-loop stimulations and cannot adaptively adjust based on the user's real-time physiological responses.

[0004] 3. Disconnect between intervention and entertainment functions. Most existing products are purely training or purely entertainment, lacking deep integration, resulting in poor user compliance.

[0005] 4. Current wearable devices have not yet effectively combined high-precision deep physiological sensing, AI frequency optimization, and closed-loop regulation. The development of microelectromechanical systems (MEMS) technology, especially ultrasonic MEMS sensors, has made it possible to solve the problem of non-invasive, real-time monitoring of deep ocular parameters. However, how to integrate them into portable glasses and build an intelligent closed-loop rehabilitation system based on physiological feedback remains a problem that urgently needs to be solved in this field. Summary of the Invention

[0006] (a) Purpose of the invention The purpose of this invention is to overcome the shortcomings of the prior art and provide a closed-loop vision rehabilitation glasses system based on ultrasonic MEMS and AI-driven frequency regulation. The system uses ultrasonic MEMS sensors to collect ciliary muscle activity in real time, combines AI algorithms to identify the optimal intervention frequency for each individual, and dynamically regulates visual stimulation based on real-time physiological feedback to achieve precise, personalized, and highly compliant closed-loop vision rehabilitation.

[0007] (II) Technical Solution To achieve the above-mentioned objectives, the present invention adopts the following technical solution: A closed-loop vision rehabilitation glasses system based on ultrasonic MEMS and AI-driven frequency modulation, characterized in that it includes: Eyeglasses body (100); An ultrasonic MEMS biosensing module (200) integrated into the main body of the glasses (100) is used to collect the user's ciliary muscle activity status in real time in a non-contact manner. A visual stimulation module (300) integrated into the main body of the glasses (100) is used to output visual stimulation to the user; The AI ​​closed-loop control module (400) integrated into the main body of the glasses (100) is electrically connected to the ultrasonic MEMS biosensing module (200) and the visual stimulation module (300), respectively. The wireless communication module (130) integrated into the main body of the glasses (100) is electrically connected to the AI ​​closed-loop control module (400) and is used to synchronize or exchange data with external terminal devices.

[0008] The AI ​​closed-loop control module (400) includes: The data acquisition unit (410) is electrically connected to the ultrasonic MEMS biosensing module (200) and is used to receive the raw signal and perform analog-to-digital conversion, preliminary filtering and buffering. The individualized frequency feature library construction unit (420) is used to automatically identify and store at least one intervention frequency that corresponds to the optimal response of the user's visual system through multi-frequency stimulation scanning and real-time ciliary muscle response analysis in the initialization mode. The real-time control unit (430) is used to generate control instructions and send them to the visual stimulation module (300) based on the changes in ciliary muscle activity state collected in real time and combined with the optimal intervention frequency in the individualized frequency feature library. Visual health state space modeling unit (440) is used to construct a user's personalized high-dimensional visual health state space based on historically collected multi-dimensional eye parameters; The state guidance unit (450) is used to calculate in real time the deviation between the current state position in the state space and the preset optimal region, and generate guidance and control instructions based on the deviation.

[0009] The visual stimulation module (300) dynamically adjusts the frequency and parameters of the visual stimulation according to the control instructions, forming a closed-loop rehabilitation circuit based on real-time feedback from the ciliary muscle.

[0010] Furthermore, the ultrasonic MEMS biosensing module (200) includes an ultrasonic MEMS sensor array (210), a signal conditioning unit (220), and a data transmission unit (230). The ultrasonic MEMS sensor array (210) is arranged on the inner side of the temple (110) and the edge of the frame (120) of the eyeglass body (100) to emit ultrasonic signals and receive reflected signals from the eye tissue, so as to resolve parameters such as ciliary muscle contraction speed, contraction amplitude, intraocular pressure fluctuation, and relaxation time.

