Intelligent self-adaptive electronic hyperopia device mirror surface angle adjusting system and method

Abstract

The invention discloses an intelligent self-adaptive electronic hyperopia device mirror surface angle adjusting system and method, and belongs to the field of mirror surface adjustment. The invention discloses an intelligent self-adaptive electronic hyperopia device mirror surface angle adjusting system and method. The method comprises the following steps: constructing a personalized perception data pool; generating an angle adjusting instruction; a pre-adjusting instruction is generated and corrected; calculating a current visual comfort score; judging whether the visual comfort score is lower than a preset threshold value or not; optimal angle adjusting parameters are screened out; performing angle adjustment; the system comprises a multi-source sensing module, a data processing module, a precise adjusting module and a visual feedback module. The problems that dynamic parameters such as wearing postures and object viewing distances are not considered in the prior art, and adaptation deviation is prone to occurring in complex scenes are solved. According to the method, the multi-user and multi-scene adaptation precision is improved, accurate synchronization of the mirror surface angle and the sight line is realized, asthenopia characteristics are captured by eye movement tracking, automatic early warning optimization is realized, and passive adaptation is changed into active pre-judgment.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of mirror surface adjustment, in particular to an intelligent self-adaptive electronic far-vision device mirror surface angle adjustment system and method. BACKGROUND

[0002] As an important tool for correcting far-vision, the fitting accuracy and dynamic following ability of the mirror surface angle of the electronic far-vision device directly determine the user's visual experience and correction effect. With the development of intelligent wearable technology, traditional mechanical adjustment type far-vision devices have gradually evolved towards intelligentization, but the existing related invention patents still have many technical bottlenecks, which are difficult to meet the high-precision visual needs of users in diversified scenarios.

[0003] The core technical features of the existing electronic far-vision device and intelligent glasses mirror surface adjustment technology have obvious limitations. First of all, the adjustment trigger mode is single and has poor adaptability. Current mainstream technologies mostly rely on manual mechanical adjustment or fixed program calibration mode, which requires users to actively operate to complete parameter fitting, and the operation is cumbersome and the user learning cost is high. Some so-called intelligent adjustment schemes only focus on adjusting the single parameter of interpupillary distance, and do not fully consider dynamic scene parameters such as user wearing posture (such as head pitch, roll angle), viewing distance, head movement state, etc., which cannot accurately adapt to individual differences of different users and may easily cause fitting deviation in complex use scenarios. Therefore, it does not meet the existing needs, and for this purpose, we propose an intelligent self-adaptive electronic far-vision device mirror surface angle adjustment system and method. SUMMARY

[0004] The purpose of the present application is to provide an intelligent self-adaptive electronic far-vision device mirror surface angle adjustment system and method, which learns user habits and dynamically optimizes adjustment strategies by constructing a mapping model combined with reinforcement learning algorithms, improves multi-user multi-scenario fitting accuracy, realizes precise synchronization of mirror surface angle and line of sight by predicting head movement trends and correcting errors, relies on eye movement tracking to capture visual fatigue characteristics, simulates matching of adjustment parameters and visual needs, automatically warns and optimizes, realizes passive fitting to active prediction, ensures long-term visual fitting and correction effect, and solves the problems raised in the above background technology.

[0005] To achieve the above purpose, the present application provides the following technical scheme: an intelligent self-adaptive electronic far-vision device mirror surface angle adjustment method, comprising the following steps:

[0006] S1: Real-time acquisition of user physiological characteristic data, dynamic scene data and basic vision data, standardization processing and denoising processing of the collected data, and construction of a personalized perception data pool;

[0007] S2: constructing a mapping model, using the mapping model to generate an initial angle adjustment parameter adapted to the user's physiological characteristics and the initial use scene, and converting the initial angle adjustment parameter into an angle adjustment instruction;

[0008] S3: driving the mirror to adjust the angle according to the instruction, generating a pre-adjustment instruction in the adjustment process, and generating an error correction signal, and correcting the angle adjustment instruction according to the error correction signal;

[0009] S4: capturing the visual comfort related feature data of the user in real time, and calculating the current visual comfort score based on the visual comfort evaluation model constructed by simulation;

[0010] S5: determining whether the visual comfort score is lower than a preset threshold; if not, returning to S1 to dynamically optimize the angle adjustment parameter and continuously following the head movement to adjust the angle; if yes, triggering an optimization warning and entering S6;

[0011] S6: iteratively optimizing through simulation technology, generating a plurality of groups of candidate angle adjustment parameters, simulating and evaluating the visual effect of each group of candidate parameters, and selecting the optimal angle adjustment parameter with the highest visual comfort score;

[0012] S7: converting the optimal angle adjustment parameter into an optimization adjustment instruction, driving the mirror to adjust the angle according to the optimization adjustment instruction, and returning to S4 to continue monitoring the visual comfort after completing the optimization adjustment.

[0013] Preferably, the iterative optimization and simulation evaluation specifically includes:

[0014] Adjusting the value range of the angle adjustment parameter based on simulation technology with the visual comfort related feature data as the constraint condition, and generating a plurality of groups of candidate angle adjustment parameters;

[0015] For each group of candidate parameters, simulating the visual imaging effect and eye muscle stress of the user in the current use scene, and calculating the corresponding visual comfort score and correction effect score;

[0016] Calculating the comprehensive score of each group of candidate parameters by using the weighted comprehensive score method, and selecting the candidate parameter with the highest comprehensive score as the optimal angle adjustment parameter;

[0017] If there are multiple groups of candidate parameters with the same comprehensive score, the parameter with the smallest adjustment amplitude is selected as the optimal angle adjustment parameter.

[0018] Preferably, S3 specifically includes:

[0019] Receiving head posture data collected by the posture sensor every 10 ms, and inputting the continuous 10 groups of data into the LSTM time series prediction model to predict the head movement angle change within the next 50 ms;

[0020] generating a pre-adjustment instruction according to the prediction result, and controlling the electrostrictive micro-nano adjustment mechanism to perform pre-adjustment in advance;

[0021] Meanwhile, mirror real-time angle data is collected every 5 ms, and the error between the real-time angle and the target angle is calculated. When the error is greater than 0.05°, an error correction signal is generated to correct the mirror angle.

[0022] Preferably, the generation of the pre-adjustment instruction specifically includes:

[0023] Collecting head real-time posture data at a preset frequency, including the real-time values and change rates of the head pitch angle and roll angle;

[0024] Time-sequencing the collected raw data to construct a one-dimensional time-sequencing data sequence in chronological order, and removing abnormal fluctuation values in the data;

[0025] Inputting the one-dimensional time-sequencing data sequence after preprocessing into a pre-trained LSTM time-sequencing prediction model as an input sequence;

[0026] The LSTM time-sequencing prediction model learns the time-sequencing characteristics of the historical posture one-dimensional time-sequencing data sequence, mines the change rule of the head movement, and establishes the time-sequencing correlation of the posture change;

[0027] The LSTM time-sequencing prediction model calculates and outputs the head movement angle change and movement direction in a future preset time period based on the time-sequencing correlation of the posture change and the latest head posture time-sequencing data segment collected in real time, and forms a head movement trend prediction result;

[0028] According to the predicted head movement trend, the initial deviation between the current mirror angle and the target angle is converted to generate a corresponding pre-adjustment instruction to determine the adjustment direction, adjustment angle, and adjustment rate in advance.

