Intelligent intraocular pressure dynamic glasses monitoring method and system based on multi-modal sensing

Abstract

The invention provides an intelligent intraocular pressure dynamic glasses monitoring method and system based on multi-modal sensing, and relates to the technical field of glasses monitoring, and the method comprises the steps: carrying out the determination of a monitoring point position and the collection of point position multi-modal data through an eye sensor and a pose sensor which are disposed in glasses, and obtaining the point position multi-modal collection data; performing multi-modal and intraocular pressure data influence analysis on the point location multi-modal acquisition data to obtain multi-modal influence analysis data and intraocular pressure influence analysis data; according to the multi-modal influence analysis data, combined with the intraocular pressure influence analysis data, multi-modal intraocular pressure dynamic analysis is carried out, multi-modal intraocular pressure dynamic monitoring data is obtained, and through multi-dimensional data collection and fusion analysis, the scene flexibility and the precision accuracy are greatly improved.

Description

Technical Field

[0001] This invention proposes a method and system for intelligent dynamic intraocular pressure monitoring of glasses based on multimodal sensing, which relates to the field of glasses monitoring technology, specifically to the field of intelligent dynamic intraocular pressure monitoring of glasses based on multimodal sensing. Background Technology

[0002] Traditional intraocular pressure (IOP) monitoring methods, exemplified by the Goldmann applanation tonometer, suffer from drawbacks such as reliance on professional medical personnel, the need for hospital settings, and the inability to achieve dynamic and continuous monitoring. They also struggle to capture fluctuations in IOP during daily activities, potentially leading to missed diagnoses of early-stage conditions. Existing portable IOP monitoring devices often employ a single sensing modality, making them susceptible to interference from head posture and eye movements, resulting in insufficient measurement accuracy. Furthermore, they lack comprehensive analysis of multi-dimensional data, failing to effectively eliminate the influence of interfering factors on IOP results. Currently, there is no mature solution that organically combines ocular physiological sensing with posture sensing to achieve dynamic and interference-resistant IOP monitoring. Summary of the Invention

[0003] This invention provides a method and system for intelligent intraocular pressure dynamic glasses monitoring based on multimodal sensing, in order to solve the above-mentioned problems:

[0004] The present invention proposes a method and system for monitoring dynamic intraocular pressure in smart glasses based on multimodal sensing, wherein the method includes: S1. The monitoring points are determined and multimodal data of the points are collected by eye sensors and posture sensors installed in the glasses, and multimodal data of the points are obtained. S2. Perform multimodal and intraocular pressure data impact analysis on the multimodal acquisition data at the locations to obtain multimodal impact analysis data and intraocular pressure impact analysis data; S3. Based on the multimodal impact analysis data and the intraocular pressure impact analysis data, perform multimodal intraocular pressure dynamic analysis to obtain multimodal intraocular pressure dynamic monitoring data.

[0005] Further, S1 includes: Eye sensors and pose sensors are set on glasses, and the eye sensors and pose sensors are combined to obtain a sensor group; Multimodal sensing data is acquired by collecting multimodal sensing data from the sensor array. Multiple monitoring points are generated based on the monitoring area of ​​the glasses. The multimodal sensor data is divided according to the monitoring point information to obtain multimodal data of the points.

[0006] Furthermore, the step of generating multiple monitoring points based on the glasses monitoring area, and dividing the multimodal sensing data according to the monitoring point information to obtain multimodal data of the monitoring points includes: The sensor group acquires information about the monitoring area of ​​the glasses, and the monitoring area information of the glasses is used to identify regional features to obtain regional feature identification information. Multiple monitoring points are generated based on the regional feature identification information; Data features are extracted from the multimodal sensor data at each monitoring point to obtain point feature extraction data; Similarity calculations are performed on the extracted point feature data to obtain point feature similarity data; Based on the similarity data of the points, the point feature extraction data is analyzed to determine the representative points, and then the multimodal acquisition data of the representative points are obtained.