[0011] Furthermore, the ultrasonic MEMS sensor array (210) employs a piezoelectric micromechanical ultrasonic sensor (PMUT), specifically a dual-electrode, dual-piezoelectric layer structure, with an operating frequency of 1-10MHz, an acquisition accuracy of ≤0.01mm, and a sampling frequency of 10-50Hz. After installation, the vertical distance between the PMUT sensor array (210) and the facial skin is controlled to ≤2mm to ensure effective penetration of the ultrasonic signal and reception of high-quality reflected signals. Compared to the traditional single-piezoelectric layer structure, the dual-electrode, dual-piezoelectric layer structure improves the center displacement by more than 400% and the linearity by 10 times, ensuring the accuracy and stability of ciliary muscle parameter acquisition.

[0012] Furthermore, the individualized frequency feature library construction unit (420) includes: A multi-frequency stimulation generator (421) is used to control the visual stimulation module (300) to output visual stimulation in a preset frequency sequence, the preset frequency sequence covering at least 5 different frequency bands of stimulation, specifically including 6Hz, 8Hz, 10Hz, 12Hz and 15Hz. The response feature extractor (422) is used to analyze the ciliary muscle response speed, amplitude and intraocular pressure fluctuation parameters collected by the ultrasonic MEMS biosensor module (200) under stimulation at various frequencies. The system can deduce the relaxation time of the ciliary muscle based on the intraocular pressure fluctuation cycle, which together serve as key indicators for assessing the ciliary muscle activity status. The optimal frequency identifier (423) is used to automatically identify the 1-3 frequencies that are most significant in response to the user based on the response characteristics and through a preset response intensity evaluation algorithm, and store them in the individualized frequency feature library (424).

[0013] Furthermore, the response intensity assessment algorithm employs a weighted comprehensive scoring method. After normalizing the ciliary muscle contraction speed, contraction amplitude, and intraocular pressure fluctuation / relaxation time parameters, the results are weighted and summed. The frequency with the highest comprehensive score is taken as the optimal intervention frequency. Specifically, contraction speed has a weight of 0.5, accommodation amplitude has a weight of 0.3, and intraocular pressure fluctuation has a weight of 0.2.

[0014] Furthermore, the visual stimulation module (300) includes a display unit (310) and an audio unit (320). The display unit (310) adopts a micro OLED flexible display screen or a MicroLED waveguide display module, which is arranged at the lens position of the main body of the glasses (100). The audio unit (320) adopts a bone conduction speaker or a micro speaker, which is arranged on the inside of the temple (110) to output an audio guidance signal synchronized with the visual stimulation.

[0015] As a preferred embodiment, the display unit (310) is a dimmable display unit, whose transmittance can be dynamically adjusted according to ambient light or the needs of the rehabilitation scenario. Specifically, the display unit (310) includes an electrochromic layer or a liquid crystal dimming layer, which is electrically connected to the AI ​​closed-loop control module (400) and adjusts the transmittance according to the control command.

[0016] As a preferred embodiment, the visual stimulation module (300) further includes a rehabilitation linkage unit (330) for binding vision rehabilitation training tasks with visual stimulation content. After the user completes a specific training action, the visual stimulation content is updated, achieving a deep integration of intervention and entertainment. The vision rehabilitation training tasks include one or more of the following: ciliary muscle relaxation training, eye rotation training, and near-far accommodation training.

[0017] As a preferred embodiment, the glasses body (100) includes a left-side light-blocking baffle and a right-side light-blocking baffle (140), extending from the front ends of the left and right temples (110) toward the sides of the face to block lateral ambient light. The light-blocking baffle (140) extends toward the sides of the face at a 30° angle relative to the temples (110) to optimize the lateral light blocking effect and ensure good contact between the sensor and the temple area. At least a portion of the ultrasonic MEMS sensor array (210) is disposed on the inner side of the left-side and right-side light-blocking baffles (140), contacting the temple area.

[0018] As a preferred embodiment, the left and right light-blocking baffles (140) are made of translucent material with a light transmittance of 30-60%, and are detachably connected to the temples.

[0019] Furthermore, it also includes a data storage module (500) electrically connected to the AI ​​closed-loop control module (400), used to store historical data of user ciliary muscle parameters, an individualized frequency feature library, and system operation logs. The data storage module (500) supports data export and cloud backup.