[0029] Preferably, the training process of the LSTM time-sequencing prediction model specifically includes

[0030] Collecting head posture time-sequencing data of different people in various typical scenarios and preprocessing, and using the preprocessed head posture time-sequencing data to construct a training data set, a verification data set, and a test data set;

[0031] Constructing a time-sequencing prediction model based on an LSTM neural network, training the time-sequencing prediction model using the training set, and obtaining the final LSTM time-sequencing prediction model after training;

[0032] Inputting the verification data set into the LSTM time-sequencing prediction model obtained by training, and calculating the mean absolute error of the predicted value and the true value;

[0033] The test data set is input into the optimal LSTM time series prediction model to verify the generalization ability of the LSTM time series prediction model on new data.

[0034] Preferably, when the visual comfort score is lower than the preset threshold, it further comprises predicting the head movement trend based on the eye movement features, specifically:

[0035] When the visual comfort score is lower than the preset threshold, the eye movement tracking sensor and the attitude sensor are synchronized in hardware time stamping, and the sampling time reference is determined;

[0036] According to the sampling time reference, the eye movement trajectory data point sequence collected by the eye movement tracking sensor and the head posture data point sequence collected by the attitude sensor are time stamped and aligned to generate a sequence of synchronized data pairs;

[0037] Extract the first feature set representing the eye movement velocity vector and the second feature set representing the head angular velocity vector from the synchronized data pair sequence;

[0038] The first feature set and the second feature set are paired in a unified time window to generate a set of eye movement-head movement correlation feature pairs;

[0039] According to the eye movement-head movement correlation feature pair set, the real-time transformation matrix between the eye movement velocity vector and the head angular velocity vector is determined, and the real-time correlation mapping between the eye movement features and the head movement is established according to the real-time transformation matrix;

[0040] Analyze the real-time correlation mapping to determine the real-time correlation degree between the eye movement features and the head movement features in the real-time correlation mapping, and obtain the reference correlation degree;

[0041] Calculate the correlation deviation between the real-time correlation degree and the reference correlation degree, and compare the correlation deviation with the preset stability threshold;

[0042] When the correlation deviation is less than the preset stability threshold, the LSTM time series prediction model based on historical head posture real-time data is maintained for head movement trend prediction;

[0043] Otherwise, input the real-time collected eye movement velocity vector into the real-time correlation matrix, determine the predicted head angular velocity vector according to the real-time correlation matrix, and generate a pre-adjustment instruction for adjusting the mirror surface angle in advance according to the predicted head angular velocity vector.

[0044] Preferably, after calculating the error between the real-time angle and the target angle, it further comprises correcting the error data, specifically:

[0045] Obtain the mirror real-time angle data in the continuous sampling period

[0046] Meanwhile, target angle data corresponding to each sampling period is acquired;

[0047] An error smoothing value is calculated according to the real-time mirror angle in the consecutive sampling periods and the target angle corresponding to each sampling period;

[0048] The environmental vibration frequency and the environmental temperature in the current dynamic scene data are acquired, and an environmental interference coefficient is calculated according to the environmental vibration frequency and the environmental temperature;

[0049] The calibrated error data is calculated according to the error smoothing value and the environmental interference coefficient;

[0050] When the calibrated error coefficient is greater than the error threshold, an error correction signal is generated, and a mirror angle correction operation is performed according to the error correction signal, wherein the correction amplitude is positively correlated with the calibrated error data.

[0051] The intelligent adaptive electronic far-vision device mirror angle adjustment system is applied in an intelligent adaptive electronic far-vision device mirror angle adjustment method, and comprises:

[0052] A multi-source perception module set is configured to collect physiological feature data, dynamic scene data, and basic vision data of a user in real time, and perform format standardization processing and denoising processing on the collected data to construct a personalized perception data pool;

[0053] A data processing module is configured to generate initial angle adjustment parameters adapted to the physiological features of the user by using a mapping model, dynamically optimize the angle adjustment parameters, and generate optimized angle adjustment instructions;

[0054] A precise adjustment module is configured to perform angle adjustment according to the angle adjustment instructions, predict the head movement trend of the user in combination with the head posture data, and generate pre-adjustment instructions in advance, and simultaneously correct the adjustment error in real time;

[0055] A visual feedback module is configured to capture visual comfort-related feature data of the user, analyze the matching degree between the current angle adjustment parameters and the visual needs of the user, filter optimal angle adjustment parameters according to the matching degree, and perform angle adjustment.

[0056] Preferably, the data processing module specifically comprises:

[0057] The personalized perception data pool data and the visual comfort-related feature data are received, and the received data is buffered and preliminarily analyzed;

[0058] A mapping model is constructed based on simulation technology, the data in the personalized perception data pool is input into the mapping model for simulation iteration, and the initial angle adjustment parameters are determined through multiple iteration calculations;

[0059] Continuous collection of user usage data in different usage scenarios to build a user usage habit dataset;

[0060] Based on the reinforcement learning algorithm, the usage habit dataset is trained and learned, the association rule between the user usage habit and the angle adjustment parameter is mined, the weight coefficient of the mapping model is dynamically adjusted, and the angle adjustment parameter is personalized optimized.

[0061] Preferably, the visual feedback module specifically comprises:

[0062] A preset visual comfort threshold value is received, visual comfort related feature data is received, and a visual comfort evaluation model is built based on simulation technology;

[0063] The visual comfort related feature data is input into the evaluation model, and the current visual comfort score is calculated;

[0064] When the calculated visual comfort score is lower than the preset threshold value, an optimization warning signal is triggered, and the angle parameter is optimized according to the warning signal and the visual comfort related feature data;

[0065] The generated initial angle adjustment parameter and the optimized angle adjustment parameter are converted into recognizable electrical signal instructions, an error correction signal is received, the angle adjustment instruction is corrected in real time, and the final angle adjustment instruction is generated.