[0007] Furthermore, the step of performing representative point analysis on the point feature extraction data based on point feature similarity data to obtain representative point multimodal acquisition data includes: Compare the location feature similarity data with the location feature similarity threshold to obtain location feature similarity comparison data; Multiple similar monitoring points are determined based on the comparison data of similarity of location features, and the similar monitoring points are combined to obtain a similar point combination; Obtain the average value of similar point feature data for similar point combinations to obtain the mean value of similar point combinations; Obtain the absolute value of the difference between the feature similarity data of each location and the mean of the combination of similar locations, and obtain the feature difference coefficient of the location; The similarity difference sequence is obtained by sorting the feature difference coefficients of multiple similar point combinations from smallest to largest. The monitoring point corresponding to the feature difference coefficient of the point ranked first in the similarity difference sequence is selected as the representative monitoring point; The multimodal acquisition data representing the monitoring points is called point multimodal acquisition data.

[0008] Further, S2 includes: The multimodal acquisition data of the points is divided according to the preset time series information to obtain the time series point information; Based on the time-series location information, acquire the time-series multimodal acquisition data of each time-series node of multiple representative monitoring points; Multimodal impact analysis was performed based on time-series multimodal acquisition data to obtain multimodal impact analysis data; Intraocular pressure (IOP) impact analysis was performed based on time-series multimodal acquisition data to obtain IOP impact analysis data.

[0009] Furthermore, the step of performing multimodal impact analysis based on time-series multimodal acquisition data to obtain multimodal impact analysis data includes: Multiple preset pose data are extracted from time-series multimodal acquisition data to obtain multiple pose extraction data; The extracted pose data is compared with preset pose data to obtain multiple pose comparison data. Based on the comparison data of various poses, normal pose extraction data and abnormal pose extraction data are determined; Determine the timing information of the influence of normal pose and the timing information of the influence of abnormal pose based on the extracted data of normal pose and abnormal pose. The timing information of the influence of normal pose and the timing information of the influence of abnormal pose constitute the multimodal influence analysis data.

[0010] Further, based on the aforementioned multiple pose comparison data, normal pose extraction data and abnormal pose extraction data are determined, including: Based on multiple pose comparison data, the difference between the extracted pose data and the preset pose data is obtained to obtain the pose difference data for each pose. The pose difference data is compared with a preset pose difference threshold to obtain the pose difference comparison result; When the pose difference data is greater than the preset pose difference threshold, the pose extraction data corresponding to the pose difference data is determined to be abnormal pose extraction data. When the pose difference data is less than or equal to the preset pose difference threshold, the pose extraction data corresponding to the pose difference data is determined to be normal pose extraction data.

[0011] Furthermore, the step of performing intraocular pressure impact analysis based on time-series multimodal acquisition data to obtain intraocular pressure impact analysis data includes: Intraocular pressure (IOP) data was extracted from time-series multimodal acquisition data to obtain IOP extracted data. The extracted intraocular pressure data is compared with a preset intraocular pressure range to obtain the intraocular pressure comparison result; Based on the intraocular pressure comparison results, intraocular pressure abnormality is determined from the extracted intraocular pressure data to obtain abnormal intraocular pressure determination data. The abnormal intraocular pressure determination data is the intraocular pressure impact analysis data.

[0012] Further, S3 includes: Based on preset time-series information, the multimodal impact analysis data and intraocular pressure impact analysis data are correlated with time-series nodes to obtain time-series multimodal intraocular pressure impact data; Based on the time-series multimodal intraocular pressure (IOP) impact data, obtain the multimodal impact analysis data corresponding to the IOP impact analysis data, and obtain the associated multimodal impact data; Clustering and combining data from the analysis of intraocular pressure impact yielded intraocular pressure clustering and combination information; Clustering and combining of associated multimodal impact data based on intraocular pressure clustering combination information yields multimodal clustering combination information; The intraocular pressure clustering information and the multimodal clustering information are combined to obtain multimodal intraocular pressure dynamic monitoring data.

[0013] Furthermore, the system includes: The data acquisition module is used to determine the monitoring points and acquire multimodal data of the points by using eye sensors and pose sensors installed in the glasses, thereby obtaining multimodal data acquisition of the points. The impact analysis module is used to perform multimodal and intraocular pressure data impact analysis on the multimodal acquisition data of the location, and obtain multimodal impact analysis data and intraocular pressure impact analysis data; The monitoring module is used to perform multimodal intraocular pressure dynamic analysis based on multimodal impact analysis data and intraocular pressure impact analysis data, and obtain multimodal intraocular pressure dynamic monitoring data.

