Dioptric compensation method and device based on bimodal perception

Abstract

The invention relates to the technical field of diopter adjustment, and discloses a diopter compensation method and device based on bimodal perception, and the method comprises the steps: firstly building a head-mounted equipment body coordinate system, defining a first region and a second region, enabling a user to carry out the round-trip watching switching between the two regions, and enabling the user to carry out the diopter adjustment. Bimodal sensing data including a physiological signal modal and a spatial motion modal are collected; an emergency or steady adjustment event is judged based on bimodal sensing data, the former triggers an advanced compensation acceleration mechanism, and the latter adopts a conventional compensation mechanism; according to the method, through bimodal data fusion and a layered compensation strategy, the problem that a traditional compensation algorithm and physiological reaction are not synchronous is solved, and in a high-frequency fixation switching scene, invalid adjustment actions can be effectively reduced, ciliary muscle fatigue can be relieved, and refraction compensation accuracy and physiological suitability can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of diopter adjustment, and more specifically, to a diopter compensation method and device based on dual-modal perception. Background Art

[0002] In the prior art, the Chinese patent application with publication number CN118131475A discloses a head-mounted display device and a method for detecting diopter, pupil distance compensation, and sight line adjustment. The disclosed head-mounted display device obtains diopter by detecting the identification component of the refractive correction lens to achieve pupil distance compensation and sight line adjustment. The Chinese patent with authorization announcement number CN115736814B proposes a naked eye retinal diopter fitting device, lens, method and storage medium. The naked eye retinal diopter fitting method in the prior art realizes personalized refractive correction by fitting the rotational symmetry surface.

[0003] However, with the increasing demand for high-frequency gaze switching in education and life scenarios, users' frequent gaze switching between near and far distances will cause ciliary muscle adjustment lags. For example, students' frequent gaze switching between desks and blackboards will cause ciliary muscle adjustment lags. Failure to perform refractive compensation in a timely manner can easily lead to high-frequency adjustment fatigue of the user's ciliary muscles. Summary of the invention

[0004] In order to overcome the above-mentioned defects of the prior art, the present invention provides a refractive compensation method and device based on dual-modal perception, which establishes a head-mounted device body coordinate system, synchronously collects ciliary electromyographic signals and dual-modal data such as head posture and gaze distance, and accurately determines emergency and stable adjustment events based on multi-dimensional characteristics of distance change rate, head movement gradient, trajectory curvature and electromyographic slope. Dynamically shorten the adjustment delay for emergency scenes and calculate the target refractive power in advance; baseline correction and periodic compensation are used for stable scenes to form a hierarchical decision-making mechanism. This scheme realizes the real-time synchronization of lens refractive compensation and ciliary muscle adjustment rhythm, effectively reduces invalid adjustment actions in high-frequency switching, reduces abnormal fluctuations in refractive power, significantly improves physiological adaptability and compensation accuracy in dynamic visual scenes, relieves ciliary muscle fatigue from the source, and provides innovative solutions for myopia prevention and control.

[0005] To achieve the above object, the present invention provides the following technical solutions: A refractive compensation method based on dual-modal perception, comprising: Establishing a head mounted device body coordinate system O-XYZ, defining a first area and a second area in the coordinate system O-XYZ; the user switches gaze back and forth between the first area and the second area at different switching speeds, and collecting bimodal perception data during the user's gaze switching process; Based on the bimodal perception data during the user's gaze switching process, determine the emergency adjustment event and the stable adjustment event; if it is determined as an emergency adjustment event, trigger the lead compensation acceleration mechanism; if it is determined as a stable adjustment event, adopt the conventional compensation mechanism.

[0006] Further, the method for establishing the coordinate system O-XYZ of the head-mounted device body is as follows: Take the midpoint of the line connecting the user's two eye pupils as the origin O, the X-axis is along the direction of the line connecting the two eye pupils from left to right, the Y-axis points to the top of the head, and the Z-axis direction is defined as the direct front of the line of sight.

[0007] Further, the method for defining the first region and the second region in the coordinate system O-XYZ is as follows: The interval Z ∈ [a, b] is defined as the first region, and the interval Z ∈ [c, d] is defined as the second region, where a, b, c, and d are positive real numbers and a < b < c < d.

[0008] Further, the bimodality refers to the physiological signal modality and the spatial motion modality, and the bimodal perception data includes physiological signal data and spatial motion data; The physiological signal data refers to the ciliary muscle electrical signal; The spatial motion data includes the user's head pose data, the gaze distance, and the three-dimensional gaze point coordinates.

[0009] Further, the method for determining the emergency adjustment event and the stable adjustment event is as follows: Based on the head pose data and the gaze distance, calculate the distance change rate V D 、the head motion gradient G H and the three-dimensional trajectory curvature κ; calculate the change slope ke of the ciliary muscle electrical signal; Based on the distance change rate V D 、the head motion gradient G H 、the three-dimensional trajectory curvature κ and the change slope ke of the ciliary muscle electrical signal, determine the emergency adjustment event and the stable adjustment event.

[0010] Further, the method for calculating the distance change rate V D is as follows: Adopt a sliding time window and calculate the distance change rate V D ; Further, the calculation method of the three-dimensional trajectory curvature is as follows: Generate a global space curve from the three-dimensional gaze point coordinates; select the current gaze point from the global space curve and the previous several gaze points to form a local space curve, and calculate the curvature of the local space curve as the three-dimensional trajectory curvature

[0011] Further, for the method of determining emergency accommodation events and smooth accommodation events based on the distance change rate V D , the head movement gradient G H , the three-dimensional trajectory curvature κ, and the change slope ke of the ciliary muscle electrical signal are as follows: If and and and > k th , it is determined as an emergency accommodation event; otherwise, it is a smooth accommodation event. Among them, is a preset distance change rate threshold, is a preset head movement gradient threshold, is a preset trajectory curvature threshold, and k th is a preset change slope threshold.

[0012] Further, the lead compensation acceleration mechanism is as follows: dynamically shorten the accommodation delay parameter to obtain the shortened compensation delay time , and based on calculate the target refractive power in advance, and adjust the lens according to ; the accommodation delay parameter is synchronously obtained when collecting the bimodal perception data during the user's fixation switching process.

[0013] A refractive compensation device based on bimodal perception is used to implement the above-mentioned refractive compensation method based on bimodal perception. The device includes: Partition module: used to establish the body coordinate system O-XYZ of the head-mounted device, and define the first region and the second region in the coordinate system O-XYZ; Bimodal data acquisition module: The user performs round-trip fixation switching between the first region and the second region at different switching speeds, and collects the bimodal perception data during the user's fixation switching process; the bimodal refers to the physiological signal modality and the spatial motion modality; Accommodation event determination module: Based on the bimodal perception data during the user's fixation switching process, determine emergency accommodation events and smooth accommodation events; Refractive compensation module: If it is determined as an emergency accommodation event, trigger the lead compensation acceleration mechanism; if it is determined as a smooth accommodation event, adopt the conventional compensation mechanism.

[0014] Compared with the prior art, the beneficial effects of the present invention are: The present invention realizes the precise positioning of the user's gaze space by constructing a coordinate system of the head-mounted device body and dividing it into a first region and a second region. Combining the acquisition of bimodal perception data, it synchronously obtains ciliary muscle electrical signals and spatial motion data such as head posture and gaze distance, and characterizes the accommodation scenario from the dual dimensions of physiological response and behavioral characteristics. Based on multi-dimensional features, the rate of distance change, head movement gradient, three-dimensional trajectory curvature, and the change slope of ciliary muscle electrical signals are calculated to accurately determine emergency and steady accommodation events, enabling the compensation strategy to be dynamically adjusted according to the accommodation load. For the lead compensation acceleration mechanism for emergency accommodation events, by dynamically shortening the accommodation delay parameter and combining the predicted accommodation lag, the refractive power of the lens is adjusted in advance to offset the physiological delay of the ciliary muscle; the conventional compensation mechanism periodically calculates the refractive compensation amount and performs baseline correction, reducing the computational load while ensuring accuracy. The combination of the two forms a hierarchical decision-making mechanism covering the full range of gaze switching scenarios, realizing the real-time synchronization of lens refractive compensation and ciliary muscle accommodation rhythm, effectively reducing ineffective accommodation actions caused by delay, reducing abnormal fluctuations in refractive power, alleviating high-frequency accommodation fatigue of the ciliary muscle, and significantly improving the refractive compensation accuracy and human physiological adaptability in dynamic visual scenarios, fundamentally solving the core problem of the disconnection between the compensation strategy and the real physiological response in traditional solutions.

[0015] The technical solution of the present invention can be widely adapted to the dynamic refractive compensation requirements in the field of light field display (such as holographic display, naked-eye 3D, stereoscopic display, etc.). In the light field display scenario, the user's line of sight often switches frequently with the depth change of the virtual / real scenario. Through bimodal data fusion and hierarchical compensation strategy, the present invention can accurately match the dynamic refractive requirements of the light field display device, effectively improving the visual clarity and physiological comfort when virtual objects are integrated with the real environment, and providing key technical support for optimizing the human-computer interaction experience of the light field display device. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0017] Figure 1 It is a principle flowchart of a refractive compensation method based on bimodal perception in the present invention; Figure 2 It is a method flowchart for determining emergency accommodation events and steady accommodation events in a refractive compensation method based on bimodal perception of the present invention; Figure 3 It is a schematic diagram of a scenario with a rapid change in gaze distance in an embodiment of the present invention; Figure 4 Schematic diagram of the scenario where the viewing distance changes slowly in the embodiment of the present invention; Figure 5 Flowchart of the method for obtaining the individual initial compensation coefficient matrix A in a refractive compensation method based on bimodal perception of the present invention; Figure 6 Functional module diagram of a refractive compensation device based on bimodal perception in the present invention.