[0066] Compared with the prior art, the present application has the following advantages:

[0067] The present application continuously learns the user usage habit by the mapping model combined with the reinforcement learning algorithm, dynamically optimizes the adjustment strategy, effectively solves the adaptation deviation problem caused by the single parameter and strong generalization of the prior art, significantly improves the adaptation accuracy in different users and different scenes, combines the LSTM time series prediction algorithm and the PID closed loop control algorithm, can predict the head movement trend in advance and correct the adjustment error in real time, ensures that the mirror angle and the line of sight are accurately synchronized during head movement, effectively solves the problems of blurred vision, ghosting and other problems in the prior art in dynamic scenes, ensures the stable visual experience in dynamic scenes, captures the user visual fatigue features through the eye tracking sensor, combines the simulation analysis of the matching degree of the adjustment parameter and the visual demand, can automatically trigger the optimization warning, and realizes passive perception and warning of the adjustment effect without active operation of the user, solves the problem that the prior art lacks real-time verification link, optimizes the angle parameter through simulation iteration based on visual feedback data, generates multiple candidate schemes and simulates and evaluates the optimal scheme, realizes dynamic iterative optimization of the adjustment parameter, predicts the user visual change trend in advance, generates the adjustment parameter plan in advance, realizes the transformation from passive adaptation to active prediction, effectively solves the problem that the prior art cannot adapt to the user visual change due to the fixed calibration, and ensures the visual adaptability and correction effect in the long-term use process. Attached Figure Description

[0068] Fig. 1 This is a schematic diagram of the mirror angle adjustment method of the intelligent adaptive electronic farsightedness device of the present invention;

[0069] Fig. 2 This is a schematic diagram of the mirror angle adjustment module of the intelligent adaptive electronic farsightedness device of the present invention;

[0070] Fig. 3 This is a schematic diagram illustrating the iterative optimization and simulation evaluation of the present invention. Detailed Implementation

[0071] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0072] To address the issues of existing technologies having a single triggering method, poor adaptability, and relying mainly on manual or fixed-procedure calibration which is cumbersome, has a high learning curve, and fails to consider dynamic parameters such as wearing posture and viewing distance, making it difficult to adapt to individual differences and prone to adaptation deviations in complex scenarios, please refer to [the relevant documentation / reference]. Figs. 1-3 This embodiment provides the following technical solution:

[0073] Example 1

[0074] User A is a 35-year-old adult with a hyperopia of +2.50D and an astigmatism axis of 180°. Their daily commute involves walking and taking the subway, and they need to use an electronic farsightedness device in different sub-scenarios such as walking outdoors, reading in the subway, and waiting on the platform.

[0075] A method for adjusting the mirror angle of an intelligent adaptive electronic farsightedness device includes the following steps:

[0076] S1: Before going out, user A wears a smart adaptive electronic farsightedness device. The multi-source sensing module starts working. The miniature eye-tracking sensor collects user A's interpupillary distance (62mm) and eye movement trajectory data in real time; the posture sensor collects user A's head tilt angle (initial 0°) and lateral tilt angle (initial 0°) data; the distance sensor collects the visual distance data between user A and objects in front of them in outdoor environment (5-10m); at the same time, it calls up the basic vision data (+2.50D farsightedness, 180° astigmatism axis) that user A has pre-entered through the mobile terminal APP, performs format standardization and Kalman filter noise reduction processing, and constructs a personalized sensing data pool.

[0077] S2: Construct a mapping model based on simulation technology, input the data in the personalized perception data pool into the mapping model for simulation iteration, generate initial angle adjustment parameters (mirror pitch angle 0.2°, roll angle 0°, rotation angle 0.1°) that adapt to the user's physiological characteristics and the initial use scenario after 60 simulation iterations, and convert the initial angle adjustment parameters into angle adjustment instructions;

[0078] S3: Drive the mirror to adjust the angle according to the adjustment instructions. During the adjustment process, the angle sensor collects mirror angle data in real time. At this time, user A starts walking, and the posture sensor detects that the head pitch angle fluctuates between -5° and +5°. Based on the LSTM time series prediction algorithm, input the continuous 10 groups of head posture data, predict that the head will pitch down 3° in the next 50 ms, and generate a pre-adjustment instruction. At the same time, using the PID closed-loop control algorithm, the error between the real-time angle data of the mirror and the target angle data is calculated as 0.03°, an error correction signal is generated, the angle adjustment instruction is corrected according to the error correction signal, and the mirror pitch angle is adjusted to 0.5° in advance to ensure that user A has a clear view during walking without blurring;

[0079] S4: Real-time capture of user A's blink frequency (12 times / minute) and pupil jitter amplitude (0.3°), real-time capture of user's visual comfort related feature data, based on the visual comfort evaluation model constructed by simulation, the current visual comfort score is calculated as 8.2 points;

[0080] S5: Determine whether the visual comfort score is lower than the preset threshold; since 8.2 points > preset threshold 6 points, no optimization warning is triggered, return to S1 to continuously collect data, record user A's commonly used head posture (-5°~+5° pitch) and commonly used viewing distance (5-10m) during walking, dynamically optimize the angle adjustment parameters, and continuously follow the head movement for angle adjustment; User A arrives at the subway station, after entering the subway, the viewing distance is captured as 30-50 cm (the user starts reading the phone information), and the adjustment parameter plan that adapts to the close-range reading scenario is quickly called through the reinforcement learning algorithm, and the optimization adjustment parameters (mirror pitch angle 0.8°, roll angle 0°, rotation angle 0.2°) are generated for angle adjustment. After user A reads in the subway for 15 minutes, when the blink frequency is detected to rise to 18 times / minute and the pupil jitter amplitude rises to 0.6°, the visual comfort score is calculated as 5.8 points, which is lower than the preset threshold of 6 points, triggering the optimization warning, entering S6 based on visual feedback data,

[0081] S6: Iterative optimization through simulation technology, generate 15 groups of candidate angle adjustment parameters, evaluate the visual effect of each candidate parameter, and select the optimal angle adjustment parameter (mirror pitch angle 1.0°, roll angle 0°, rotation angle 0.25°) with the highest visual comfort score;

[0082] S7: convert the optimal angle adjustment parameter into an optimal adjustment instruction, drive the mirror to adjust the angle according to the optimal adjustment instruction, and return to S4 to continue monitoring the visual comfort after completing the optimal adjustment, the user A's blinking frequency returns to 13 times per minute, the pupil tremor amplitude returns to 0.35°, the visual comfort score improves to 7.5 points, forming a closed-loop adjustment process of perception, adjustment, feedback and optimization.

[0083] The iterative optimization and simulation evaluation specifically includes:

[0084] Based on the simulation technology, adjust the value range of the angle adjustment parameter based on the visual comfort related feature data as the constraint condition, and generate 10-20 groups of candidate angle adjustment parameters;

[0085] For each group of candidate parameters, simulate the visual imaging effect and eye muscle stress of the user in the current use scenario, and calculate the corresponding visual comfort score and correction effect score;

[0086] Calculate the comprehensive score of each group of candidate parameters by using the weighted comprehensive score method, and select the candidate parameter with the highest comprehensive score as the optimal angle adjustment parameter;

[0087] If there are multiple groups of candidate parameters with the same comprehensive score, select the parameter with the smallest adjustment amplitude as the optimal angle adjustment parameter.

[0088] S3 specifically includes:

[0089] Receive head posture data collected by the posture sensor every 10 ms, input the continuous 10 groups of data into the LSTM time series prediction model, and predict the head movement angle change within the next 50 ms;

[0090] Generate a pre-adjustment instruction based on the prediction result to control the electrostrictive micro-nano adjustment mechanism to pre-adjust in advance;

[0091] At the same time, collect real-time angle data of the mirror every 5 ms, calculate the error between the real-time angle and the target angle, and generate an error correction signal when the error is > 0.05° to correct the mirror angle.