[0014] The beneficial effects of this invention are as follows: This method avoids the problem that traditional intraocular pressure monitoring usually relies on single sensor data, which leads to inaccurate and inflexible monitoring results. Through a closed-loop process of collection, analysis and fusion, it realizes the collaborative analysis of ocular physiological information and head posture information, which greatly reduces the impact of interference factors such as posture changes on the monitoring results and greatly improves the accuracy of the monitoring results. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of a smart intraocular pressure dynamic glasses monitoring method based on multimodal sensing. Detailed Implementation

[0016] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0017] In one embodiment of the present invention, the present invention proposes a method and system for monitoring dynamic intraocular pressure in smart glasses based on multimodal sensing, the method comprising: S1. The monitoring points are determined and multimodal data of the points are collected by eye sensors and posture sensors installed in the glasses, and multimodal data of the points are obtained. S2. Perform multimodal and intraocular pressure data impact analysis on the multimodal acquisition data at the locations to obtain multimodal impact analysis data and intraocular pressure impact analysis data; S3. Based on the multimodal impact analysis data and the intraocular pressure impact analysis data, perform multimodal intraocular pressure dynamic analysis to obtain multimodal intraocular pressure dynamic monitoring data.

[0018] The working principle and technical effects of the above-mentioned technical solution are as follows: This method uses smart glasses to monitor the eyes and posture, integrating eye sensors and posture sensors to achieve multimodal monitoring and analysis. Monitoring points are analyzed and determined using sensors, and eye physiological data and head posture data are collected at each monitoring point to obtain a set of multimodal data. Based on the collected data, a dual-dimensional analysis of eye and posture is performed to analyze the interference patterns of posture and other multimodal data on intraocular pressure measurement, and to assess the abnormal characteristics of the intraocular pressure data itself. The two types of analysis results are then temporally correlated and fused to achieve multimodal data-driven dynamic intraocular pressure monitoring, outputting comprehensive monitoring results. This method avoids the problem of inaccurate and inflexible monitoring results caused by traditional intraocular pressure monitoring that typically relies on single sensor data. Through a closed-loop process of acquisition, analysis, and fusion, it achieves collaborative analysis of eye physiological information and head posture information, greatly reducing the impact of interference factors such as posture changes on the monitoring results and significantly improving the accuracy of the monitoring results.

[0019] In one embodiment of the present invention, S1 includes: An eye sensor and a pose sensor are installed on the glasses, and the eye sensor and the pose sensor are combined to obtain a sensor group; the eye sensor includes an intraocular pressure sensor and an eye shape sensor, etc., and the pose sensor includes a triaxial accelerometer, a distance sensor and a tilt sensor, etc. Multimodal sensing data is acquired by using a sensor array; the multimodal sensing data includes intraocular pressure data, triaxial acceleration data, distance data (distance between the eye and the screen, etc.), and tilt angle data, etc. Multiple monitoring points are generated based on the glasses monitoring area. The multimodal sensor data is then segmented based on the monitoring point information to obtain multimodal data for each point. The glasses monitoring area includes an eye detection area and a pose detection area, etc. The working principle and technical effects of the above solution are as follows: Eye sensors and posture sensors are installed in the smart glasses, and these two types of sensors are combined into a sensor group. The eye sensors collect data directly related to eye condition, such as intraocular pressure, while the posture sensors collect data reflecting the wearer's posture and eye habits, such as triaxial acceleration, viewing distance, head tilt angle, and body displacement. The sensor group simultaneously collects multimodal sensing data across all dimensions. Based on the glasses' monitoring range, two monitoring areas are defined: an eye monitoring area and a posture monitoring area. Multiple monitoring points (randomly determined) are set within each monitoring area. The collected multimodal data is classified according to the data collection attribution information of each monitoring point, obtaining the multimodal acquisition data corresponding to each monitoring point. This method, through the combined configuration of sensors, achieves the synchronous collection of eye physiological indicators and head posture parameters, thereby enabling the corresponding analysis of abnormal displacement and abnormal physiological phenomena. Dividing monitoring points and classifying data based on the monitoring area avoids the mixing of data from different areas and types, greatly improving the targeted analysis of the data.