[0018] Explanation of reference numerals: 1. User, 2. Desk, 3. Blackboard. Detailed implementation manners

[0019] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0020] Embodiment 1 Please refer to Figure 1 As shown, this embodiment provides a refractive compensation method based on bimodal perception, including: Step S1000, establish a body coordinate system O-XYZ of the head-mounted device, and define a first area and a second area in the coordinate system O-XYZ; the user performs a reciprocating gaze switch between the first area and the second area at different switching speeds, and collect bimodal perception data during the user's gaze switch; Further, step S1000 includes: Step S1100, establish a body coordinate system O-XYZ of the head-mounted device, and define a first area and a second area in the coordinate system O-XYZ; Further, step S1100 includes: Step S1110, take the midpoint of the line connecting the user's two eyes' pupils as the origin O, and establish a body coordinate system O-XYZ of the head-mounted device. This coordinate system is a right-handed Cartesian coordinate system, the X-axis is along the line connecting the two eyes' pupils from left to right, the Y-axis points to the top of the head, and the Z-axis direction is defined as the straight-ahead direction of the line of sight; Step S1120, define a first area and a second area in the body coordinate system O-XYZ of the head-mounted device; the close-range interval Z∈[a, b] is defined as the first area, and the far-range interval Z∈[c, d] is defined as the second area, where a, b, c, and d are positive real numbers and a < b < c < d.

[0021] Specifically, the X-axis is along the direction of the line connecting the pupils of the two eyes, with the positive direction from left to right, parallel to the plane of the first area (such as a desk, etc.), and is used to represent the horizontal movement of the user's head from left to right; the Y-axis is perpendicular to the line connecting the pupils, with the positive direction pointing towards the top of the head, and is used to represent the vertical movement of the head up and down; the Z-axis is along the straight-ahead direction of the user's line of sight, perpendicular to the lens plane, and is used to represent the front-back distance of the fixation target (i.e., the depth-of-field direction). This coordinate system is calibrated in real time by a nine-axis inertial sensor (integrated at the bridge of the frame) to ensure that the coordinate system dynamically adjusts with the head posture and always remains consistent with the user's visual axis. The pupil midpoint is selected as the origin because the pupil position directly reflects the geometric center of the line of sight and can most accurately map the spatial position of the fixation target. The orthogonal design of the X / Y / Z axes conforms to ergonomics, decomposing the head movement into three independent dimensions: horizontal, vertical, and depth of field, which facilitates subsequent quantitative analysis of the fixation distance (Z-axis coordinate) and head rotation (X / Y-axis attitude angle). If a non-pupil origin (such as the geometric center of the device) is used, it may cause a deviation between the line-of-sight coordinates and the actual fixation point, affecting the accuracy of distance measurement.

[0022] In the coordinate system O-XYZ, two core functional areas are divided according to the Z-axis coordinate range: The first area: is defined as the Z-axis range [a, b]. Exemplarily, the values are [0.2m, 0.6m], corresponding to the typical distance range between a primary school student's desk and the eyes (covering near-distance fixation targets such as books and stationery); The second area: is defined as the Z-axis range [c, d]. Exemplarily, the values are [2m, 15m], covering the common distance range of the classroom blackboard (adapting to different classroom space sizes).

[0023] The interval parameters a, b, c, and d satisfy a < b < c < d. The depth-of-field data of the environment is measured in real time by a ToF (Time of Flight) sensor built into the glasses and dynamically calibrated in combination with historical statistical data to ensure that the area division conforms to the actual usage scenario. Using the Z-axis as the main basis for division is because the ciliary muscle adjustment load is mainly caused by changes in the fixation distance (the curvature of the lens needs to be changed when switching between far and near distances). The clear definition of the first area and the second area simplifies the complex three-dimensional space into an interval judgment in the depth-of-field direction, greatly reducing the computational complexity of subsequent algorithms. If the area is not divided, the system needs to process the fixation points in the entire spatial range in real time, resulting in waste of computing resources and a decrease in the recognition efficiency of adjustment events.

[0024] Step S1100 solves the core problem of "lack of spatial positioning reference" by constructing an ontology coordinate system and dividing functional regions, laying a geometric foundation for subsequent data collection and adjustment event analysis. Incorporating the user's head posture, fixation point coordinates, and lens refractive power adjustment direction into the same coordinate system enables the possibility of spatio-temporal alignment between physiological signals (such as ciliary muscle electrical signals) and spatial motion data (such as changes in fixation distance). For example, when it is detected that the fixation point rapidly moves from the first region (Z = 0.3 m) to the second region (Z = 5 m), the distance change ΔD = 4.7 m provided by the coordinate system can be directly used as a quantitative indicator of the adjustment load. Through the division of the Z-axis interval, near-distance (desk) and far-distance (blackboard) fixation targets are clearly distinguished, and the most frequent adjustment switching scenarios in primary school classrooms are specifically captured (according to educational scenario statistics, the average number of switches per single class period reaches 80 - 120 times). Without regional definition, the system cannot quickly identify effective adjustment events, resulting in invalid data interfering with the decision-making of the compensation strategy. The real-time calibration function of the coordinate system (based on inertial sensors) ensures that head movement does not affect the accuracy of regional division. Even when the student lowers their head to write (Y-axis posture change) or turns their head to communicate (X-axis rotation), the depth-of-field measurement on the Z-axis can still accurately reflect the actual distance of the fixation target. This robust design avoids the positioning error of traditional fixed coordinate systems when the head posture changes (such as misjudging the first region as a farther distance when the head is lowered).

[0025] In step S1200, based on the ontology coordinate system O-XYZ of the head-mounted device, the user performs round-trip fixation switching between the first region and the second region at different switching speeds, and collects bimodal perception data during the user's fixation switching; The bimodal refers to the physiological signal modality and the spatial motion modality, and the bimodal perception data includes physiological signal data and spatial motion data; the physiological signal data refers to ciliary muscle electrical signals; the spatial motion data includes the user's head posture data, fixation distance, and three-dimensional fixation point coordinates; a global spatial curve is generated from the three-dimensional fixation point coordinates; The acquisition method of the user's head posture data is: integrating a nine-axis inertial sensor at the bridge of the spectacle frame to collect head posture data; the head posture data includes the head pitch angle , yaw angle , roll angle , and , and the change rates of ; The acquisition method of the ciliary muscle electrical signals is: arranging an n1-channel dry electrode bioelectric sensor on the inner side of the temple to collect ciliary muscle electrical signals; Specifically, in step S1200, by integrating a multi-sensor module, spatial motion data reflecting the fixation behavior and electro-signal data of the ciliary muscle physiological response are synchronously collected to form a bimodal perception data set. In this embodiment, the physiological signal modality specifically refers to the ciliary muscle electro-signal (EMG), which is collected by an n1-channel dry electrode sensor arranged on the inner side of the temple. The dry electrode is made of silver / silver chloride material and can be attached to the periorbital skin without conductive gel, providing high comfort for long-term wearing. After the signal is band-pass filtered at 20 - 500 Hz (to remove low-frequency noise and high-frequency interference), the root mean square value (RMS) is extracted as a quantitative index of muscle activation level.

[0026] Spatial motion modality: It includes head pose data, fixation distance, and three-dimensional fixation point coordinates. The head pose data is collected by a nine-axis inertial sensor (integrating an accelerometer, a gyroscope, and a magnetometer) at the bridge of the frame nose, and outputs the head pitch angle θ (rotation angle around the X-axis), yaw angle ϕ (rotation angle around the Y-axis), roll angle ψ (rotation angle around the Z-axis), and the real-time change rate of each angle ( ), reflecting the direction and intensity of head movement. The fixation distance is measured in real time by the lens ToF sensor for the target distance in the line-of-sight direction. The three-dimensional fixation point coordinates combine the eye tracking module (an infrared camera integrated above the lens) and the body coordinate system to map the retinal imaging position to the O-XYZ coordinate system, generating a continuous sequence of fixation points, which is further fitted into a global space curve for analyzing the curvature and direction changes of the fixation trajectory.

[0027] Guide the user to perform N1 round-trip fixation switches (from the first area to the second area and back). Each single switch includes the following key steps: Trigger condition: When the fixation point stays in the first area for ≥100 ms and then first enters the second area and stays for ≥100 ms for the first time, it is determined as a valid switch; Data storage: The bimodal perception data corresponding to each switch (including θ, ϕ, ψ and their change rates, fixation distance sequence, EMG RMS sequence, three-dimensional fixation point coordinates) are synchronously stored according to the time stamp, forming a data set containing N1 samples for subsequent model training and event determination.

[0028] By synchronously collecting the change in fixation distance (spatial motion data) and the peak delay of the EMG signal (physiological data), the accurate characterization of the individual's "accommodation lag characteristic" is achieved. Combining the head pose data (θ, ϕ, ψ and their change rates) with the fixation point coordinates can distinguish between two behaviors of "active head-turning fixation" and "eye-rotation fixation". For example, when the head yaw angle ϕ changes rapidly, even if the fixation distance remains unchanged, there may be a coordinated movement of the ciliary muscle (due to the head movement causing fine adjustment of the line-of-sight direction), which needs to be included in the accommodation load calculation. If only the fixation distance is collected and the head movement is ignored, such hidden accommodation demands will be missed, resulting in insufficient compensation.

[0029] The establishment of the coordinate system provides a spatial positioning reference for data acquisition, while the spatio-temporal characteristics of the bimodal data feed back the dynamic calibration of the coordinate system, forming a closed-loop support; when the head pose data (θ, ϕ, ψ) shows that the user bows their head by 15°, the coordinate system automatically adjusts the direction of the Z-axis (tilts with the line of sight), ensuring that the Z-axis interval of the first region is still based on the actual line-of-sight distance, rather than a fixed vertical distance, to avoid misjudgment of the region caused by changes in head pose; through the correlation analysis of the fixation point coordinates and the electromyogram signals, it can be identified whether an abnormal accommodation load is triggered in a specific spatial region (such as the upper left corner of the first region), providing a basis for subsequent personalized regional compensation. This coordination can still maintain a precise perception of the accommodation demand in a complex classroom environment (such as non-standard postures like students tilting their heads to write or looking up at the projection), fundamentally solving the defect of the traditional solution of "only relying on a fixed distance threshold and ignoring individual posture differences", and improving the spatio-temporal accuracy of dynamic refractive compensation to the same frequency level as the physiological signal response.