[0092] Based on the LSTM time series prediction algorithm, the head movement trend is predicted based on the collected real-time head posture data to generate a pre-adjustment instruction, specifically including:

[0093] Collect real-time posture data of the head at a preset frequency, including the real-time values and change rates of the head pitch angle and roll angle;

[0094] The collected raw data is time-sequenced, a one-dimensional time sequence data sequence is constructed according to time sequence, and abnormal fluctuation values in the data are removed to ensure the continuity and stability of the time sequence, thereby laying a data foundation for subsequent prediction analysis;

[0095] The one-dimensional time sequence data sequence after preprocessing is input into the LSTM time sequence prediction model which has been trained in advance as an input sequence;

[0096] The LSTM time sequence prediction model learns the time sequence characteristics of the historical posture one-dimensional time sequence data sequence, mines the change rule of the head movement, and establishes the time sequence correlation of the posture change;

[0097] The LSTM time sequence prediction model calculates and outputs the head movement angle change and movement direction in a future preset time period based on the time sequence correlation of the posture change and the latest head posture time sequence data segment collected in real time, and forms a head movement trend prediction result;

[0098] According to the predicted head movement trend, the initial deviation between the current mirror angle and the target angle is converted to generate a corresponding pre-adjustment instruction to determine the adjustment direction, adjustment angle and adjustment rate in advance.

[0099] The training process of the LSTM time sequence prediction model specifically includes

[0100] The head posture time sequence data of different people in various typical scenes is collected and preprocessed, and the preprocessed head posture time sequence data is used to construct a training data set, a verification data set and a test data set;

[0101] A time sequence prediction model is constructed based on an LSTM neural network, and the model includes an input layer, three LSTM hidden layers and an output layer; the number of input layer neurons is 6 (corresponding to the head pitch angle, pitch angle change rate, roll angle, roll angle change rate and comprehensive posture features of the previous two time steps); the first layer of LSTM hidden layer is set to 64 neurons, the second layer is set to 32 neurons, and the third layer is set to 16 neurons; each hidden layer uses a Dropout mechanism (Dropout rate is set to 0.2) to prevent overfitting, and the activation function uses a tanh function; the number of output layer neurons is 4, corresponding to the head pitch angle prediction value and roll angle prediction value at two time points of 50ms and 100ms in the future;

[0102] The time series prediction model is trained by using the training set, the mean square error is used as the loss function, the Adam optimizer is used for model parameter optimization, the initial learning rate is set to 0.001, and the learning rate decay strategy is adopted (the learning rate is reduced to 0.9 of the original value every 100 epochs); In the training process, the early stopping mechanism is adopted, when the loss function value of the validation data set does not decrease continuously for 20 epochs, the training is stopped, and the current optimal model parameters are saved; The number of LSTM hidden layer neurons, the Dropout rate and the learning rate are optimized by grid search method to determine the optimal hyperparameter combination, and the final LSTM time series prediction model is obtained after training;

[0103] The validation data set is input into the LSTM time series prediction model trained, and the mean absolute error of the predicted value and the true value is calculated;

[0104] The test data set is input into the optimal LSTM time series prediction model, and the generalization ability of the LSTM time series prediction model on new data is verified.

[0105] The intelligent adaptive electronic far-vision device mirror surface angle adjustment system is applied in the intelligent adaptive electronic far-vision device mirror surface angle adjustment method, comprising:

[0106] A multi-source perception module set is integrated on the frame of the electronic far-vision device, used for real-time collection of physiological characteristic data, dynamic scene data and basic vision data of the user, and format standardization processing and denoising processing of the collected data, and construction of a personalized perception data pool;

[0107] A data processing module is used for constructing a mapping model based on simulation technology, inputting data in the personalized perception data pool into the mapping model for simulation iteration, generating initial angle adjustment parameters adapted to the physiological characteristics of the user; At the same time, the use habit data of the user is continuously learned through the reinforcement learning algorithm, the angle adjustment parameters are dynamically optimized, and the optimized angle adjustment instructions are generated;

[0108] A precise adjustment module is used for angle adjustment according to the angle adjustment instructions, based on the LSTM time series prediction algorithm and the PID closed-loop control algorithm, in combination with the head posture data collected by the multi-source perception module, to predict the head movement trend of the user and generate pre-adjustment instructions in advance, and to correct the adjustment error in real time;

[0109] A visual feedback module is used for capturing visual comfort related feature data of the user through eye tracking, and transmitting the data to the data processing module. The data processing module analyzes the matching degree of the current angle adjustment parameters and the visual demand of the user based on simulation technology, triggers optimization warning when the matching degree is lower than the preset threshold, and performs simulation iterative optimization based on the visual comfort related feature data, generates multiple candidate angle adjustment parameters, and evaluates the visual effect through simulation, selects the optimal angle adjustment parameter, and performs angle adjustment.

[0110] a data processing module, specifically comprising:

[0111] receiving personalized perception data pool data and visual comfort related feature data, buffering and preliminarily analyzing the received data;

[0112] constructing a mapping model based on simulation technology, the input parameters of the mapping model including interpupillary distance, eye movement trajectory, head posture data, viewing distance data, hyperopia degree and astigmatism axis, and the output parameter being a mirror angle adjustment parameter;

[0113] inputting the data in the personalized perception data pool into the mapping model for simulation iteration, and determining the initial angle adjustment parameter through multiple iteration calculations;

[0114] continuously collecting the use data of the user in different use scenarios, including use time, use environment, commonly used viewing distance, commonly used head posture and adjustment parameter use frequency, and constructing a user use habit data set;

[0115] training and learning the use habit data set based on reinforcement learning algorithm, mining the association rule between the user use habit and the angle adjustment parameter, dynamically adjusting the weight coefficient of the mapping model, and realizing the personalized optimization of the angle adjustment parameter.

[0116] a visual feedback module, specifically comprising:

[0117] presetting a visual comfort threshold, receiving visual comfort related feature data, and constructing a visual comfort evaluation model based on simulation technology;

[0118] inputting the visual comfort related feature data into the evaluation model to calculate the current visual comfort score;

[0119] when the calculated visual comfort score is lower than the preset threshold, triggering an optimization warning signal, and optimizing the angle parameter according to the warning signal and the visual comfort related feature data;

[0120] converting the generated initial angle adjustment parameter and the optimized angle adjustment parameter into recognizable electrical signal instructions, receiving error correction signals at the same time, and correcting the angle adjustment instructions in real time to generate the final angle adjustment instructions

[0121] Embodiment 2

[0122] User B is a 60-year-old old man, with a hyperopia degree of +4.00D and an astigmatism axis of 90°. His daily commuting method is walking + taking the bus, and he needs to use the electronic hyperopia device in different sub-scenarios such as walking in strong outdoor light, resting in the bus, and checking the bus stop on the platform. In addition, the user B's vision has a slow downward trend.