[0020] In one embodiment of the present invention, the step of generating multiple monitoring points based on the glasses monitoring area, and dividing the multimodal sensing data according to the monitoring point information to obtain multimodal data of the monitoring points includes: The system acquires information about the glasses' monitoring area using a sensor array, and then performs regional feature recognition based on this information to obtain regional feature recognition information. This regional feature recognition information includes eye feature information and pose feature information, among others. Multiple monitoring points are generated based on the regional feature identification information; Data features are extracted from the multimodal sensor data at each monitoring point to obtain point feature extraction data; Similarity calculations are performed on the extracted point feature data to obtain point feature similarity data; Based on the similarity data of the points, the point feature extraction data is analyzed to determine the representative points, and then the multimodal acquisition data of the representative points are obtained.

[0021] The following is a specific example of similarity calculation (this is just an example and does not represent the actual calculation): Assume that three monitoring points P1, P2, and P3 are generated within the eyeglass monitoring area. The multimodal feature extraction data of each point includes three core dimensions (intraocular pressure data stability, triaxial acceleration fluctuation value, and average eye distance). The similarity calculation adopts feature standardization and cosine similarity methods. The specific process is as follows: data that is not available is not included in the calculation.

[0022] The feature extraction data from the three locations are standardized (to eliminate the difference in dimensions between different data points) to obtain the standardized feature vector: P1 standardized eigenvector: [0.82, 0.61, 0.78] P2 standardized eigenvectors: [0.79, 0.58, 0.76] P3 standardized eigenvectors: [0.45, 0.89, 0.42] Calculate the cosine similarity between the feature vectors of any two points (the closer the cosine value is to 1, the higher the similarity; the value ranges from 0 to 1): The similarity between P1 and P2 was calculated using the vector dot product formula. The numerical differences in the feature values ​​of each dimension of P1 and P2 were small, and the final similarity was 0.97, indicating that the two features were highly similar. Calculating the similarity between P1 and P3 includes calculating the similarity of intraocular pressure stability and the similarity of triaxial acceleration fluctuation values ​​between P1 and P3. The similarity between P2 and P3 was calculated by considering that the core feature trend of P2 is consistent with that of P1, while the feature trend of P3 is significantly opposite, resulting in a similarity of 0.60.

[0023] The obtained data on similarity of location features include P1-P2 (0.97), P1-P3 (0.62), and P2-P3 (0.60). Based on this, it can be determined that P1 and P2 are similar monitoring locations, and P3 is an independent monitoring location. Representative locations will be selected based on the combination of P1-P2, and P3 will be treated as a separate type of location.

[0024] The working principle and technical effects of the above technical solution are as follows: Based on the raw data collected by the sensor group, eye feature information (such as intraocular pressure data distribution) and posture feature information (such as head tilt angle change range, limb acceleration fluctuation range, etc.) are extracted from the eye monitoring area. Multiple comprehensive monitoring points are generated in both the eye monitoring area and the posture monitoring area. For each monitoring point, corresponding feature data is extracted from the multimodal sensor acquisition data. The correlation of data at each point is measured by calculating the similarity of feature data at different points. Based on the similarity analysis results, representative monitoring points that can represent the data characteristics of similar points are selected, and their corresponding core data source of multimodal acquisition data is determined. By generating monitoring points through regional feature recognition, the comprehensiveness of point coverage is ensured, avoiding data omissions due to monitoring blind spots. By selecting representative points through similarity calculation, redundant data is effectively eliminated, greatly reducing the amount of computation and analysis, and improving data processing efficiency. At the same time, it is ensured that the data of representative points can reflect the core characteristics of similar points.