[0020] Furthermore, it also includes an interactive module (600) electrically connected to the AI ​​closed-loop control module (400). The interactive module (600) includes one or more of physical buttons, a touch screen, and a voice recognition unit, and is used for users to set system parameters, view rehabilitation data, and manually mark the best state period.

[0021] Furthermore, it also includes a power module (700) that provides power to each module. The power module (700) uses a rechargeable lithium battery with a capacity of 500-1000mAh and integrates a power management unit for monitoring battery power and controlling charging speed.

[0022] Furthermore, the main body of the glasses (100) is made of lightweight and adjustable material, with an overall weight of ≤50g; the temples (110) are foldable and the frame (120) is adjustable, suitable for users of different ages.

[0023] (III) Beneficial Effects Compared with the prior art, the present invention has the following beneficial effects: 1. First realization of closed-loop vision rehabilitation based on real-time feedback from the ciliary muscle: Unlike existing PMUT glasses which are only used for blink monitoring, this invention uses ultrasonic MEMS sensors to collect deep physiological parameters such as the contraction speed and amplitude of the ciliary muscle in real time, and dynamically adjusts visual stimulation based on the real-time changes of these parameters, forming a complete closed loop of "monitoring-analysis-regulation-feedback", realizing a fundamental leap from passive monitoring to active intervention.

[0024] 2. This invention is the first to combine AI frequency modulation with ocular physiological monitoring: Unlike frequency stimulation (used only for command decoding) and open-loop frequency intervention devices (such as 40Hz phototherapy) in the field of brain-computer interfaces, this invention identifies the optimal intervention frequency for each individual through multi-frequency scanning and dynamically optimizes it based on real-time ciliary muscle response. A functional interdependence exists between the AI ​​algorithm and the ultrasonic MEMS sensor, producing non-obvious technical effects—experiments show that after adopting individualized frequency modulation, the speed of ciliary muscle relaxation increases by 37%, and the training effect is significantly better than fixed-frequency training.

[0025] 3. Deep integration of interventional entertainment enhances adherence: By deeply binding training tasks with visual stimulation content through rehabilitation linkage units, users complete rehabilitation training in an entertaining experience, effectively solving the problems of boredom and difficulty in users persisting with traditional rehabilitation equipment, and improving adherence by more than 50%.

[0026] 4. Dimmable Display Enhances Scene Adaptability: As a preferred solution, the dimmable display unit can dynamically adjust its transmittance according to ambient light and the rehabilitation scene, ensuring usability in different indoor and outdoor scenarios while enhancing the immersive rehabilitation experience. Three dimming modes (transparent, semi-transparent, and immersive) are intelligently linked with the rehabilitation content to further optimize training effects.

[0027] 5. Integrated design of the light shield and sensor: The light shield not only blocks ambient light from the sides to enhance immersion, but also serves as an integrated carrier for the PMUT sensor. The 30° tilt design ensures optimal fit to the temple area, achieving high-quality signal acquisition. The detachable design allows the glasses to flexibly switch between immersive mode and daily mode, expanding application scenarios.

[0028] 6. Visual health status space guides intelligent repair: By constructing a personalized visual health status space, the rehabilitation goal is upgraded from "completing training tasks" to "guiding users back to the best physiological state", drawing a personalized rehabilitation roadmap for each user and achieving true intelligent repair.

[0029] 7. Compact structure and comfortable wear: All functional modules are highly integrated into a lightweight, adjustable glasses body, with the overall weight controlled to within 50g. The sensor's distance from the face is ≤2mm, ensuring signal quality while maintaining wearing comfort. Actual testing showed no pressure marks on the bridge of the nose and no ear pain after 4 hours of continuous wear.