[0123] The intelligent self-adaptive electronic far-vision device mirror surface angle adjusting method comprises the following steps:

[0124] S1: User B wears the intelligent self-adaptive electronic far-vision device, the multi-source perception module is started, the micro eye movement tracking sensor collects the interpupillary distance (58 mm) and eye movement trajectory data of user B; the posture sensor collects the initial head posture data (pitch 0°, roll 0°); the distance sensor collects the visual distance data (3-5 m) of user B from the front bus stop in an outdoor strong light environment; the basic vision data (+4.00D far vision degree, 90° astigmatism axis) of user B is called, data standardization and denoising processing are completed, a personalized perception data pool is constructed and transmitted to the data processing module;

[0125] S2: A mapping model is constructed based on simulation technology, the personalized perception data pool data is input into the mapping model, and after 55 simulation iterations, initial angle adjusting parameters (mirror surface pitch angle 0.5°, roll angle 0.1°, rotation angle 0.3°) are generated, and the initial angle adjusting parameters are converted into angle adjusting instructions;

[0126] S3: The mirror surface is driven to adjust the angle according to the adjusting instructions, and the angle sensor feeds back the angle data in real time; user B starts walking to the bus stop, and the posture sensor detects that the head pitch angle fluctuates between -3° and +3°, predicts that the head will pitch up by 2° in the next 50 ms, and generates a pre-adjusting instruction; the adjusting error is calculated to be 0.02° by using the PID closed-loop control algorithm, the adjusting instruction is corrected, and the mirror surface is controlled to adjust the pitch angle to 0.7° in advance, so that user B can clearly see the front bus stop;

[0127] S4: The visual comfort characteristics of user B are monitored, the blinking frequency is 10 times per minute, the pupil tremor amplitude is 0.25°, and the visual comfort score is calculated to be 8.5;

[0128] S5: If the score is higher than the threshold value, the data is continuously collected, the commonly used visual distance (3-5 m) and head posture (-3°~+3° pitch) of user B when looking at the bus stop are recorded, user B takes the bus and rests on the seat, the head posture becomes pitch -10° (leaning against the back of the chair), the distance sensor collects the visual distance as 2-3 m (looking at the interior scene), and based on the use habits learned through reinforcement learning, the adjusting parameters (mirror surface pitch angle 1.2°, roll angle 0.1°, rotation angle 0.35°) suitable for the rest scene are quickly generated, and the adjustment is completed; After user B rests for 20 minutes, it is detected that the blinking frequency rises to 20 times per minute, the pupil tremor amplitude is 0.7°, the visual comfort score is 5.2, and the optimization warning is triggered;

[0129] S6: Based on the visual feedback data, 20 groups of candidate parameters are generated through simulation, and the optimal parameters (mirror pitch angle 1.5°, roll angle 0.1°, rotation angle 0.4°) are selected after simulation evaluation;

[0130] S7: Convert the optimal angle adjustment parameters into optimization adjustment instructions, drive the mirror to adjust the angle according to the optimization adjustment instructions, and complete the optimization adjustment; After adjustment, the visual comfort score is improved to 7.8 points.

[0131] Example 3

[0132] User C is a 12-year-old teenager with a hyperopia degree of +1.75D and an astigmatism axis of 120°. The daily commuting method is walking + shared bicycle, and the electronic hyperopia device needs to be used in different sub-scenes such as outdoor cycling, waiting for traffic lights, checking mobile phone navigation, etc. User C has a high head movement frequency, and the dynamic scene demand is prominent.

[0133] The intelligent adaptive electronic hyperopia device mirror angle adjustment method comprises the following steps:

[0134] S1: User C wears an intelligent adaptive electronic hyperopia device, and a multi-source perception module is started. A miniature eye movement tracking sensor collects the interpupillary distance (55mm) and eye movement trajectory data of user C. A posture sensor collects the initial head posture data (pitch 0°, roll 0°). A distance sensor collects the visual distance data of user and the front road and vehicle in the outdoor cycling environment (10-20m). The basic vision data of user C (hyperopia degree +1.75D, astigmatism axis 120°) is called and processed to complete data standardization and denoising, and a personalized perception data pool is constructed.

[0135] S2: Input the data into the mapping model, generate the initial angle adjustment parameters (mirror pitch angle 0.3°, roll angle 0°, rotation angle 0.2°) through 58 times of simulation iteration, and convert them into adjustment instructions;

[0136] S3: According to the adjustment instructions, drive the mirror to adjust the angle, and the angle sensor feedbacks the data in real time. User C starts cycling, and the head movement frequency is high. The posture sensor detects that the head pitch angle fluctuates between -10°~+10° and the roll angle fluctuates between -5°~+5°, so the data acquisition and prediction frequency is increased (collect data every 5ms, and predict every 10ms). It is predicted that the head will roll 4° to the left within the next 50ms, and a pre-adjustment instruction is generated. The PID closed-loop control algorithm is used to calculate the adjustment error of 0.04°, quickly generate an error correction signal, correct the adjustment instruction, and control the mirror to adjust the roll angle to 0.4° synchronously, ensuring that the field of view is stable during cycling without ghosting.

[0137] S4: Monitor the visual comfort features of user C. Due to the user's focus during the exercise mode, the blink rate is 8 times per minute, the pupil tremor amplitude is 0.2°, and the visual comfort score is 8.8 points;

[0138] S5: The score is higher than the threshold, and the data is continuously collected. The head movement pattern and the commonly used viewing distance (10-20m) of user C are recorded when riding; user C stops at a red light intersection and checks the phone navigation, and the distance sensor collects the viewing distance as 25cm. The adjustment parameters for viewing navigation at close range are generated (mirror pitch angle 0.9°, roll angle 0°, rotation angle 0.3°), and the adjustment is completed. After user C checks the navigation for 10 minutes, the blink rate is detected to rise to 16 times per minute, the pupil tremor amplitude is 0.5°, the visual comfort score is 5.9 points, and the optimization warning is triggered;

[0139] S6: Based on the visual feedback data, 18 groups of candidate parameters are generated through simulation, and the optimal parameters (mirror pitch angle 1.1°, roll angle 0°, rotation angle 0.35°) are selected after simulation evaluation;

[0140] S7: Convert the optimal angle adjustment parameters into optimization adjustment instructions, and complete the optimization adjustment; after adjustment, the visual comfort score is improved to 7.6 points.