[0025] In one embodiment of the present invention, the step of performing representative point analysis on the point feature extraction data based on point feature similarity data to obtain representative point multimodal acquisition data includes: Compare the location feature similarity data with the location feature similarity threshold to obtain location feature similarity comparison data; Multiple similar monitoring points are determined based on the comparison data of similarity of location features, and the similar monitoring points are combined to obtain a similar point combination; Obtain the average value of similar point feature data for similar point combinations to obtain the mean value of similar point combinations; Obtain the absolute value of the difference between the feature similarity data of each location and the mean of the combination of similar locations, and obtain the feature difference coefficient of the location; The similarity difference sequence is obtained by sorting the feature difference coefficients of multiple similar point combinations from smallest to largest. The monitoring point corresponding to the feature difference coefficient of the point ranked first in the similarity difference sequence is selected as the representative monitoring point; The multimodal acquisition data representing the monitoring points is called point multimodal acquisition data.

[0026] The working principle and technical effect of the above technical solution are as follows: A similarity threshold for location features is set based on historical data classification experience. The calculated similarity data for location features is compared with the threshold to filter out similar monitoring points that meet the similarity requirements and combine them. The average value of the similarity data for all locations within each similar location combination is calculated to obtain the mean of the similar location combination. Then, the absolute value of the difference between the similarity data for each location within the combination and the mean of the combination is calculated. This absolute value of the difference is used as the location feature difference coefficient to measure the degree of deviation between the location data and the average feature of the combination. The feature difference coefficients of all locations within the combination are sorted in ascending order to form a similarity difference sequence. The monitoring point corresponding to the first-ranked feature difference coefficient in the sequence, i.e., the one with the smallest deviation, is selected as the representative monitoring point. The multimodal acquisition data of the representative monitoring point is the final multimodal acquisition data for the location. By setting similarity thresholds to filter similar points, accurate representative classification of points of the same type is achieved, avoiding interference from data of dissimilar points. The point with the smallest feature difference coefficient is used as the representative point to ensure that the data of the selected point best matches the average characteristics of points of the same type, thereby improving the representativeness and reliability of the data. At the same time, representative points are determined by quantitative calculation, replacing subjective judgment by manual screening.

[0027] In one embodiment of the present invention, S2 includes: The multimodal acquisition data of the points is divided according to the preset time series information to obtain the time series point information; Based on the time-series location information, acquire the time-series multimodal acquisition data of each time-series node of multiple representative monitoring points; Multimodal impact analysis was performed based on time-series multimodal acquisition data to obtain multimodal impact analysis data; Intraocular pressure (IOP) impact analysis was performed based on time-series multimodal acquisition data to obtain IOP impact analysis data.

[0028] The working principle and technical effects of the above technical solution are as follows: Based on preset time-series segmentation rules, the multimodal acquisition data at various points is segmented into datasets of different time-series nodes, forming time-series point information. Based on the time-series point information, the time-series multimodal acquisition data corresponding to each representative monitoring point at each time-series node is extracted. A two-dimensional analysis is performed on the time-series data, including multimodal influence analysis, to obtain the interference patterns of pose and other multimodal data on intraocular pressure (IOP) monitoring. IOP influence analysis is then performed to identify abnormal characteristics of the IOP data, and both multimodal influence analysis data and IOP influence analysis data are output separately. This method transforms the acquired data into a time-series dataset through time-series segmentation, enabling dynamic tracking and analysis of IOP changes. The multimodal influence analysis and IOP influence analysis clarify the interference of external posture factors on the monitoring results and also grasp the abnormal state of IOP itself. The results of the two-dimensional analysis complement each other, improving the comprehensive understanding of IOP monitoring data.

[0029] A two-dimensional analysis of time-series data is performed, including multimodal impact analysis, to obtain the interference patterns of multimodal data such as pose on intraocular pressure monitoring. Intraocular pressure impact analysis is then conducted to identify abnormal characteristics of the intraocular pressure data. The specific analysis process for outputting multimodal impact analysis data and intraocular pressure impact analysis data includes: Head tilt angle, eye distance and other pose data are extracted from temporal multimodal data, compared with preset standard pose data, the difference is calculated and compared with a threshold to distinguish normal / abnormal pose data, determine the corresponding temporal interval, and output multimodal influence analysis data.

[0030] Intraocular pressure (IOP) data is extracted from time-series data and compared with a preset normal IOP range to determine whether it is abnormal (such as exceeding the range or excessive fluctuation). The output includes IOP impact analysis data containing abnormal IOP and the corresponding time series.