[0030] 8. Wireless communication and ecological interconnection: By integrating a wireless communication module (130), the system can synchronize data with mobile APP and cloud server, support remote rehabilitation guidance, multi-person data comparison and AI model cloud update, and expand the functional boundaries of the system. Attached Figure Description

[0031] Figure 1 This is a block diagram of the overall structure of the system of the present invention; Figure 2 This is a schematic diagram of the structure and sensor layout of the ultrasonic MEMS biosensing module of the present invention, showing the design of the sensor being ≤2mm away from the face and the light-shielding baffle being tilted at 30°. Figure 3 This is a block diagram of the internal functional units of the AI ​​closed-loop control module of the present invention; Figure 4 This is a flowchart of the individualized frequency feature library construction unit of the present invention, showing multi-frequency stimulation sequences (6 / 8 / 10 / 12 / 15Hz) and a weighted comprehensive scoring method (weight: contraction velocity 0.5 / accommodation amplitude 0.3 / intraocular pressure fluctuation 0.2). Figure 5This is a schematic diagram of the internal structure of the visual stimulation module of the present invention and its interaction logic with the rehabilitation linkage unit; Figure 6 This is a schematic diagram of the main body of the glasses of the present invention, showing key design parameters such as a 30° tilt of the light-blocking baffle and a vertical distance of ≤2mm between the sensor and the face contact surface.

[0032] Explanation of markings in the diagram 100 glasses body; 110 temples; 120 frames; 130 wireless communication module; 140 light-blocking baffle; 200 ultrasonic MEMS biosensing modules; 210 Ultrasonic MEMS Sensor Array (PMUT); 220 signal conditioning unit; 230 data transmission units; 300 visual stimulation modules; 310 display unit; 320 audio units; 330 Rehabilitation Linkage Unit; 400 AI closed-loop control module; 410 Data Acquisition Unit; 420 Individualized Frequency Feature Library Construction Unit; 421 Multi-Frequency Stimulator; 422 Response Feature Extractor; 423 Optimal Frequency Recognition; 424 individualized frequency feature libraries; 430 real-time control unit; 440 Visual Health Status Spatial Modeling Units; 450 State Guidance Unit; 500 data storage modules; 600 interactive modules; 700 power supply module. Detailed Implementation

[0033] Example 1: Basic System This embodiment provides a closed-loop vision rehabilitation glasses system based on ultrasonic MEMS and AI-driven frequency modulation, suitable for myopia prevention and control and eye strain relief in adolescents. The system structure is as follows: Figure 1 and Figure 6As shown, the glasses include a main body (100), and an ultrasonic MEMS biosensing module (200), a visual stimulation module (300), an AI closed-loop control module (400), a wireless communication module (130), a data storage module (500), an interaction module (600), and a power supply module (700) integrated into the main body.

[0034] The main body of the glasses (100) adopts a composite structure of magnesium-lithium alloy frame and TR90 nylon temples, with an overall weight of 48g. The top bridge of the frame (120) is 2.2mm thick, the temples (110) are 7.8mm wide, and the length is adjustable from 135-145mm. The front ends of the left and right temples are respectively provided with translucent light-blocking baffles (140) extending to the side of the face, with a light transmittance of 40%. The light-blocking baffles (140) extend to the side of the face at a 30° angle relative to the temples (110) to optimize the side light blocking effect and ensure good fit between the sensor and the temple area.

[0035] The ultrasonic MEMS biosensing module (200) includes a PMUT sensor array (210), a signal conditioning unit (220), and a data transmission unit (230). The PMUT sensor array (210) adopts a dual-electrode dual-piezoelectric layer structure, operates at a frequency of 5MHz, has an acquisition accuracy of 0.005mm, and a sampling frequency of 30Hz. After installation, the vertical distance between the PMUT sensor array (210) and the facial skin is controlled to ≤2mm to ensure effective penetration of ultrasonic signals and reception of high-quality reflected signals. Eight PMUT sensors are respectively arranged on the inner side of the left and right light-shielding baffles (140) (two each, fitting the temple area) and the inner side of the upper beam of the eyeglass frame (four). The signal conditioning unit (220) uses a low-noise amplifier and a bandpass filter to perform noise reduction, filtering, and amplification processing on the original signal. The data transmission unit (230) is connected to the AI ​​closed-loop control module (400) via a wired connection.

[0036] The visual stimulation module (300) includes a display unit (310) and an audio unit (320). The display unit (310) uses a MicroLED waveguide display module with 85% light transmittance and 1080P resolution, and is located at the lens position. The audio unit (320) uses a bone conduction speaker and is located on the inside of the temple, used to output audio guidance signals synchronized with the visual stimulation.