[0141] In one embodiment, an intelligent self-adaptive electronic far vision device mirror angle adjustment method is provided, and when the visual comfort score is lower than the preset threshold, it further includes predicting the head movement trend based on the eye movement features, specifically:

[0142] When the visual comfort score is lower than the preset threshold, the eye movement tracking sensor and the attitude sensor are synchronized in hardware time stamp, and the sampling time reference is determined;

[0143] According to the sampling time reference, the eye movement trajectory data point sequence collected by the eye movement tracking sensor and the head posture data point sequence collected by the attitude sensor are time stamped and aligned to generate a sequence of synchronized data pairs;

[0144] Extract the first feature set representing the eye movement velocity vector and the second feature set representing the head angular velocity vector from the synchronized data pair sequence;

[0145] Pair the first feature set and the second feature set in a unified time window to generate a set of eye movement-head movement correlation feature pairs;

[0146] Determine the real-time transformation matrix between the eye movement velocity vector and the head angular velocity vector according to the eye movement-head movement correlation feature pair set, and establish the real-time correlation mapping between the eye movement features and the head movement according to the real-time transformation matrix;

[0147] Analyze the real-time correlation map to determine the real-time correlation degree between the eye movement features and the head movement features in the real-time correlation map, and simultaneously obtain a reference correlation degree;

[0148] Calculate the correlation deviation between the real-time correlation degree and the reference correlation degree, and compare the correlation deviation with a preset stability threshold;

[0149] When the correlation deviation is less than the preset stability threshold, the LSTM time series prediction model based on historical head posture real-time data is maintained for head movement trend prediction;

[0150] Otherwise, the eye movement velocity vector collected in real time is input into the real-time correlation matrix, and a predicted head angular velocity vector is determined according to the real-time correlation matrix, and a pre-adjustment instruction for adjusting the mirror surface angle in advance is generated according to the predicted head angular velocity vector.

[0151] In this embodiment, determining the real-time transformation matrix between the eye movement velocity vector and the head angular velocity vector according to the set of eye movement-head movement correlation feature pairs comprises: taking the eye movement velocity vector and the head angular velocity vector at each time point in the set of eye movement-head movement correlation feature pairs as input vectors and target vectors, respectively; initializing a weight matrix whose dimensions are determined by the dimensions of the input vectors and the target vectors, and initializing it as an identity matrix; inputting the input vector and the target vector at the first time point into a recursive update formula; the recursive update formula first calculates a predicted target vector according to the current weight matrix and the input vector; then calculates a residual vector between the predicted target vector and the actual target vector; updates the element values of the weight matrix according to the residual vector and a dynamically adjusted gain vector; takes the input vector and the target vector at the next time point as new inputs, and repeats the steps of prediction, residual calculation and weight matrix update; after processing all data points in the time window, the finally updated weight matrix is output as the real-time transformation matrix.

[0152] In this embodiment, the visual comfort score refers to a quantitative score obtained by evaluating the user's visual state, which is used to measure the comfort level when using electronic far vision devices.

[0153] In this embodiment, the preset threshold refers to a pre-set visual comfort score threshold value, which is used to trigger the subsequent eye movement and head movement analysis process.

[0154] In this embodiment, hardware timestamp synchronization refers to aligning the internal clocks of the eye movement tracking sensor and the posture sensor to ensure data collection time consistency.

[0155] In this embodiment, the sampling time reference refers to a unified time point or time sequence as a reference for time alignment.

[0156] In this embodiment, the synchronized data pair sequence refers to the paired data sequence formed by aligning eye-tracking trajectory data points and head posture data points with timestamps.

[0157] In this embodiment, the first feature set refers to the set of features extracted from eye-tracking trajectory data that characterize the eye-tracking velocity vector.

[0158] In this embodiment, the second feature set refers to the feature set extracted from the head posture data that characterizes the head angular velocity vector.

[0159] In this embodiment, the eye-movement-head motion associated feature pair set refers to the feature pair set generated after the first feature set and the second feature set are paired within a unified time window.

[0160] In this embodiment, the real-time transformation matrix refers to the mathematical matrix that describes the real-time mapping relationship between the eye movement velocity vector and the head angular velocity vector.

[0161] In this embodiment, real-time association mapping refers to the dynamic association between eye movement features and head movement features established through a real-time transformation matrix.

[0162] In this embodiment, real-time correlation refers to the degree of real-time correlation between current eye movement features and head movement features.

[0163] In this embodiment, the baseline correlation refers to the degree of reference correlation between eye movement features and head movement features under stable or normal conditions.

[0164] In this embodiment, the correlation deviation refers to the amount of deviation between the real-time correlation degree and the baseline correlation degree.

[0165] In this embodiment, the preset stability threshold refers to a preset value used to determine whether the correlation deviation is within a stable range.

[0166] In this embodiment, the LSTM time-series prediction model based on historical head posture real-time data refers to a machine learning model that uses a long short-term memory network to predict future head movement trends based on historical head posture data.

[0167] In this embodiment, the predicted head angular velocity vector refers to the future head angular velocity vector calculated through a real-time correlation matrix or LSTM model.

[0168] In this embodiment, the pre-adjustment command refers to the control command generated based on the predicted head angular velocity vector, used to adjust the mirror angle of the electronic farsightedness device in advance.

[0169] The working principle and beneficial effects of the above technical solution are as follows: Accurate alignment of eye-tracking and head-movement data is achieved through hardware timestamp synchronization, and feature sets of eye-tracking velocity vectors and head angular velocity vectors are extracted to establish a real-time correlation mapping. When the visual comfort score is low, the correlation deviation is calculated to assess data stability: if the deviation is small, the LSTM model based on historical data is used to predict head trends; if the deviation is large, the head angular velocity is directly predicted based on real-time eye-tracking features through a transformation matrix, and a mirror angle pre-adjustment command is generated. This effectively improves the accuracy and timeliness of head-movement trend prediction, realizes intelligent pre-adjustment of mirror angles, thereby significantly enhancing visual comfort, reducing image lag or blurring, and improving the device's adaptive response capability.

[0170] In one embodiment, a method for adjusting the mirror angle of an intelligent adaptive electronic farsightedness device is provided. After calculating the error between the real-time angle and the target angle, the method further includes correcting the error data, specifically:

[0171] Obtain real-time mirror angle data over m consecutive sampling periods. , ... ,in, This represents the real-time angle data of the mirror surface during the first sampling period; This represents the real-time angle data of the mirror surface during the second sampling period; This represents the real-time angle data of the mirror surface during the m-th sampling period;

[0172] Simultaneously, acquire the target angle data corresponding to m sampling periods. , ... ,in, This represents the target angle data for the first sampling period; This represents the target angle data for the second sampling period; This represents the target angle data in the m-th sampling period;

[0173] The error smoothing value is calculated based on the real-time angle of the mirror surface within m consecutive sampling periods and the target angle corresponding to each sampling period.

[0174] ;

[0175] in, Indicates the error smoothing value; This represents the real-time angle data of the mirror surface during the i-th sampling period; This represents the target angle data corresponding to the i-th sampling period; i represents the index of the sampling period.

[0176] Obtain the environmental vibration frequency and ambient temperature from the current dynamic scene data, and calculate the environmental interference coefficient based on the environmental vibration frequency and ambient temperature;

[0177] ;

[0178] in, Indicates the environmental interference coefficient; This indicates the environmental vibration frequency in the current dynamic scene data; Indicates the standard environmental vibration frequency; Indicates ambient temperature; Indicates the standard operating temperature; , Represents the weighting coefficient, and ;

[0179] Calculate the calibrated error data based on the error smoothing value and the environmental interference coefficient;

[0180] ;

[0181] in, This represents the calibrated error data;

[0182] When the calibrated error coefficient When the error correction signal is generated, a mirror angle correction operation is performed based on the error correction signal. The correction magnitude is positively correlated with the calibrated error data.