[0031] In one embodiment of the present invention, the step of performing multimodal impact analysis based on time-series multimodal acquisition data to obtain multimodal impact analysis data includes: Multiple preset pose data are extracted from time-series multimodal acquisition data to obtain multiple pose extraction data; The extracted pose data is compared with preset pose data to obtain multiple pose comparison data. Based on the comparison data of various poses, normal pose extraction data and abnormal pose extraction data are determined; Based on the extracted data of normal pose and the extracted data of abnormal pose, determine the timing information of the influence of normal pose and the timing information of the influence of abnormal pose; the preset timing information corresponding to the extracted data of normal pose is the timing information of the influence of normal pose, and the preset timing information corresponding to the extracted data of abnormal pose is the timing information of the influence of abnormal pose. The timing information of the influence of normal pose and the timing information of the influence of abnormal pose constitute the multimodal influence analysis data.

[0032] The working principle and technical effects of the above technical solution are as follows: Multiple pose-related data, such as head tilt angle, eye distance, and triaxial acceleration, are extracted from temporal multimodal acquisition data to form multiple pose extraction data. These pose extraction data are compared with pre-set standard pose data to obtain multiple pose comparison data. Based on the comparison results, the pose extraction data are divided into normal pose extraction data that conforms to the standard and abnormal pose extraction data that deviates from the standard. Then, the temporal information corresponding to the two types of data is determined, namely, the temporal information of normal pose influence and the temporal information of abnormal pose influence. These two types of temporal information together constitute multimodal influence analysis data. By extracting pose data and comparing it with standard data, accurate determination of the wearer's eye posture is achieved; the temporal intervals corresponding to normal and abnormal poses are clearly distinguished, and the period of influence of posture interference on intraocular pressure monitoring is clearly defined, improving the accuracy of distinguishing between true intraocular pressure changes and pseudo-intraocular pressure changes caused by posture interference, and improving the accuracy of intraocular pressure monitoring results.

[0033] One embodiment of the present invention, determining normal pose extraction data and abnormal pose extraction data based on the multiple pose comparison data, includes: Based on multiple pose comparison data, the difference between the extracted pose data and the preset pose data is obtained to obtain the pose difference data for each pose. The pose difference data is compared with a preset pose difference threshold to obtain the pose difference comparison result; the preset pose difference threshold is the preset correct eye posture extraction data, such as maintaining a preset distance from the computer screen, etc. When the pose difference data is greater than the preset pose difference threshold, the pose extraction data corresponding to the pose difference data is determined to be abnormal pose extraction data. When the pose difference data is less than or equal to the preset pose difference threshold, the pose extraction data corresponding to the pose difference data is determined to be normal pose extraction data.

[0034] The working principle and technical effects of the above technical solution are as follows: Based on the comparison data of various poses obtained in multimodal influence analysis, the difference between the extracted data of each pose and the preset standard pose data is calculated to obtain the difference data of each pose; a preset pose difference threshold is set, which is determined with reference to the parameter range of correct eye posture (such as reasonable distance from the computer screen and standard head tilt angle); the calculated pose difference data is compared with the threshold. If the difference data is greater than the threshold, the corresponding pose extraction data is determined to be abnormal pose extraction data; if the difference data is less than or equal to the threshold, it is determined to be normal pose extraction data. By quantifying the difference data and comparing it with the threshold, the objective determination of normal and abnormal poses is achieved, avoiding errors caused by subjective judgment; the threshold setting refers to the standard of reasonable eye posture, making the judgment results more in line with the actual application scenario, and effectively identifying the pose data corresponding to poor eye posture; thus improving the operability and accuracy of multimodal influence analysis.

[0035] In one embodiment of the present invention, the step of performing intraocular pressure impact analysis based on time-series multimodal acquisition data to obtain intraocular pressure impact analysis data includes: Intraocular pressure (IOP) data was extracted from time-series multimodal acquisition data to obtain IOP extracted data. The extracted intraocular pressure data is compared with a preset intraocular pressure range to obtain the intraocular pressure comparison result; Based on the intraocular pressure comparison results, intraocular pressure abnormality is determined from the extracted intraocular pressure data to obtain abnormal intraocular pressure determination data. The abnormal intraocular pressure determination data is the intraocular pressure impact analysis data.