[0037] The AI ​​closed-loop control module (400) uses an STM32U5 series microprocessor and integrates a data acquisition unit (410), an individualized frequency feature library construction unit (420), a real-time control unit (430), a visual health state space modeling unit (440), and a state guidance unit (450). The data acquisition unit (410) receives raw signals from the ultrasonic MEMS biosensor module (200), performs 12-bit analog-to-digital conversion, digital filtering, and ring buffering, and the preprocessed data is used by other units.

[0038] The wireless communication module (130) uses a Bluetooth 5.0 chip and is electrically connected to the AI ​​closed-loop control module (400) for data synchronization with a mobile APP. Users can view rehabilitation progress, receive training reports, synchronize individualized frequency feature libraries to the cloud, and receive remote guidance and suggestions from ophthalmologists through the mobile APP.

[0039] The data storage module (500) uses 16GB of non-volatile flash memory. The interaction module (600) includes three physical buttons (outer temples) and a 1.5-inch touchscreen (inner temples). The power module (700) uses a 600mAh rechargeable lithium battery that supports fast charging and can provide up to 8 hours of continuous use on a single charge.

[0040] Personalized frequency feature library construction: When a user uses the system for the first time, they enter the initialization mode through the interaction module (600). The multi-frequency stimulation generator (421) of the personalized frequency feature library construction unit (420) controls the visual stimulation module (300) to output visual stimuli sequentially at frequencies of 6Hz, 8Hz, 10Hz, 12Hz, and 15Hz (each frequency lasts for 30 seconds, with a 10-second interval). The response feature extractor (422) analyzes the ciliary muscle response data collected by the PMUT sensor array (210) under each frequency stimulation, extracts the ciliary muscle contraction velocity (mm / s), contraction amplitude (relative change), and intraocular pressure fluctuation parameters, and the system derives the ciliary muscle relaxation time (ms) based on the intraocular pressure fluctuation cycle, which together serve as response feature indicators. The optimal frequency identifier (423) uses a weighted comprehensive scoring method (ciliary muscle contraction speed weight 0.5, contraction amplitude weight 0.3, intraocular pressure fluctuation / relaxation time weight 0.2). After normalizing each indicator, the weighted sum is calculated to identify the two frequencies with the highest comprehensive scores as 10Hz and 12Hz, which are then stored in the individualized frequency feature library (424).

[0041] Real-time closed-loop control: In daily use, the real-time control unit (430) acquires processed ciliary muscle contraction speed data from the data acquisition unit (410) in real time. When the ciliary muscle contraction speed is detected to be higher than the threshold (30% higher than the individual baseline value) for more than 30 seconds, the system determines that the user is in a state of visual fatigue. The real-time control unit (430) generates a control command based on the optimal frequency (10Hz) in the individualized frequency feature library (424) and the current visual fatigue state, and sends it to the visual stimulation module (300). The visual stimulation module (300) outputs a guide light point that moves from near to far at a frequency of 10Hz (10 near-far switching per second), while the audio unit (320) outputs a synchronized guide sound to prompt the user to follow the training. During the training process, the PMUT sensor array (210) monitors the contraction-relaxation cycle of the ciliary muscle in real time to verify the correctness of the user's movements. The data is fed back to the real-time control unit (430) after signal conditioning and acquisition. When the ciliary muscle parameters are detected to return to normal, the system automatically reduces the training intensity or exits the training mode, forming a closed-loop feedback.

[0042] The data storage module (500) records all physiological data and regulatory parameters for this training session. After the training is completed, the wireless communication module (130) automatically synchronizes the data to the user's mobile APP and generates a visual rehabilitation report.