[0030] Step S2000: Based on the bimodal perception data during the user's fixation switching, determine the emergency accommodation event and the smooth accommodation event; if it is determined to be an emergency accommodation event, trigger the early compensation acceleration mechanism; if it is determined to be a smooth accommodation event, adopt the conventional compensation mechanism.

[0031] Furthermore, step S2000 includes: Step S2100: Based on the bimodal perception data during the user's fixation switching, determine the emergency accommodation event and the smooth accommodation event; Furthermore, as Figure 2 shown, step S2100 includes: Step S2110: Based on the head pose data and the fixation distance, calculate the distance change rate V D , the head movement gradient G H and the three-dimensional trajectory curvature κ; Furthermore, step S2110 includes: Step S2111: Adopt a sliding time window and calculate the distance change rate V D based on the fixation distance; Specifically, use a sliding time window to calculate, the length of the time window is , the step size is , and the calculation formula is: where is the fixation distance at the current moment , is the fixation distance at the previous sampling moment, is the sampling time interval, . The distance change rate It reflects the speed at which the user switches the gaze target between the near distance (first area) and the far distance (second area). The larger the value, the faster the switching. Figures 3 - 4 As shown, Figure 3 In the previous At the sampling moment, the eyes look at the desk, that is, the first area where the target is close to the eye. When the user looks at the blackboard, that is, the target is the second area at a long distance; the calculated distance change rate at this time is V D1 . Figure 4 In the previous At the sampling moment, the eyes look at the desk, and the first area of ​​the gaze target is close to the previous moment. When the user's gaze target is still the first area at a short distance, the calculated distance change rate is V D2 ; Figure 3 V in application scenarios D1 The larger the value, the delayed physiological response of the ciliary muscle to complete the adjustment may cause a deviation between the instantaneous refractive state and the actual demand. The use of sliding time windows avoids noise interference from single-time data. The trend analysis of each sampling point can accurately capture the transient characteristics of distance changes. If the speed is calculated directly by the difference between adjacent moments, it is easily affected by sensor noise and leads to misjudgment. However, the average processing of the fixed window can smooth high-frequency fluctuations and retain the real trend.

[0032] Step S2112, based on the head posture data , calculate the head motion gradient G H ; Specifically, , convert the movement speed in three directions into a comprehensive gradient value. Head movement gradient It comprehensively reflects the intensity of the head's movement in three directions. The larger the value, the more intense the head movement. The head movement gradient reflects the coordinated load of ciliary muscle adjustment. When the head turns quickly (such as when the head suddenly looks up at the blackboard accompanied by neck movement), the extraocular muscles need to adjust the direction of sight synchronously, which increases the adjustment pressure of the ciliary muscles. Traditional solutions only focus on distance changes and ignore such coordinated movements, resulting in compensation strategies that do not fully match the actual physiological load. The introduction of fills this gap.

[0033] Step S2113, select the current gaze point from the global space curve With the previous fixation point The local space curve is constructed, and the curvature of the local space curve is calculated as the curvature of the three-dimensional trajectory. ; Specifically, ,in, is the parametric equation of the fixation point, and are the first derivative (tangent vector) and the second derivative (curvature vector), respectively.

[0034] 3D trajectory curvature reflects the curvature of the gaze trajectory. The larger the value, the more curved the trajectory is, that is, the more drastic the line of sight shift is. The calculation is based on continuous (For example, M=5) three-dimensional gaze point coordinates rely on the real-time positioning of the gaze target by the ToF sensor and the coordinate mapping of the eye tracking module to ensure the spatiotemporal alignment of continuous gaze points. If only a single gaze point coordinate is used, the dynamic changes of the trajectory cannot be captured, while the curvature calculation effectively identifies "non-linear" gaze transfers through differential geometry methods. Such scenes are often accompanied by sudden adjustment needs of the ciliary muscles.

[0035] Step S2120, calculating the change slope ke of the ciliary muscle electroencephalogram signal; Specifically, after collecting the ciliary muscle electroencephalogram signal through the n1 channel dry electrode sensor on the inner side of the temple, it is first band-pass filtered (e.g., 5-50 Hz) to remove power frequency noise and high-frequency interference, in order to filter out power frequency noise (such as 50 Hz AC interference) and low-frequency drift (such as skin electrode contact noise), and then the sliding window method is used to calculate the root mean square value (RMS) of the signal to form a time series E(t) that represents the degree of muscle activation.

[0036] Exemplarily, the calculation method of the change slope ke is: at the event triggering time T when the gaze distance change is detected start Then select T start To T start The +300ms time window is used as the analysis interval. First, determine the signal baseline value E baseline :T start The average value of E(t) in the first 100ms is used as the reference value when the ciliary muscle is not activated. Then the signal extreme value E is found in the window. peak and its appearance time T peak (i.e., the maximum or minimum point of E(t)). The slope of change ke is defined as the linear rate of change of the signal from the baseline value rising (or falling) to the extreme value. For example, when a student suddenly looks up at the blackboard, the ciliary muscle needs to relax quickly. At this time, the absolute value of the slope of the RMS value of the electromyographic signal can also be used as a basis for judgment to ensure the responsiveness evaluation of the relaxation process. The slope reflects the response speed of the ciliary muscle during the adjustment process. If the signal does not have a valid extreme value in the window, it is judged as invalid data and is not included in the subsequent judgment.

[0037] Steps S2110 and S2120 solve the core problem of single judgment basis for accommodation events through multi-dimensional feature extraction. 、 、 、ke respectively depict the accommodation scenario from four dimensions of "distance change", "head movement", "gaze trajectory", and "physiological response", forming a three-dimensional judgment model. Quantifying the speed of distance change is the core driving force for triggering accommodation demand; Capturing the co-load brought by head movement solves the problem that traditional solutions ignore the influence of neck movement; κ Identifying the dynamic characteristics of the gaze trajectory covers high-frequency scenarios of non-linear gaze transfer (such as intensive accommodation when scanning text and pictures). ke verifies the accommodation state from the internal physiological level to ensure that the judgment result fits the individual muscle response ability. Through and κ combination, the intensity of accommodation load can be predicted: for example, high combined with high κ indicates that large and frequent lens accommodation is required in a short time, and the risk of lag is extremely high; through and the association of ke, the efficiency of physiological accommodation can be evaluated: high but low indicates that head movement exacerbates muscle response delay and stronger anticipatory compensation is required.

[0038] Step S2130 determines emergency accommodation events and stable accommodation events based on the distance change rate V D 、the head movement gradient G H 、the three-dimensional trajectory curvature κ and the change slope ke of the ciliary muscle electrical signal; If and and and >k th , it is determined as an emergency accommodation event, otherwise, it is a stable accommodation event; where is the preset distance change rate threshold for judging whether the distance change rate reaches the "fast" standard. is the preset head movement gradient threshold for judging whether the head movement is "significant". is the preset trajectory curvature threshold for judging whether the trajectory is "sharp turning", k th is the preset change slope threshold.

[0039] Specifically, in step S2120, by establishing a multi-feature joint determination model, accurate classification of emergency adjustment events and stable adjustment events is achieved. The setting of the three types of thresholds (V1, G1, K1) follows the principle of "physiological response boundary constraint + scene feature statistics + individual difference adaptation". Threshold of distance change rate The setting is based on the physiological delay characteristics of ciliary muscle accommodation. When the change rate of the fixation distance exceeds the maximum response ability of the ciliary muscle, the instantaneous refractive error will accumulate significantly. By collecting the peak delay time of the ciliary muscle electrical signal when different users switch their gazes, the critical value of the distance change rate is determined - this value corresponds to the boundary condition of "the ciliary muscle cannot complete accommodation within the effective time, resulting in the lens compensation requirement being significantly ahead of the physiological response". Exemplarily, in a classroom scenario, when a student quickly looks from the first area to the second area (3m), if the switching time is shorter than the average accommodation time of the ciliary muscle (about 150 ms), then the distance change rate . Considering individual differences and sensor noise, the actual threshold is set slightly lower than this theoretical value to ensure coverage of the above rapid adjustment scenarios.

[0040] The head movement gradient reflects the comprehensive movement intensity of the head in the three directions of pitch, yaw, and roll. When the head rotates rapidly, the extraocular muscles need to synchronously adjust the line of sight direction, causing the ciliary muscle to bear additional co-regulation load. By analyzing the correlation between the head posture change rate and the rising edge slope of the ciliary muscle electrical signal, G1 is determined as the critical value of "the head movement begins to significantly affect the ciliary muscle accommodation efficiency" - that is, when G H > G1, the response speed of the electromyogram signal decreases by more than 15%, indicating that acceleration compensation is required to offset the delay. Exemplarily, when a student suddenly turns his head to look at the blackboard on the side of the classroom, if the pitch angle change rate is 40° / s, the yaw angle change rate is 50° / s, and the roll angle change rate is 20° / s, and the combined movement gradient exceeds the normal value, it is determined as a significant head movement at this time, triggering the candidate condition for an emergency event.

[0041] The three-dimensional trajectory curvature quantifies the intensity of the line-of-sight transfer. When there is a sharp turn in the line-of-sight trajectory (such as quickly scanning from the corner of the desk to the edge of the blackboard), the ciliary muscle needs to complete asymmetric accommodation within a short time. At this time, the curvature value exceeds the critical value K1, indicating that the visual system's demand for clarity suddenly increases, and the traditional compensation algorithm is prone to cause blurred vision due to response delay. The threshold is determined by fitting the curvature distribution of a large number of real fixation trajectories, and the value that can identify 90% of the sharp line-of-sight transfer scenarios is taken. Exemplarily, when a student is reading the text and pictures in a textbook, if the line of sight quickly switches direction within a distance of 0.3m, and the formed local trajectory curvature exceeds K1, it indicates that there is a high-frequency and small-range accommodation demand, and the corresponding compensation mechanism needs to be triggered.