[0183] In this embodiment, The weighting coefficient representing the environmental vibration frequency; environmental vibration frequency Relative to standard environmental vibration frequency The degree of deviation contributes to the interference of mirror angle adjustment error; its value ranges from 0 to... <1, When vibration interference is dominant in the usage scenario of the electronic farsightedness device (such as outdoor cycling, bumpy public transportation, etc.), The value is closer to 1 (e.g.) =0.7), which means that the influence of vibration frequency on error is given a higher weight.

[0184] In this embodiment, This indicates the deviation of ambient temperature from the weighting coefficient, representing the ambient temperature. Relative to standard operating temperature The degree of deviation contributes to the interference of mirror angle adjustment error. Its value ranges from 0 to... <1, when temperature fluctuations dominate the usage scenario (such as switching between high-temperature outdoor and low-temperature indoor / outdoor environments), The value is closer to 1 (e.g.) =0.6), which means that the impact of temperature deviation on the error is given a higher weight.

[0185] In this embodiment, the current environmental vibration frequency vibration frequency with standard environment The relationship between them is: through Quantify the intensity level of the current vibration disturbance, where, when When the ratio is greater than 1, it indicates that the current vibration intensity exceeds the standard state, and the degree of disturbance is enhanced; when When the ratio is 1, it indicates that the current vibration is at a standard level with no additional interference; when If the ratio is less than 1, it indicates that the current vibration intensity is lower than the standard state and the degree of interference is weak.

[0186] In this embodiment, the positive correlation between the correction magnitude and the calibrated error data can be expressed as follows: Let the correction magnitude be... (unit: ), calibrated error data (unit: The positive linear correlation between the two satisfies the formula: ;in, Represents the proportionality coefficient ( To ensure a positive correlation, the response sensitivity of the electronic farsightedness device is determined by the performance of its adjustment mechanism (such as the response sensitivity of an electrostrictive micro / nano adjustment mechanism), and requires experimental calibration, for example... ,express Each increase The correction range increased. ; Baseline correction amount The minimum response threshold used to compensate for the regulating mechanism (e.g.) ,avoid A value less than the minimum adjustment threshold for a single adjustment results in a correction amplitude lower than the minimum adjustment accuracy of the mechanism; simultaneously... , The maximum single correction range of the regulating mechanism, such as To prevent due to Exceeding the maximum threshold for a single adjustment leads to over-adjustment.

[0187] In this embodiment, the environmental interference coefficient is a quantitative representation of interference factors such as vibration and temperature in the environment. Among them, the vibration frequency will affect the sensor acquisition accuracy and the response stability of the adjustment mechanism, and the temperature will cause thermal expansion and contraction of the frame material or performance drift of the adjustment mechanism (such as electrostrictive micro-nano mechanism). Both of these will make the actual error greater than the statistical result. Therefore, it is necessary to quantify the degree of influence of additional interference on the actual error through the environmental interference coefficient.

[0188] The working principle and beneficial effects of the above technical solution are as follows: In the dynamic calibration of error correction signal, the influence of environmental interference is quantified, a positive linear relationship between correction error and correction amplitude is established and constraints are set, and each link is progressively advanced to realize real-time perception of user visual state and environmental interference, ensuring the accuracy, stability and dynamic adaptability of mirror angle adjustment, and improving the correction effect and user experience of electronic farsightedness device.

[0189] In summary, the three embodiments described above are all based on the unified usage scenario of users' daily commutes, and are designed for users of different ages, vision conditions, and needs in different sub-scenarios. They fully present the working process of the intelligent adaptive electronic farsightedness device mirror angle adjustment system and method of the present invention, verify the precise adjustment capability, personalized adaptation capability and closed-loop optimization effect of the present invention in dynamic scenarios, and fully demonstrate the technical advantages and application value of the present invention.

[0190] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not specifically listed, or elements inherent to such process, method, article, or apparatus.

[0191] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention.

Claims

1. A method for adjusting the mirror angle of an intelligent adaptive electronic farsightedness device, characterized in that, Includes the following steps: S1: Real-time collection of users' physiological characteristic data, dynamic scene data and basic vision data; standardization and noise reduction of the collected data; construction of a personalized perception data pool. S2: Construct a mapping model, use the mapping model to generate initial angle adjustment parameters that adapt to the user's physiological characteristics and initial usage scenario, and convert the initial angle adjustment parameters into angle adjustment commands; S3: Drive the mirror to adjust the angle according to the command. During the adjustment process, a pre-adjustment command is generated, and an error correction signal is generated at the same time. The angle adjustment command is corrected according to the error correction signal. S4: Capture user visual comfort-related feature data in real time, and calculate the current visual comfort score based on the visual comfort evaluation model built by simulation; S5: Determine if the visual comfort score is lower than the preset threshold; if not, return to S1 to dynamically optimize the angle adjustment parameters and continuously adjust the angle according to head movement; if yes, trigger the optimization warning and proceed to S6. S6: Iterative optimization is performed through simulation technology to generate multiple sets of candidate angle adjustment parameters. Visual effect simulation evaluation is performed on each set of candidate parameters to select the optimal angle adjustment parameter with the highest visual comfort score. S7: Convert the optimal angle adjustment parameters into an optimized adjustment command, drive the mirror to adjust the angle according to the optimized adjustment command, and return to S4 after the optimized adjustment is completed to continue visual comfort monitoring.

2. The method for adjusting the mirror angle of an intelligent adaptive electronic farsightedness device according to claim 1, characterized in that, The iterative optimization and simulation evaluation specifically include: Using visual comfort-related feature data as constraints, the range of angle adjustment parameters is adjusted based on simulation technology to generate multiple sets of candidate angle adjustment parameters; For each set of candidate parameters, the simulation simulates the visual imaging effect and eye muscle stress of the user in the current usage scenario, and calculates the corresponding visual comfort score and correction effect score. The weighted comprehensive scoring method is used to calculate the comprehensive score of each group of candidate parameters, and the candidate parameter with the highest comprehensive score is selected as the optimal angle adjustment parameter. If multiple candidate parameters have the same overall score, the parameter with the smallest adjustment range is selected as the optimal angle adjustment parameter.

3. The method for adjusting the mirror angle of an intelligent adaptive electronic farsightedness device according to claim 2, characterized in that, S3 specifically includes: The system receives head posture data from the posture sensor every 10ms and inputs 10 consecutive sets of data into the LSTM time-series prediction model to predict the change in head motion angle within the next 50ms. Based on the prediction results, a pre-adjustment command is generated to control the electrostrictive micro / nano adjustment mechanism to perform pre-adjustment in advance. Simultaneously, real-time mirror angle data is collected every 5ms, and the error between the real-time angle and the target angle is calculated. When the error is greater than 0.05°, an error correction signal is generated to correct the mirror angle.