[0043] Example 2: An enhanced system integrating interventional entertainment functions This embodiment enhances the functionality of the visual stimulation module (300) based on Embodiment 1. The visual stimulation module (300) adds a rehabilitation linkage unit (330) to deeply bind the vision rehabilitation training task with the visual stimulation content. In this embodiment, the visual stimulation content is a "Star Trek" game. Binding logic: The game task requires the user to follow a guide light point to complete near-far adjustment training; the training frequency is 10Hz (determined by an individualized frequency feature library); completion is determined by the PMUT sensor verifying the ciliary muscle's contraction-relaxation cycle in real time, with one correct contraction-relaxation cycle counted as one instance; the reward mechanism is that after completing a set (10 times) of correct training, the game character receives energy replenishment and unlocks the next level; the difficulty progression is that as training progresses, the required number of training sessions gradually increases, and the game difficulty increases synchronously. Testing showed that after adopting the interventional entertainment binding method, the user's average daily training time increased from 12 minutes to 28 minutes, and the compliance rate reached 82% after 3 months, significantly better than traditional training equipment (usually below 30%).

[0044] Example 3: Enhanced System with Integrated Dimmable Functionality This embodiment enhances the display unit (310) based on embodiment 1. The display unit (310) adopts a multi-layer structure, consisting of the following layers from the outside to the inside: an outer protective glass layer (0.3mm thick), an electrochromic color-changing layer (transmittance adjustment range 10%-85%, response time <0.5 seconds), a MicroLED display layer (0.8mm thick), and an inner protective glass layer (0.2mm thick). The electrochromic color-changing layer is electrically connected to the AI ​​closed-loop control module (400) and can dynamically adjust the transmittance according to ambient light sensor data or rehabilitation scenario requirements. Three dimming modes are available: transparent mode (85% transmittance, suitable for outdoor walking and talking), semi-transparent mode (40-50% transmittance, suitable for indoor daily training), and immersive mode (10-20% transmittance, suitable for deep rehabilitation training). The rehabilitation linkage unit (330) can automatically switch dimming modes according to the type of training task: semi-transparent mode is used for near-far adjustment training; immersive mode is used for eye movement training; and transparent mode is used for relaxation training. Users can also quickly switch modes using physical buttons.

[0045] Example 4: An Enhanced System Integrating Spatial Guidance for Visual Health Status This embodiment enhances the functionality of the AI ​​closed-loop control module (400) based on embodiment 1. The visual health state space modeling unit (440) continuously records parameters such as ciliary muscle contraction speed, contraction amplitude, intraocular pressure fluctuation, and relaxation time for the user over 30 days, generating a set of multidimensional feature vectors daily. Principal component analysis is used to reduce the dimensionality of the multidimensional data and map it to a two-dimensional plane to construct the user's personalized "visual health state space". The user marks their "current feeling state" (1-5 points) after waking up each morning through the interaction module (600). The system automatically extracts the physiological data of the time period marked by the user as "5 points (best state)" and generates the optimal state benchmark area in the state space. The state guidance unit (450) calculates the coordinates of the current state in the state space in real time during daily use and calculates the deviation vector between the current state and the center of the optimal benchmark area. The training strategy is dynamically adjusted according to the characteristics of the deviation vector: if the contraction speed is too fast, the relaxation training time is increased; if the relaxation time is too long, dynamic adjustment training is increased; if there is a deviation from multiple dimensions, a comprehensive training scheme is adopted. Experimental data show that users who adopted state-space guidance improved their adjustment range by 37% more over 3 months than users with fixed training programs, and reduced the frequency of visual fatigue attacks by 52%.

[0046] Example 5: Integrated Design of Light-Shielding Baffle and Sensor This embodiment details the structural design of the light-shielding baffle (140). The left and right light-shielding baffles are injection molded from semi-transparent PC / ABS material with a light transmittance of 40%. The light-shielding baffle (140) extends towards the face at a 30° angle relative to the temple (110), aiming to optimize the blocking effect of lateral ambient light and ensure optimal fit between the integrated PMUT sensor and the temple area. Two grooves are provided on the inner side of the baffle for embedding the PMUT sensor. The sensor surface is covered with a 0.2mm medical-grade silicone protective layer, which ensures good fit with the skin and improves wearing comfort. After installation, the vertical distance between the transmitting / receiving surface of the PMUT sensor array (210) and the facial skin is controlled to ≤2mm to ensure effective penetration of ultrasound signals and reception of high-quality reflected signals. Two PMUT sensors are integrated on the inner side of each light-shielding baffle. The front sensor is aligned with the front of the temple to monitor the anterior ciliary muscle region; the rear sensor is aligned with the posterior of the temple to monitor the posterior ciliary muscle region and intraocular pressure changes. The sunshade and temples are magnetically connected, with three pairs of neodymium magnets at the connection point to ensure wearing stability. Users can remove the sunshade as needed: with the sunshade on, it's in immersion mode for deep rehabilitation training; without it, it's in daily use. During reinstallation, the magnetic structure automatically aligns, ensuring accurate sensor positioning. Ergonomic testing: Tested on 30 subjects, the wearing comfort score was 4.6 / 5, signal acquisition stability was 98.7%, and the lateral sunshade effect reduced ambient light interference by 76%.