[0042] By analyzing the electromyogram signals in different adjustment scenarios, it is found that when is lower than a certain critical value, the response ability of the ciliary muscle to rapid distance changes significantly decreases, resulting in an extended adjustment lag time; k th is defined as the minimum slope value of "the electromyogram signal response speed is sufficient to support the conventional adjustment requirements". Exemplarily, by collecting the ke data of more than 200 adolescent users at different fixation switching speeds, the distribution characteristics of ke in the emergency scenario are fitted, and the 75th percentile is taken as k th to ensure coverage of 75% of the efficient adjustment response scenarios.

[0043] According to the distance change rate V D and the head movement gradient G H and the three-dimensional trajectory curvature κ and the change slope ke of the ciliary muscle electromyogram signal, the determination model forms a dual verification mechanism of "behavioral characteristics + physiological indicators". V D and G H and κ identify high-load adjustment scenarios from the perspective of fixation behavior, and ke confirms whether there is a risk of decreased adjustment efficiency from the perspective of muscle response. Only when the behavioral characteristics indicate high load ( and and ) and the physiological indicators show sufficient muscle response ( > k th ), it is determined as an emergency adjustment event, otherwise it is a stable adjustment event, avoiding overcompensation caused by misjudgment of simple behavioral characteristics (such as not triggering unnecessary acceleration compensation when switching at high speed but with good muscle response). The introduction of k th fills the gap in the traditional determination model of "only relying on external behavioral data and ignoring individual muscle differences". For example, users with higher myopia degrees may have lower due to long-term tension of the ciliary muscle. Even if the behavioral characteristics meet the standards, if ≤ k th , it is still determined as a stable event and the compensation time constant is extended, avoiding blurred vision caused by premature compensation due to insufficient muscle response.

[0044] Step S2100 solves the problem of inaccurate recognition of adjustment events through multi-modal feature fusion and joint determination; the distance change rate directly reflects the core driving force of ciliary muscle adjustment, the speed of far and near distance switching, and is the most direct factor triggering adjustment requirements; the head movement gradient supplements the additional load brought by head movement, identifies the co-regulation requirements caused by neck movements, and avoids missed judgments caused by only focusing on distance changes in traditional solutions; the trajectory curvature κCapture the dynamic characteristics of saccade, identify the sudden accommodation demands in non-linear fixation scenarios (such as rapid saccades). Although the distance change is small in such scenarios, the instantaneous response requirement of the ciliary muscle is extremely high. Ke verifies the accommodation state from the intrinsic physiological level to ensure that the judgment result conforms to the individual muscle response ability. The four combined judgments form a three-dimensional judgment model of "driving force - collaborative load - dynamic trajectory - physiological response", which effectively improves the recognition accuracy of accommodation events compared with single-feature judgment.

[0045] Step S2200, if it is determined to be an emergency accommodation event, trigger the lead compensation acceleration mechanism; Specifically, when it is determined to be an emergency accommodation event in step S2130, trigger the lead compensation acceleration mechanism, which specifically includes the following sub-steps: First, based on the real-time distance change rate V in the bimodal perception data D , head movement gradient G H , three-dimensional trajectory curvature κ, and the change slope ke of the ciliary muscle electrical signal, identify that the current fixation switch has the characteristics of rapid distance change, significant head movement, sharp trajectory turning, and severe fluctuation of the ciliary muscle electrical signal. In such scenarios, the accommodation load of the ciliary muscle suddenly increases, and the traditional fixed-delay compensation strategy is prone to cause serious lag in refractive compensation.

[0046] To solve this problem, the lead compensation acceleration mechanism first shortens the response cycle of the lens refractive compensation to match the dynamic characteristics of the emergency event. For example, it increases the data processing frequency to twice the original sampling frequency (such as from 100Hz to 200Hz) to reduce the signal processing delay. Secondly, introduce a feed-forward prediction model based on the change slope ke of the ciliary muscle electrical signal: by analyzing the correlation between ke and the accommodation lag in historical data, establish a real-time mapping function. When exceeds the preset threshold k th , estimate the change trend of the accommodation demand within the next 100ms according to the current value of ke, and adjust the lens refractive power in advance. For example, if the current ciliary muscle electrical signal rises at a slope of 50μV per second (significantly higher than the average slope of 10μV / s in the stable scenario), the prediction model determines that the ciliary muscle is rapidly contracting to cope with near fixation, and then applies a refractive compensation increment of +0.5D in advance, responding 200ms earlier than the traditional delayed compensation.

[0047] In addition, for the case where the head movement gradient G H is relatively high, the compensation mechanism synchronously introduces a head movement compensation coefficient: this coefficient is dynamically adjusted according to the real-time value of G H . When G H exceeds G1, superimpose an adjustment correction amount related to the head movement direction and speed on the original compensation algorithm. For example, when the head quickly tilts up (accompanied by a far fixation switch to the blackboard), according to the yaw rate and the pitch angle change rate The vector synthesis result, an additional 0.3D hyperopic compensation is added to offset the synergistic load of extraocular muscle movement on ciliary muscle accommodation.

[0048] The lead compensation acceleration mechanism accurately captures the sudden demands of the ciliary muscle during emergency accommodation events through the deep fusion of bimodal data, solving the problem of compensation lag caused by fixed delays in traditional solutions. Exemplarily, when a student quickly raises their head from a desk (0.3m) to look at the blackboard (5m), and the pitch angle change rate of the head reaches 30° / s (exceeding the threshold of 20° / s), the lead compensation acceleration mechanism can complete the refractive power adjustment in advance, in a shorter time than traditional solutions, effectively reducing the instantaneous refractive error caused by accommodation lag, avoiding the accumulation of ineffective accommodation actions, and alleviating the excessive fatigue of the ciliary muscle in high-frequency switching scenarios.

[0049] Step S2300, if it is determined to be a stable accommodation event, then a conventional compensation mechanism is adopted.

[0050] Specifically, the conventional compensation mechanism constructs a steady-state compensation model based on the historically statistically average accommodation delay: by using a sliding time window (such as a 1s window length) to smooth the fixation distance, head pose data, and ciliary muscle electrical signals, and calculating the mean value of the data within the window as the input parameter of the current accommodation state. For example, when the fixation distance slowly switches between 0.5m (the first region) and 3m (the second region), and the distance change rate V D is 0.8m / s (below the threshold of 1.5m / s), the system uses the average fixation distance (2.0m) of 20 consecutive sampling points (sampling interval 50ms) within the window as the basis for calculating the target refractive power, avoiding the interference of a single noise point on the compensation strategy.

[0051] Secondly, the conventional mechanism introduces a baseline drift correction module for ciliary muscle electrical signals: Since the change slope of ciliary muscle electrical signals in a stable scenario is small and is easily affected by noise such as skin impedance fluctuations, by calculating the difference between the current signal and the baseline signal n2 seconds ago, the effective component that truly reflects the accommodation state is extracted to ensure that the refractive compensation amount matches the actual physiological demand. For example, when the baseline signal is 20μV and the current signal is 25μV, it is determined that the ciliary muscle is in a slightly contracted state, corresponding to a 0.75D myopic compensation, rather than misjudging due to directly using the original signal (such as misjudging as a 1.5D compensation corresponding to 30μV under noise interference).

[0052] The conventional compensation mechanism targets the physiological characteristics of stable adjustment scenarios. Through data smoothing and baseline correction, it reduces the system's computational complexity while ensuring compensation accuracy. Exemplarily, when a student slowly moves their line of sight in the first region (such as a slight head rotation when reading a textbook), the conventional compensation mechanism can control the data processing delay within 150 ms. At the same time, through baseline correction, it reduces the noise interference of the ciliary muscle electrical signal by 60%, avoiding frequent adjustment of refractive power caused by overresponding to low-frequency signal fluctuations (such as reducing the ineffective adjustments from 5 times per minute in traditional solutions to 1 time), and improving visual stability and wearing comfort in dynamic visual scenarios. In addition, the conventional compensation mechanism complements the lead compensation mechanism, covering the full range of gaze switching scenarios from low-frequency stable to high-frequency urgent, ensuring that the lens refractive compensation strategy can synchronize with the ciliary muscle adjustment rhythm under different loads, and overall reducing the abnormal fluctuation amplitude of refractive power, achieving a balance between physiological adaptability and compensation accuracy.

[0053] The emergency event determination result directly drives the lead compensation acceleration mechanism to offset the physiological delay by pre-adjusting the lens refractive power, solving the problem of instantaneous error amplification during high-speed switching; the stable event determination result triggers the conventional compensation mechanism to ensure visual comfort through progressive adjustment, avoiding user discomfort caused by frequent rapid adjustments. This hierarchical decision-making mechanism maximizes the fusion value of bimodal perception data (physiological signals and spatial motion data), avoiding the misjudgment risk of a single modality (such as relying only on distance changes), and achieving "on-demand compensation" through dynamic parameter adjustment, fundamentally solving the core problem of the out-of-sync compensation strategy and physiological response in existing designs, providing key technical support for the precise adaptation of myopia control lenses in real dynamic visual scenarios.

[0054] Embodiment 2 Based on Embodiment 1, this embodiment provides a refractive compensation method based on bimodal perception, including: Step S1000, establish a body coordinate system O-XYZ of the head-mounted device, and define a first region and a second region in the coordinate system O-XYZ; the user performs round-trip gaze switching between the first region and the second region at different switching speeds, and collect bimodal perception data during the user's gaze switching process to obtain an individual initial compensation coefficient matrix A; Further, step S1000 includes: Step S1100, establish a body coordinate system O-XYZ of the head-mounted device, and define a first region and a second region in the coordinate system O-XYZ; The method of establishing the body coordinate system O-XYZ of the head-mounted device and defining a first region and a second region in the coordinate system O-XYZ is the same as that in Embodiment 1.