4. The method for adjusting the mirror angle of the intelligent adaptive electronic farsightedness device according to claim 3, characterized in that, The generation of pre-adjustment instructions specifically includes: Real-time head posture data is collected at a preset frequency, including the real-time values ​​and rate of change of head pitch and tilt angles. The collected raw data is processed in a time series manner to construct a one-dimensional time series data sequence in chronological order, while removing abnormal fluctuation values ​​from the data. The preprocessed one-dimensional time series data sequence is used as the input sequence and input into the pre-trained LSTM time series prediction model; The LSTM temporal prediction model learns the temporal features of historical one-dimensional temporal data sequences to uncover the changing patterns of head movements and establish temporal correlations of posture changes. The LSTM time-series prediction model calculates and outputs the change in head motion angle and direction within a preset time period based on the temporal correlation of posture changes and the latest head posture time-series data segments collected in real time, thus forming a head motion trend prediction result. Based on the predicted head movement trend, combined with the initial deviation between the current mirror angle and the target angle, the corresponding pre-adjustment command is generated to determine the pre-adjustment direction, adjustment angle, and adjustment rate.

5. The method for adjusting the mirror angle of an intelligent adaptive electronic farsightedness device according to claim 3, characterized in that, The training process of the LSTM time series prediction model specifically includes: We collected and preprocessed head pose time-series data from different groups of people in various typical scenarios, and used the preprocessed head pose time-series data to construct training datasets, validation datasets and test datasets. A time series prediction model is constructed based on an LSTM neural network. The time series prediction model is trained using a training set, and the final LSTM time series prediction model is obtained after training. Input the validation dataset into the trained LSTM time series prediction model and calculate the mean absolute error between the predicted values ​​and the true values. Input the test dataset into the optimal LSTM time series prediction model to verify the generalization ability of the LSTM time series prediction model on new, unseen data.

6. The method for adjusting the mirror angle of an intelligent adaptive electronic farsightedness device according to claim 1 is characterized in that, When the visual comfort score is lower than a preset threshold, the method further includes predicting head movement trends based on eye movement features, specifically: When the visual comfort score is lower than a preset threshold, the eye-tracking sensor and the posture sensor are synchronized with hardware timestamps, and a sampling time reference is determined. Based on the sampling time reference, the sequence of eye trajectory data points collected by the eye tracking sensor and the sequence of head posture data points collected by the posture sensor are timestamped to generate a synchronized data pair sequence. Extract the first feature set representing the eye movement velocity vector and the second feature set representing the head angular velocity vector from the synchronized data pair sequence; The first feature set and the second feature set are paired within a unified time window to generate a set of eye-tracking-head motion associated feature pairs; Based on the eye-head motion correlation features, determine the real-time transformation matrix between the eye movement velocity vector and the head angular velocity vector, and establish the real-time correlation mapping between eye movement features and head motion based on the real-time transformation matrix; The real-time correlation mapping is analyzed to determine the real-time correlation degree between eye movement features and head movement features in the real-time correlation mapping, and at the same time, the baseline correlation degree is obtained. Calculate the correlation deviation between the real-time correlation degree and the baseline correlation degree, and compare the correlation deviation with a preset stability threshold; When the correlation deviation is less than the preset stability threshold, the LSTM time series prediction model based on historical head posture real-time data is maintained to predict the head movement trend. Otherwise, the real-time eye movement velocity vector is input into the real-time correlation matrix, and the predicted head angular velocity vector is determined based on the real-time correlation matrix. Based on the predicted head angular velocity vector, a pre-adjustment instruction for adjusting the mirror angle in advance is generated.

7. The method for adjusting the mirror angle of an intelligent adaptive electronic farsightedness device according to claim 3, characterized in that, After calculating the error between the real-time angle and the target angle, the process also includes correcting the error data, specifically: Acquire real-time mirror angle data within consecutive sampling periods At the same time, acquire the target angle data corresponding to each sampling period; The error smoothing value is calculated based on the real-time angle of the mirror in consecutive sampling periods and the target angle corresponding to each sampling period. Obtain the environmental vibration frequency and ambient temperature from the current dynamic scene data, and calculate the environmental interference coefficient based on the environmental vibration frequency and ambient temperature; Calculate the calibrated error data based on the error smoothing value and the environmental interference coefficient; When the calibrated error coefficient When the error correction signal is generated, a mirror angle correction operation is performed based on the error correction signal. The correction magnitude is positively correlated with the calibrated error data.

8. An intelligent adaptive electronic hyperopia device mirror angle adjustment system, applied in the intelligent adaptive electronic hyperopia device mirror angle adjustment method as described in claim 7, characterized in that, include: The multi-source perception module set is used to collect users' physiological characteristic data, dynamic scene data and basic vision data in real time, and to perform format standardization and noise reduction processing on the collected data to build a personalized perception data pool. The data processing module is used to generate initial angle adjustment parameters adapted to the user's physiological characteristics using a mapping model, dynamically optimize the angle adjustment parameters, and generate optimized angle adjustment instructions. The precision adjustment module is used to adjust the angle according to the angle adjustment command, combine head posture data to predict the user's head movement trend and generate pre-adjustment commands in advance, while correcting the adjustment error in real time. The visual feedback module is used to capture user visual comfort-related feature data, analyze the matching degree between the current angle adjustment parameters and the user's visual needs, select the optimal angle adjustment parameters based on the matching degree, and adjust the angle accordingly.

9. The intelligent adaptive electronic farsightedness device mirror angle adjustment system according to claim 8, characterized in that, The data processing module specifically includes: Receive personalized perception data and visual comfort-related feature data, and cache and perform preliminary analysis on the received data; A mapping model is constructed based on simulation technology. Data from the personalized perception data pool is input into the mapping model for simulation iteration. The initial angle adjustment parameters are determined through multiple iterations. Continuously collect user usage data in different usage scenarios to build a user usage habit dataset; The algorithm is used to train and learn from a dataset of user habits using reinforcement learning. This allows for the discovery of the correlation between user habits and angle adjustment parameters. The weight coefficients of the mapping model are then dynamically adjusted to optimize the angle adjustment parameters in a personalized manner.

10. The intelligent adaptive electronic farsightedness device mirror angle adjustment system according to claim 8, characterized in that, The visual feedback module specifically includes: A visual comfort threshold is preset, visual comfort-related feature data is received, and a visual comfort evaluation model is constructed based on simulation technology. Input visual comfort-related feature data into the evaluation model to calculate the current visual comfort score; When the calculated visual comfort score is lower than the preset threshold, an optimization warning signal is triggered, and the angle parameters are optimized based on the warning signal and visual comfort-related feature data. The generated initial angle adjustment parameters and optimized angle adjustment parameters are converted into recognizable electrical signal commands. At the same time, error correction signals are received to correct the angle adjustment commands in real time and generate the final angle adjustment commands.

Images