[0047] Summary of differences from existing technologies 1. PMUT Glasses: Existing technology is used for blink monitoring, while this invention is used for ciliary muscle activity monitoring + closed-loop rehabilitation, achieving a fundamental leap from passive monitoring to active intervention.

[0048] 2. Frequency visual stimulation: Existing technologies are used for BCI instruction decoding, while this invention is used for rehabilitation intervention based on ciliary muscle feedback. The application scenarios, feedback mechanisms, and technical objectives are all different.

[0049] 3. Frequency intervention: Existing technologies involve fixed-frequency open-loop intervention, while this invention uses AI to identify the optimal frequency for each individual and then implements closed-loop control to achieve personalized closed-loop functionality.

[0050] 4. AI Adaptive Control: Existing technologies use multimodal stimulation control, while this invention is deeply coupled with ultrasonic MEMS sensing, rather than a simple superposition.

[0051] 5. Dimmable glasses: Existing technology uses independent dimming, while the dimming mode of this invention is intelligently linked with the rehabilitation scenario.

[0052] 6. Wireless Ecosystem: Existing technologies are mostly local standalone devices. This invention integrates a wireless communication module to achieve data cloud synchronization, remote guidance, and AI model updates.

[0053] Industrial application This invention deeply integrates ultrasonic MEMS sensing technology, AI frequency modulation, and closed-loop control, resulting in a compact structure, comfortable wear, and convenient operation. It can be widely applied in the following scenarios: myopia prevention and control in adolescents, relief of visual fatigue in adults, intervention for accommodative function decline in middle-aged and elderly individuals, amblyopia assisted intervention, post-ophthalmic rehabilitation, and visual function training centers. The system supports connection to a mobile APP via a wireless communication module (130), facilitating remote guidance and rehabilitation effect evaluation by medical personnel. With approximately 250 million potential users in China, it possesses a vast market application space.

[0054] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Those skilled in the art can make various improvements and modifications without departing from the spirit and principles of the invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A closed-loop vision rehabilitation glasses system based on ultrasonic MEMS and AI-driven frequency modulation, characterized in that, include: Eyeglasses body (100); An ultrasonic MEMS biosensing module (200) integrated into the main body of the glasses (100) is used to collect the user's ciliary muscle activity status in real time in a non-contact manner. A visual stimulation module (300) integrated into the main body of the glasses (100) is used to output visual stimulation to the user; The AI ​​closed-loop control module (400) integrated into the main body of the glasses (100) is electrically connected to the ultrasonic MEMS biosensing module (200) and the visual stimulation module (300), respectively. The wireless communication module (130) integrated into the main body of the glasses (100) is electrically connected to the AI ​​closed-loop control module (400) and is used to synchronize or exchange data with external terminal devices; The AI ​​closed-loop control module (400) includes: The data acquisition unit (410) is electrically connected to the ultrasonic MEMS biosensing module (200) and is used to receive raw signals and perform preprocessing. The individualized frequency feature library construction unit (420) is used to automatically identify and store at least one intervention frequency that corresponds to the optimal response of the user's visual system through multi-frequency stimulation scanning and real-time ciliary muscle response analysis in the initialization mode. The real-time control unit (430) is used to generate control instructions and send them to the visual stimulation module (300) based on the changes in ciliary muscle activity state collected in real time and combined with the optimal intervention frequency in the individualized frequency feature library. Visual health status space modeling unit (440) is used to construct a user's personalized visual health status space based on historically collected multi-dimensional eye parameters; The state guidance unit (450) is used to calculate in real time the deviation between the current state position in the state space and the preset optimal region, and generate guidance and control instructions based on the deviation. The visual stimulation module (300) dynamically adjusts the frequency and parameters of the visual stimulation according to the control instructions, forming a closed-loop rehabilitation circuit based on real-time feedback from the ciliary muscle.