[0055] Step S1200: Based on the body coordinate system O-XYZ of the head-mounted device, perform personalized parameter initialization to obtain the individual initial compensation coefficient matrix A; Further, step S1200 includes: Step S1210: Based on the body coordinate system O-XYZ of the head-mounted device, require the user to perform N1 round-trip fixation switches between the first area and the second area at different switching speeds, with each speed repeated m times; Step S1220: Synchronously collect the bimodal perception data and accommodation delay parameters during the user's fixation switch ; The bimodal refers to the physiological signal modality and the spatial motion modality. The bimodal perception data includes physiological signal data and spatial motion data; the physiological signal data refers to ciliary muscle electrical signals; the spatial motion data includes the user's head pose data, fixation distance, and three-dimensional fixation point coordinates; a global spatial curve is generated from the three-dimensional fixation point coordinates; The acquisition method of the user's head pose data is: integrate a nine-axis inertial sensor at the bridge of the frame to collect the head pose data; the head pose data includes the head pitch angle , yaw angle , roll angle , and , and the change rates of ; The acquisition method of the ciliary muscle electrical signals is: arrange n1-channel dry electrode bioelectric sensors on the inner side of the temple to collect the ciliary muscle electrical signals; Specifically, the user performs a round-trip gaze switch between the first area and the second area at different switching speeds. For example, in a typical scenario where students frequently switch between a close-range desk (about 0.3-0.5m) and a distant blackboard (about 2-10m), multiple switching speeds are set, such as low speed (gaze distance change rate ≤ 0.5m / s), medium speed (0.5m / s<change rate<1.5m / s), and high speed (change rate ≥1.5m / s), covering different behavioral patterns from slow head raising to fast scanning in daily learning. Each speed level is repeated m times (such as m=10) to reduce the interference of individual accidental actions on data collection and ensure that the sample is statistically significant. In the coordinate system of the head-mounted device, the user completes the round-trip gaze from the first area to the second area according to the preset path, that is, from a close-range target (such as text on a textbook) to a distant target (such as a blackboard on a blackboard), and then returns to the close-range target, forming a complete gaze cycle. The head posture and gaze point coordinates are monitored in real time through the nine-axis inertial sensor and ToF sensor. Through repeated tests at multiple speeds and times, a data set containing individual dynamic adjustment characteristics is constructed, providing a basis for the subsequent establishment of a mapping relationship between "gaze speed-physiological response-compensation needs". Without switching tests at different speeds, the compensation strategy will not be able to distinguish between fast glances and slow adjustment scenes, resulting in the accumulation of compensation delays during high-frequency switching, which increases the ciliary muscle adjustment load.

[0056] An n1-channel dry electrode bioelectric sensor is arranged on the inner side of the temple (e.g. n1=4, symmetrically distributed on the left and right temples), and the surface electrode technology is used to non-invasively collect ciliary muscle electrical signals. The dry electrode does not require conductive gel, which is suitable for children to wear and reduces skin irritation. The original signal is band-pass filtered (10-500Hz) to remove power frequency interference (such as 50Hz mains noise) and DC drift, and then baseline correction is used to eliminate long-term trends to ensure the accuracy of subsequent feature extraction.

[0057] The nine-axis inertial sensor (including three-axis accelerometer, three-axis gyroscope, and three-axis magnetometer) integrated at the bridge of the nose collects the head pitch angle θ (rotation around the X-axis, lowering the head is positive), yaw angle ϕ (rotation around the Y-axis, turning left is positive), roll angle ψ (rotation around the Z-axis, tilting right is positive), and the rate of change of each angle ( ), reflecting the direction and speed of head movement. The distance of the gaze target is measured in real time by the ToF sensor, and the three-dimensional gaze point coordinates are determined in combination with the head-mounted device coordinate system. Continuous gaze points are connected to generate a global space curve for subsequent trajectory curvature calculation.

[0058] Single adjustment delay time τ i', i'=1,2,...,N1: In each gaze switching process, the time difference from the moment when the gaze distance starts to change (i.e., the moment when the Z axis coordinate is detected to exceed the current area threshold, such as the moment when the change from the first area Z≤0.6m to the second area Z≥2m) to the time when the ciliary muscle electrical signal reaches the peak value. This time difference reflects the physiological delay of the ciliary muscle from receiving the visual distance change signal to completing the contraction adjustment.

[0059] Adjusting Delay Parameters Determination of: τ for N1 round-trip gaze switching i' (i'=1,2,…,N1) take the arithmetic mean and get , Characterizes the average time lag of the ciliary muscle accommodation response relative to lens compensation under regular gaze switching, and is the initial benchmark for the timing parameters in subsequent compensation strategies.

[0060] By synchronously collecting dual-modal data, the association of "spatial motion trajectory → physiological regulation signal → delay parameter" is established to avoid the disconnection between compensation strategy and physiological response caused by traditional solutions that rely only on single-modal data (such as only gaze distance). Adjusting delay parameters As an individual basic physiological characteristic, it is used to correct the delay time of the advance compensation acceleration mechanism in step S2300 to ensure that the compensation timing is synchronized with the ciliary muscle adjustment rhythm.

[0061] If the dual-modal data collection of this step is missing, the subsequent individual initial compensation coefficient matrix A will lack the coupling information of physiological signals and spatial motion, resulting in the inability of the adjustment lag prediction model to accurately map the relationship between "gaze trajectory changes-ciliary electromyography characteristics-compensation needs", which ultimately causes the compensation strategy to be out of sync with the actual physiological response and cannot solve the problem of invalid adjustment load in high-frequency switching scenarios. Through the synchronous collection of dual-modal data in this step, the individual physiological characteristics (ciliary electromyography signals) and spatial motion behaviors (head posture, gaze trajectory) are quantitatively associated, providing an indispensable data source for the precision of subsequent dynamic compensation strategies, which is a key technical link in achieving "physiologically adaptive refractive compensation".

[0062] Step S1230, obtaining an individual initial compensation coefficient matrix A based on the bimodal perception data during the user's gaze switching process.

[0063] Furthermore, if Figure 5 As shown, step S1230 includes: Step S1231, performing spatiotemporal alignment on the user's head posture data, ciliary muscle electromyography signal, and gaze distance; Convert the user's head posture data, ciliary muscle electroencephalogram and gaze distance to the head-mounted device's main coordinate system O-XYZ; Step S1232, extracting features from the ciliary electromyography signals after time-space alignment to obtain electromyography response features reflecting the contraction activity of the user's ciliary muscles; the electromyography response features include electromyography energy E EMG ; Step S1233, based on the head posture data, gaze distance and electromyographic response characteristics after time-space alignment, the refractive compensation coefficient is calculated through a multivariate regression model to form an individual initial compensation coefficient matrix A.

[0064] Specifically, step S1230 aims to construct an individual initial compensation coefficient matrix A by processing and modeling the bimodal perception data, establish a quantitative mapping relationship of "physiological signal-spatial position-compensation demand", and solve the technical problem of the disconnection between the compensation strategy and the actual physiological response in the existing design.

[0065] In step S1231, the midpoint of the pupils of both eyes is taken as the origin, the X-axis is horizontally to the right along the line connecting the pupils, the Y-axis is vertically pointing to the top of the head, and the Z-axis is along the front of the line of sight (vertical to the lens plane), ensuring that all spatial motion data (such as the coordinates of the gaze point) are strictly aligned with the physical position of the device. The output data of the nine-axis inertial sensor (collecting head posture data), the dry electrode sensor (collecting electromyographic signals) and the ToF sensor (collecting gaze distance) are timestamped, and the asynchronous data of different sensors are unified to a common sampling frequency using linear interpolation to eliminate the timing deviation caused by differences in sampling rates. If spatiotemporal alignment is not performed, the spatial reference system and time base of different modal data are inconsistent, which will lead to errors in subsequent feature association. For example, if the change in the head pitch angle θ is not associated with the spatial position of the gaze distance in the same coordinate system, the adjustment load when looking down at the desk may be misjudged, causing the compensation strategy to deviate from actual needs.

[0066] In step S1232, the ciliary electromyography signals after time-space alignment are preprocessed and feature calculated, including bandpass filtering (10-500Hz), using Butterworth filter to remove power frequency interference (such as 50Hz mains noise) and DC drift, retaining the effective electrical signal generated by ciliary muscle contraction (typical frequency range 10-500Hz); baseline correction: remove signal baseline drift by polynomial fitting, and use the resting state mean value 2 seconds before signal acquisition as the baseline reference to ensure the accuracy of subsequent energy calculation. The root mean square (RMS) algorithm is used to calculate the electromyography energy value of the preprocessed signal within a sliding time window (such as 50ms). The ciliary electromyography signal is a direct representation of the physiological state of regulation, but the original signal is susceptible to noise interference. Through filtering, baseline correction and RMS calculation, the physiological signal is converted into a quantifiable regulation load index, providing a physiological driving basis for the subsequent compensation strategy. For example, when the user quickly switches from the second area to the first area, the EEMG increases significantly, indicating that the ciliary muscle needs to contract quickly to adapt to close-range gaze, and the refractive compensation force needs to be increased to offset the accommodation lag.

[0067] In step S1233, the preprocessed bimodal data is grouped by multi-dimensional features, and a linear regression model is fitted by the least squares method to generate an individual-specific refractive compensation coefficient matrix A. The specific steps are as follows: The times of round-trip fixation switching data executed by the user in step S1210 are grouped according to the following dimensions: Head pose grouping: Using (0° - 180°) and (0° - 360°) discretization intervals (such as each each as an interval) to divide subsets; Fixation distance grouping: Discretized according to 0.1m intervals for the first region and the second region, including typical fixation distances from the first region at close range to the second region at long range; Electromyogram energy stratification: The E of all samples EMG is divided into low, medium, and high three layers according to the 25% and 75% quantiles, reflecting different levels of ciliary muscle regulation load.

[0068] For the data within each group, the least squares method is used to fit a linear regression model. With the refractive power compensation value as the dependent variable and E EMG as the independent variable, a linear equation is fitted by the least squares method to obtain the refractive compensation coefficient , indicating the refractive power compensation value of the lens corresponding to the change in unit electromyogram energy (i.e., the refractive power compensation corresponding to each microvolt of electromyogram signal) under a specific head pose , fixation distance and ciliary muscle load . i is the discretization interval index of the head pitch angle θ , j is the discretization interval index of the head yaw angle ϕ , k is the discretization point index of the fixation distance, and l is the stratification index of the electromyogram energy E EMG .

[0069] The regression coefficients of all groups are stored according to to form an individual initial compensation coefficient matrix . Among them, is the number of discretization intervals of the pitch angle (such as 36, corresponding to , interval); is the number of discretization intervals of the yaw angle (such as 72, corresponding to interval); is the fixation distance The discretization points of (such as 150, corresponding to interval); is the EMG energy The number of layers of (such as 3, low / mid / high).

[0070] matrix Through regression analysis, in the head pose data , the fixation distance , the EMG energy and the refractive compensation coefficient are directly correlated to form a mapping relationship of "physiological signal - spatial position - compensation demand". For example, when the user fixates at a distance of 0.3m (close distance) with θ = 10° (looking down at the desk) and ϕ = 0° (looking straight ahead) and E EMG is at the high level, the corresponding a ijk,l in matrix A will indicate a greater refractive compensation force to offset the ciliary muscle accommodation lag. Since step S1210 requires the user to perform back-and-forth fixation switching at different speeds, matrix A can capture the individual's accommodation response differences during fast / slow switching, providing a personalized training data basis for the subsequent accommodation lag prediction model. The individual initial compensation coefficient matrix A not only contains the coupling relationship of bimodal perception data, but also explicitly expresses the quantitative relationship between the ciliary muscle electrical signal intensity and refractive compensation through regression modeling, providing the core benchmark data for the subsequent dynamic compensation strategy and solving the technical problem of "the compensation strategy is out of sync with the real physiological response".

[0071] Step S1230 unifies the spatio-temporal benchmarks of head pose, fixation position and EMG signals through spatio-temporal alignment, solving the problem of the disconnection between physiological signals and spatial motion data in traditional solutions. The modal coupling mechanism enables the compensation strategy to have the dual adjustment capabilities of "physiological signal driven + spatial position perception", significantly improving the adaptability in dynamic scenarios. The EMG energy feature extraction and multiple regression modeling convert the ciliary muscle accommodation load into computable compensation parameters. Traditional solutions rely on fixed delay parameters and cannot reflect the impact of EMG intensity changes on compensation demand; while this step quantifies EEMG through the RMS algorithm and establishes a linear mapping through the regression coefficient to dynamically adjust the compensation force according to the EMG load. As the mapping hub of "physiological signal - spatial position - compensation demand", matrix A connects the previous data acquisition with the subsequent accommodation lag prediction and compensation mechanism. Through multi-speed data acquisition and hierarchical modeling, matrix A can capture the special accommodation demands during fast switching. When the user travels back and forth between the first area and the second area at high speed, the corresponding group will trigger a greater , prompting the lens to adjust its refractive power in advance to offset the physiological delay during rapid switching. This refined grouping strategy upgrades the compensation system from "fixed-parameter response" to "dynamic feature matching", effectively reducing abnormal fluctuations in refractive power and alleviating high-frequency regulatory fatigue of the ciliary muscle.

[0072] Step S2000: Based on the bimodal perception data during the user's gaze switching process, determine emergency adjustment events and stable adjustment events; if an emergency adjustment event is determined, trigger the advanced compensation acceleration mechanism; if a stable adjustment event is determined, adopt the conventional compensation mechanism.

[0073] Furthermore, step S2000 includes: Step S2100: Based on the bimodal perception data during the user's gaze switching process, determine emergency adjustment events and stable adjustment events; Specifically, define the minimum interval ΔD between the first region and the second region th , when the gaze distance switches from the first region to the second region or vice versa, the actual change amplitude needs to exceed ΔD th to trigger event determination. Among them, is the current gaze distance, is the gaze distance at the previous moment of the current moment t. By statistically analyzing the average distance change rate of the user at different switching speeds in step S1210 (i.e., ), take the average rate in the high-frequency switching scenario (such as switching more than 2 times per second) as the distance change rate threshold (for example ). Calculate the myoelectric energy change rate , where, is the myoelectric energy at the current moment, is the myoelectric energy at the previous moment; reflects the instantaneous change intensity of the ciliary muscle regulation load, and the myoelectric energy change rate threshold is set to the average level of the myoelectric energy mutation of an individual during rapid switching (for example ).

[0074] Adopt a dual-condition joint determination mechanism: Emergency adjustment event determination: When the following two conditions are simultaneously met, it is determined as an emergency adjustment event: (1) The change amplitude of the gaze distance , and the change rate (indicating that the fixation point quickly switches between the first region and the second region, such as a student quickly looking up at the blackboard or looking down to take notes due to the teacher's explanation); (2) The myoelectric energy change rate (indicating that significant load changes occur in the ciliary muscle due to rapid accommodation, such as a sharp increase in the electromyogram signal within a short period, reflecting a sudden increase in the accommodation demand).

[0075] Determination of smooth accommodation events: If the above two conditions are not simultaneously met, it is determined as a smooth accommodation event, including the following scenarios: (1) The change amplitude of the fixation distance is less than (such as slightly moving the line of sight within the first area, the axial coordinate changes within the range of ); (2) The rate of distance change is lower than (such as slowly looking up at the blackboard, and the ciliary muscle has sufficient time to complete accommodation); (3) The rate of change of electromyogram energy is lower than (indicating that there is no significant mutation in the accommodation load of the ciliary muscle and it is in a conventional accommodation state).

[0076] This determination method solves the key problem of "how to identify the mutation scenario of ciliary muscle accommodation load in real time" through the joint analysis of bimodal data. The emergency accommodation event corresponds to the scenario with a relatively high risk of ciliary muscle accommodation lag (such as when quickly switching between near and far, the physiological reaction delay is likely to cause the accumulation of instantaneous refractive errors). By identifying such events in advance, the lead compensation acceleration mechanism in step S2300 can be triggered, shortening the compensation delay time , enabling the refractive power of the lens to match the target distance demand in advance and avoiding the accumulation of ineffective accommodation load. Without this determination step, the system will not be able to distinguish the urgency of accommodation events, which may lead to the use of a conventional compensation cycle during rapid switching, exacerbating the asynchrony between refractive compensation and physiological responses and unable to effectively relieve ciliary muscle fatigue. By combining spatial motion data (distance change amplitude / rate) with physiological signal data (rate of change of electromyogram energy), a mapping relationship of "scene feature - physiological response" is established, enabling the individual initial compensation coefficient matrix A (including compensation coefficients under different head postures, fixation distances, and electromyogram loads) collected in step S1000 to be dynamically associated with real-time accommodation events. For example, when it is determined as an emergency event, the system can preferentially call the compensation coefficients corresponding to the high electromyogram energy layer ( A = 3) and rapid distance change in the matrix l , improving the pertinence of the compensation strategy. The personalized thresholds ( , , ) set based on individual historical data can adapt to the differences in accommodation abilities of different users (such as the different reaction speeds of the ciliary muscles between children and adults), avoiding the deficiencies of traditional fixed-parameter algorithms in individual adaptability. For example, if a user shows an average rate of change of electromyogram energy of 40% during rapid switching in step S1210, then his It can be set to 35%, ensuring that it is triggered only when significantly exceeding its regular adjustment load, reducing the probability of misjudgment.

[0077] The coordinate system defined in step S1000 O − XYZ provides a spatial positioning reference for the calculation of the viewing distance, and the bimodal data collected at different switching speeds in step S1210 provides a basis for individual differences in threshold setting (such as 、 ). The determination result (emergency / steady event) is used as a prerequisite for the accommodation lag prediction model (see step S2200), enabling the model to call different compensation strategies (lead compensation or regular compensation) according to the event type, ensuring the dynamic matching of "physiological signal characteristics - spatial movement trajectory - compensation strategy", and ultimately achieving the synchronization of lens refractive compensation and ciliary muscle accommodation rhythm, solving the core defect of "asynchronous compensation strategy and real physiological response" in the original technology problem. Through the above determination method, high-risk adjustment scenarios can be identified in real time, providing a decision-making basis for the differential execution of subsequent compensation mechanisms, forming a complete closed loop from data collection, feature analysis to strategy response, and significantly improving the adaptation accuracy and physiological compatibility of myopia control lenses in dynamic visual scenarios.