2. The system according to claim 1, characterized in that, The ultrasonic MEMS biosensing module (200) includes an ultrasonic MEMS sensor array (210), a signal conditioning unit (220), and a data transmission unit (230); the ultrasonic MEMS sensor array (210) is arranged on the inner side of the temple (110) and the edge of the frame (120) of the main body of the glasses (100); the ultrasonic MEMS sensor array (210) adopts a piezoelectric micromechanical ultrasonic sensor (PMUT), with a working frequency of 1-10MHz and an acquisition accuracy of ≤0.01mm; after installation, the vertical distance between the PMUT sensor array (210) and the facial skin is ≤2mm.

3. The system according to claim 1, characterized in that, The individualized frequency feature library construction unit (420) includes: A multi-frequency stimulation generator (421) is used to control the visual stimulation module (300) to output visual stimulation in a preset frequency sequence, the preset frequency sequence including 6Hz, 8Hz, 10Hz, 12Hz and 15Hz. The response feature extractor (422) is used to analyze the ciliary muscle contraction speed, contraction amplitude and intraocular pressure fluctuation parameters collected by the ultrasonic MEMS biosensing module (200) under stimulation at various frequencies. The optimal frequency identifier (423) is used to automatically identify the 1-3 frequencies that are most significant in response to the user based on the response characteristics by a weighted comprehensive scoring method, and store them in the individualized frequency feature library (424); wherein, the contraction speed weight is 0.5, the accommodation amplitude weight is 0.3, and the intraocular pressure fluctuation weight is 0.

2.

4. The system according to claim 1, characterized in that, The visual stimulation module (300) includes a display unit (310) and an audio unit (320). The display unit (310) is located at the lens position of the main body of the glasses (100). The audio unit (320) is located on the inside of the temple (110) and is used to output an audio guidance signal synchronized with the visual stimulation.

5. The system according to claim 4, characterized in that, The display unit (310) is a dimmable display unit, and its transmittance can be dynamically adjusted according to the ambient light or rehabilitation scene requirements. The display unit (310) includes an electrochromic layer or a liquid crystal dimming layer, which is electrically connected to the AI ​​closed-loop control module (400) and adjusts the transmittance according to the control command.

6. The system according to claim 1, characterized in that, The visual stimulation module (300) also includes a rehabilitation linkage unit (330), which is used to bind vision rehabilitation training tasks with visual stimulation content. After the user completes a specific training action, the visual stimulation content is updated.

7. The system according to claim 1, characterized in that, The main body of the glasses (100) includes a left light-blocking baffle and a right light-blocking baffle (140), which extend from the front end of the left and right temples (110) toward the side of the face to block lateral ambient light; the light-blocking baffle (140) extends toward the side of the face at a 30° angle relative to the temples (110); at least a portion of the ultrasonic MEMS sensor array (210) is arranged inside the left and right light-blocking baffles (140); the light-blocking baffle (140) is made of a semi-transparent material and is detachably connected to the temples.

8. The system according to claim 1, characterized in that, It also includes a data storage module (500) electrically connected to the AI ​​closed-loop control module (400), used to store historical data of user ciliary muscle parameters, individualized frequency feature library and system operation log.

9. The system according to claim 1, characterized in that, It also includes an interactive module (600) electrically connected to the AI ​​closed-loop control module (400), the interactive module (600) including one or more of physical buttons, a touch screen and a voice recognition unit.

10. The system according to claim 1, characterized in that, It also includes a power module (700) that provides power to each module, the power module (700) using a rechargeable lithium battery and integrating a power management unit.

Images