[0078] Step S2200, based on the individual initial compensation coefficient matrix A, constructs an accommodation lag prediction model to obtain the predicted accommodation lag amount ΔF pred ; Furthermore, step S2200 includes: Step S2210, based on the individual initial compensation coefficient matrix A, three-dimensional trajectory curvature 、head movement gradient G H 、distance change rate V D and the change slope ke of the ciliary muscle electrical signal, constructs an accommodation lag prediction model; Step S2220, according to the constructed accommodation lag prediction model, obtains the predicted accommodation lag amount ΔF pred 。

[0079] Specifically, step S2200 aims to construct an accommodation lag prediction model, dynamically predict the ciliary muscle accommodation lag amount by fusing multi-dimensional perception data, so as to provide a real-time correction basis for the lens compensation strategy. This step solves the core problem that traditional compensation algorithms rely on fixed delay parameters resulting in error amplification, and realizes the synchronization of refractive compensation and physiological accommodation rhythm. The individual initial compensation coefficient matrix A, three-dimensional trajectory curvature 、head movement gradient G H 、distance change rate V DThe method for obtaining the change slope ke of the ciliary muscle electrical signal is the same as that in Embodiment 1. The accommodation lag prediction model adopts a temporal convolutional neural network (TCN) architecture based on the attention mechanism. The model input layer contains four parallel branches: (1) the vector after flattening the individual initial compensation coefficient matrix A, which characterizes the user's personalized accommodation characteristics; (2) the time series of the three-dimensional trajectory curvature κ, which reflects the mutation characteristics of the fixation trajectory; (3) the sliding window mean of the head movement gradient G H , which characterizes the intensity of head movement; (4) the difference sequence of the distance change rate V D , which captures the dynamic characteristics of fixation switching. After the data of each branch are time-aligned, the temporal features are respectively extracted through a one-dimensional convolutional layer. The attention mechanism module dynamically weights and fuses the features of each branch to highlight the dominant influencing factors at the current moment. The fully connected layer maps the fused features to the predicted accommodation lag ΔF pred . During training, the actual accommodation delay time τ i' collected in step S1200 is used as the supervision signal, and the training set and the test set are divided by time series cross-validation. The loss function is the Huber loss to balance the influence of outliers. During the model deployment phase, the slice of the A matrix, κ, G H , V D at the current moment and the historical time series window data are input in real time, and the predicted value of ΔF pred within the future Δt'' time window is output. The predicted accommodation lag ΔF pred is the instantaneous deviation between the actual accommodation force of the ciliary muscle and the target refractive demand. This quantity is obtained by inverse quantization of the normalized value output by the TCN model. During real-time prediction, the model updates ΔF pred every 50 ms to ensure the timeliness of the compensation strategy.

[0080] The traditional fixed delay parameter τ A cannot reflect the non-linear accommodation characteristics of individuals during rapid fixation switching. By introducing multi-dimensional dynamic feature modeling, the prediction accuracy of the lag quantity is significantly improved; the spatio-temporal correlation between head movement and fixation trajectory affects the accommodation lag pattern, and the temporal modeling ability of TCN effectively captures such complex correlations; furthermore, the individual initial compensation coefficient matrix A provides personalized prior knowledge for the model, avoiding the slow convergence problem caused by learning from scratch. The multi-speed switching data collected in step S1210 provides rich training samples for the TCN model, enabling the model to learn the non-linear mapping relationship between ΔF pred and V D at different speeds; the matrix A generated in step S1233 is used as the model input, embedding the offline calibrated personalized features into the real-time prediction process, improving the model convergence speed and generalization ability. This synergy realizes a fast response to sudden accommodation events while ensuring the baseline accuracy.

[0081] The absence of this step will lead to three technical obstacles: (1) The lead compensation acceleration mechanism in step S2300 lacks a quantitative correction basis and cannot dynamically adjust τ real , resulting in inaccurate compensation timing in case of emergencies; (2) The individual initial compensation coefficient matrix A generated in step S1233 only contains static correlation relationships and cannot adapt to the real-time changing adjustment requirements, leading to the disconnection between the compensation strategy and the dynamic physiological response; (3) The online update mechanism in step S2500 loses the benchmark model and cannot achieve continuous optimization of the compensation parameters. This step combines the personalized static parameters (matrix A) with the dynamic perception data (κ, G H , V D ) through the establishment of an interpretable prediction model, enabling the lens compensation system to have the ability of adaptive evolution.

[0082] When the user performs high-frequency gaze switching, the TCN model judges the severity of head rotation based on the real-time head motion gradient G H . When G H exceeds G1, the model automatically increases the weight coefficient of the three-dimensional trajectory curvature κ to promptly capture the sudden change in the accommodation demand caused by the rapid eye movement. For example, when a student quickly turns their head to look at the side blackboard, the increased κ value triggers the model to output a higher ΔF pred , driving the lens to increase the positive refractive compensation amount in advance to offset the accommodation lag caused by the rapid head rotation. At the same time, the personalized compensation coefficients stored in matrix A ensure that different users obtain differentiated compensation amounts adapted to their physiological characteristics when facing the same G H and κ values.

[0083] In step S2300, if it is determined to be an emergency accommodation event, the lead compensation acceleration mechanism is triggered; the lead compensation acceleration mechanism is to dynamically shorten the accommodation delay parameter , to obtain the shortened compensation delay time , and based on calculate the target refractive power , and adjust the lens according to .

[0084] The method for obtaining the shortened compensation delay time is: obtain the current moment's electromyogram energy , and shorten the accommodation delay parameter according to the current moment's electromyogram energy , to obtain the compensation delay time .

[0085] The method for calculating the target refractive power is: obtain the current moment's fixation distance and the previous moment's fixation distance , and according to , and the predicted adjustment lag ΔF pred Determine the target refractive power .

[0086] The method of adjusting the lens according to is as follows: According to Generate a motor angle control command, flip the entire lens, and achieve dynamic adjustment of the refractive power.

[0087] Specifically, the ciliary muscle electrical signal is collected in real time through the dry electrode bioelectric sensor arranged on the inner side of the temple. After band-pass filtering (10 - 500 Hz) and baseline correction processing of the signal, the root mean square (RMS) algorithm is used to calculate the myoelectric energy within the current time window . This energy value reflects the instantaneous contraction intensity of the ciliary muscle during an emergency adjustment event. The higher the energy, the greater the load borne by the ciliary muscle due to adjustment lag. According to the current myoelectric energy and the preset energy threshold The ratio of, adjust the delay parameter according to the following rules: If , then , where is the shortening factor is the maximum value of the individual's historical myoelectric energy.

[0088] If , then .

[0089] is a dimensionless proportionality coefficient used to map the excess part of the myoelectric energy to the shortening amount of the delay time. Its value range is , ensuring that the shortened delay time is always less than the original delay parameter , but will not be overly shortened to zero to avoid refractive power oscillation caused by premature triggering of compensation.

[0090] Lower limit : Ensure that when the myoelectric energy exceeds the threshold, the compensation delay time must be shortened, reflecting the necessity of "advance compensation".

[0091] Upper limit Prevent the myoelectric energy from reaching the maximum value When, is shortened to zero, resulting in the lens compensation being completely out of the physiological adjustment rhythm and causing overcompensation (such as dizziness caused by sudden changes in refractive power).

[0092] This step realizes the forward shift of the compensation timing by linearly correlating the high myoelectric energy with the shortening amount of the delay time. For example, when the user quickly switches from the second area to the first area, if it is detected significantly higher than the energy threshold, then shorten to trigger the refractive compensation of the lens in advance, so as to offset the accommodation lag caused by the high-speed switching of the ciliary muscle.

[0093] According to 、 and the predicted accommodation lag amount ΔF pred determine the target refractive power , the formula is ; where is used to judge the gaze switching direction (negative when far-to-near, positive when near-to-far). By introducing correct the ideal refractive power to offset the physiological delay of the ciliary muscle and make the refractive power of the lens match the target distance requirement in advance.

[0094] Integrate a dual-lens flipping module inside the frame, which includes two sets of independently driven refractive lenses (L1, L2), corresponding to different refractive compensation degrees (such as L1 is +2.0D and L2 is -2.0D) respectively. Each set of lenses is driven by a micro servo motor, and the angle adjustment within the range of 180° is realized through a gear-rack mechanism. The process of converting the target refractive power into a motor angle control command is as follows: Establish an "angle-refractive power" mapping relationship: According to the optical principle of dual-lens superposition, the target refractive power is determined by the effective angles of the two lenses in the optical path. For example, when at 0°, L1 is fully inserted, when at 90°, L2 is fully inserted, and the intermediate angles are realized through trigonometric function combinations for continuous adjustment; Driving strategy: When it is determined as an emergency event, the motor rotates quickly to the target angle with the maximum acceleration, and at the same time activates the dual-lens synchronous calibration mode, and the position of the lens is real-time fed back through the built-in Hall sensor to ensure independent adjustment of the left and right eyes (to adapt to users with anisometropia).

[0095] The above mechanical control scheme realizes distortion-free refractive adjustment through physical optical switching. For example, for users with binocular degree differences, the left / right lenses can be independently adjusted to different angles to provide personalized compensation. If a single-lens structure is adopted, it will not be able to meet the precise adjustment requirements in the anisometropia scenario.

[0096] Step S2300 solves the problem that the traditional fixed delay parameter cannot adapt to the rapid distance switching in the emergency adjustment event. By dynamically shortening the delay time ( ) and combining the predicted lag amount ( ), synchronize the lens compensation with the actual response rhythm of the ciliary muscle. For example, in a classroom scenario where the user frequently switches the fixation target, without this step, the lens compensation will lag behind the physiological adjustment demand, resulting in the accumulation of ineffective adjustment actions, exacerbating refractive fluctuations and visual fatigue. This step adjusts the compensation timing in real time through electromyogram energy feedback, enabling the lens to reach the target refractive power before the ciliary muscle completes the adjustment, reducing the ineffective load, and significantly improving the dynamic visual clarity. Solve the refractive anisometropia adaptation problem through independent dual-lens drive, expanding the applicable population of the device (such as strabismus patients).

[0097] Step S2400, if it is determined to be a stable adjustment event, adopt a conventional compensation mechanism; based on the fixation distance at the current moment Periodically calculate the refractive compensation amount , according to Adjust the lens.

[0098] Specifically, the conventional compensation mechanism triggers the adjustment of the lens refractive power at a fixed time interval, and the length of the time interval is determined by the compensation response time corresponding to the discretization intervals of the pitch angle θ and yaw angle φ in the individual initial compensation coefficient matrix A. Specifically, in a stable adjustment event, the changes in the user's head posture and fixation trajectory are relatively gentle, and the ciliary muscle adjustment demand does not exceed the conventional physiological response range. At this time, the lens compensation strategy samples the fixation distance D t at a predetermined time period (for example, every 200 milliseconds), combines the θ and φ discretization indices in the current head posture data, retrieves the corresponding refractive compensation coefficient a ijk,l from the individual initial compensation coefficient matrix A, and performs a weighted calculation of this coefficient and the current electromyogram energy to obtain F normal . When performing the weighted calculation, different weight factors are assigned according to the stratification (low, medium, high) of the electromyogram energy to dynamically adapt to the intensity difference of the ciliary muscle adjustment load. If F normalIf the actual refractive power difference corresponding to the current lens angle exceeds a preset refractive threshold (such as 0.1 D), the smooth switching mode is triggered: the motor rotates at a constant speed to the target angle to avoid mechanical wear caused by high-frequency micro-adjustments. For example, when switching from 0.5 D compensation to 0.7 D compensation, the motor rotates at a constant speed to reduce the impact load on the rack and pinion mechanism; in the non-switching state, the motor enters the "zero-torque standby mode", only keeping the angle encoder working to reduce system energy consumption. The angle encoder is a high-precision position sensor integrated inside the micro servo motor, which is used to monitor the rotation angle of the lens module in real time and provide closed-loop position feedback for the lens adjustment system. This strategy solves the problem of hardware resource waste caused by over-reliance on dynamic prediction models in stable scenarios. For example, in the slow reading scenario, the system samples at a fixed period, avoiding the high-frequency operation of the real-time prediction model and extending the device battery life. Without periodic control, the system may still frequently adjust the lens even in low-load scenarios, resulting in shortened motor life and increased power consumption.

[0099] Step S2400 uses the fast look-up table mechanism of matrix A to convert complex real-time prediction into parameter retrieval and weighted calculation, significantly reducing the processor operation pressure. The smooth switching mode avoids frequent start-stop and rapid acceleration of the motor. Combined with the zero-torque standby mode, it can improve the lifespan of mechanical components; in coordination with the online update mechanism in step S2500, the periodic compensation data provides samples for the incremental learning of matrix A, enabling it to adapt to long-term changes in individual adjustment habits (such as the improvement of accommodation ability due to the development of the ciliary muscle in children).

[0100] Step S2500, after each refractive compensation of the lens, synchronously collects new bimodal perception data during the user's gaze switching process for online update of the individual initial compensation coefficient matrix A for online update.

[0101] Specifically, after each refractive compensation of the lens, synchronously collects new bimodal perception data during the user's gaze switching process for online update of the individual initial compensation coefficient matrix A. Specifically, after completing the lens adjustment in step S2300 or S2400, through the nine-axis inertial sensor and dry electrode bioelectric sensor built into the head-mounted device, the updated head pose data, gaze distance, and ciliary muscle electrical signal are obtained in real time, and the new data is spatio-temporally aligned with the historical data. After alignment, the data is processed by band-pass filtering and baseline correction to extract the updated myoelectric energy. At the same time, based on the discretized interval index of the current head pitch angle and yaw angle, the coefficient a to be updated in the individual initial compensation coefficient matrix A is located. ijk,l。Online update adopts a sliding time window mechanism. Taking the refractive compensation operation data of the most recent N2 times (for example, N2 = 50 times) as the training set, the coefficients in matrix A are gradually corrected through an incremental learning algorithm. During the incremental learning process, the weight distribution between new data and original data is dynamically adjusted according to the data timestamp. The newer the data, the higher the weight is assigned (for example, the linear decay factor λ = 0.95), so as to ensure that matrix A can adapt to the changes in the adjustment habits of individual users (such as the migration of head movement patterns caused by changes in the learning environment). This step solves the problem of mismatch between the initial compensation coefficient matrix A and the real-time physiological state caused by long-term user use. For example, during the growth and development of child users, the ciliary muscle accommodation ability gradually increases, and the initial modeling data cannot cover this dynamic change. Without this step, the compensation strategy will gradually deviate from the actual needs of users, resulting in the accumulation of refractive compensation lag errors, and ultimately leading to an increase in the proportion of ineffective adjustment actions and exacerbating visual fatigue. This step and the multiple regression modeling in step S1233 form an iterative optimization closed-loop. Through the online update mechanism, the dual-modal perception data collected in real time is injected back into the initial compensation model, enabling matrix A to evolve from a static reference data to a dynamically adaptable model, thereby improving the long-term stability of the lens compensation strategy. In addition, there is data interaction between this step and the accommodation lag prediction model in step S2200. The updated matrix A provides the latest coefficient benchmark for the prediction model, avoiding the amplification of prediction deviation caused by outdated compensation coefficients. After the two cooperate, the triggering timing of the lead compensation acceleration mechanism in emergency events is more accurate.

[0102] Embodiment 3 Based on Embodiment 1 and Embodiment 2, this embodiment provides a refractive compensation device based on dual-modal perception, as Figure 6 shown, including: Partition module: used to establish the body coordinate system O-XYZ of the head-mounted device, and define the first region and the second region in the coordinate system O-XYZ; Dual-modal data acquisition module: The user performs back-and-forth fixation switching between the first region and the second region at different switching speeds, and acquires the dual-modal perception data during the user's fixation switching process; the dual-modal refers to the physiological signal modality and the spatial motion modality; Adjustment event determination module: Based on the dual-modal perception data during the user's fixation switching process, determine the emergency adjustment event and the stable adjustment event; Refractive compensation module: If it is determined as an emergency adjustment event, trigger the lead compensation acceleration mechanism; if it is determined as a stable adjustment event, adopt the conventional compensation mechanism.

[0103] In addition, the parts of the above technical solutions provided in the embodiments of the present application that are consistent with the corresponding technical solutions in the prior art are not described in detail to avoid excessive elaboration.

[0104] The specific embodiments described above further elaborate in detail the objectives, technical solutions and beneficial effects of the present invention. It should be understood that the above description is only the specific embodiments of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. A refractive compensation method based on bimodal perception, characterized in that, The method includes: Establishing a coordinate system O-XYZ of the head-mounted device body, and defining a first region and a second region in the coordinate system O-XYZ; the user performs back-and-forth gaze switching between the first region and the second region at different switching speeds, and collects bimodal perception data during the user's gaze switching process; the bimodality refers to the physiological signal modality and the spatial motion modality; Based on the bimodal perception data during the user's gaze switching process, determining an emergency adjustment event and a smooth adjustment event; if it is determined to be an emergency adjustment event, triggering an advanced compensation acceleration mechanism; if it is determined to be a smooth adjustment event, adopting a conventional compensation mechanism.

2. The refractive compensation method based on bimodal perception according to claim 1, wherein The method for establishing the coordinate system O-XYZ of the head-mounted device body is: taking the midpoint of the line connecting the user's two eye pupils as the origin O, the X-axis is along the direction of the line connecting the two eye pupils from left to right, the Y-axis points to the top of the head, and the Z-axis direction is defined as the straight-ahead direction of the line of sight.

3. The refractive compensation method based on bimodal perception according to claim 2, wherein The method for defining the first region and the second region in the coordinate system O-XYZ is: the interval Z∈[a, b] is defined as the first region, and the interval Z∈[c, d] is defined as the second region, where a, b, c, and d are positive real numbers and a < b < c < d.

4. A refractive compensation method based on bimodal perception according to claim 1, characterized in that The bimodal perception data includes physiological signal data and spatial motion data; The physiological signal data refers to ciliary muscle electrical signals; The spatial motion data includes the user's head pose data, gaze distance, and three-dimensional gaze point coordinates.

5. A refractive compensation method based on bimodal perception according to claim 4, characterized in that The method for determining the emergency adjustment event and the smooth adjustment event is: Calculate the distance change rate V based on the head pose data and the fixation distance D , the head movement gradient G H and the three-dimensional trajectory curvature κ; Calculating the change slope ke of the ciliary muscle electrical signal; Based on the rate of change of distance V D , the head movement gradient G H , the three-dimensional trajectory curvature κ, and the change slope ke of the ciliary muscle electrical signal, determine emergency accommodation events and smooth accommodation events.

6. A refractive compensation method based on bimodal perception according to claim 5, characterized in that, The method for calculating the distance change rate V D is as follows: A sliding time window is adopted, and based on the fixation distance, the distance change rate V D is calculated.

7. A refractive compensation method based on bimodal perception according to claim 5, characterized in that The three-dimensional trajectory curvature is calculated as follows: Generate a global space curve from the three-dimensional fixation point coordinates; select the current fixation point from the global space curve With the previous fixation points to form a local space curve, and calculate the curvature of the local space curve as the three-dimensional trajectory curvature .

8. A refractive compensation method based on bimodal perception according to claim 5, characterized in that The distance change rate V D , the head movement gradient G H , the three-dimensional trajectory curvature κ, and the change slope ke of the ciliary muscle electrical signal. The method for determining emergency accommodation events and smooth accommodation events is as follows: If and and and > k th , it is determined as an emergency adjustment event; otherwise, it is a stable adjustment event. Among them, is a preset distance change rate threshold, is a preset head movement gradient threshold, is a preset trajectory curvature threshold, k th is a preset change slope threshold.

9. A refractive compensation method based on bimodal perception according to claim 1, characterized in that The above-mentioned lead compensation acceleration mechanism is: dynamically shortening the adjustment delay parameter , to obtain the shortened compensation delay time , based on calculate the target refractive power in advance , according to adjust the lens; the adjustment delay parameter is synchronously obtained when collecting bimodal perception data during the user's gaze switching process.

10. A refractive compensation device based on bimodal perception, which is used to implement a refractive compensation method based on bimodal perception according to any one of claims 1-9, characterized in that, The device includes: A partition module: used for establishing a coordinate system O-XYZ of the head-mounted device body and defining a first region and a second region in the coordinate system O-XYZ; A bimodal data acquisition module: the user performs back-and-forth gaze switching between the first region and the second region at different switching speeds, and collects bimodal perception data during the user's gaze switching process; the bimodality refers to the physiological signal modality and the spatial motion modality; An adjustment event determination module: based on the bimodal perception data during the user's gaze switching process, determining an emergency adjustment event and a smooth adjustment event; A refractive compensation module: if it is determined to be an emergency adjustment event, triggering an advanced compensation acceleration mechanism; if it is determined to be a smooth adjustment event, adopting a conventional compensation mechanism.

Images