[0036] The working principle and technical effects of the above technical solution are as follows: Intraocular pressure (IOP) related data are extracted from time-series multimodal acquisition data to form an IOP extraction dataset; a preset IOP range is set (referencing the standard IOP interval for eye health), and the extracted IOP data is compared with this range to obtain IOP comparison results; based on the comparison results, IOP abnormality judgment is performed. If the IOP data exceeds the preset range, it is judged as abnormal IOP data. Finally, all abnormal judgment results are integrated into abnormal IOP judgment data, i.e., IOP impact analysis data. By selectively extracting IOP data and focusing on core monitoring indicators for analysis, the relevance of the analysis is improved; abnormality judgment based on the preset IOP range can quickly identify abnormal characteristics of IOP data; the judgment results are output in the form of abnormal IOP judgment data, with a clear structure.

[0037] In one embodiment of the present invention, S3 includes: Based on preset time-series information, the multimodal impact analysis data and intraocular pressure impact analysis data are correlated with time-series nodes to obtain time-series multimodal intraocular pressure impact data; Based on the time-series multimodal intraocular pressure (IOP) impact data, obtain the multimodal impact analysis data corresponding to the IOP impact analysis data, and obtain the associated multimodal impact data; Clustering and combining data from the analysis of intraocular pressure impact yielded intraocular pressure clustering and combination information; Clustering and combining of associated multimodal impact data based on intraocular pressure clustering combination information yields multimodal clustering combination information; The intraocular pressure clustering information and the multimodal clustering information are combined to obtain multimodal intraocular pressure dynamic monitoring data.

[0038] Specific examples are as follows (these are just examples and do not represent the actual situation): First, map the two types of data at each time point to form a time-series multimodal intraocular pressure impact data set. For example: 08:00 Node: Abnormal posture (looking down at a mobile phone, tilt angle difference exceeding the threshold) and abnormal intraocular pressure (25 mmHg, exceeding the 10-21 mmHg range); 09:15 Node: Normal posture (sitting posture with eyes level with the computer, distance and tilt angle meet the standards) and normal intraocular pressure (18 mmHg); 14:30 node: abnormal posture (bending over to pick up an object, acceleration fluctuation exceeding the threshold) and abnormal intraocular pressure (23 mmHg); 19:45 Node: Normal posture (sitting reading, distance meets standard) and normal intraocular pressure (17 mmHg); 21:00 node: abnormal posture (lying on one's side while looking at a mobile phone, with both tilt angle and distance exceeding the standard) and abnormal intraocular pressure (24 mmHg); From the above correlated data, we extracted the pose influence data corresponding to the abnormal intraocular pressure nodes, and obtained: Abnormal intraocular pressure node (08:00) corresponds to abnormal posture (looking down at a mobile phone); Abnormal intraocular pressure node (14:30) corresponds to abnormal posture (bending over to pick up an object); Abnormal intraocular pressure node (21:00) corresponds to abnormal posture (lying on one's side while looking at a mobile phone); Abnormal intraocular pressure data were clustered according to numerical range. Since the three abnormal intraocular pressure values ​​(25, 23, and 24 mmHg) were all in the range of 22-26 mmHg, they were grouped into the same cluster group. The intraocular pressure cluster group information was high intraocular pressure group (22-26 mmHg), which included nodes 08:00, 14:30, and 21:00.

[0039] Based on the intraocular pressure clustering results above, the corresponding associated pose data were categorized to obtain multimodal clustering information as high intraocular pressure associated abnormal pose groups, including three types of poses: looking down at a mobile phone, bending over to pick up an object, and lying on one's side while looking at a mobile phone.

[0040] By integrating the two types of clustering information, multimodal intraocular pressure dynamic monitoring data is formed: the high intraocular pressure group (22-26 mmHg) corresponds to abnormal postures (looking down at a mobile phone, bending over to pick up an object, lying on one's side to look at a mobile phone), involving monitoring nodes 08:00, 14:30, and 21:00, clearly showing the correlation between high intraocular pressure and poor posture.

[0041] The working principle and technical effect of the above technical solution are as follows: Based on preset time-series information, multimodal impact analysis data and intraocular pressure (IOP) impact analysis data are precisely correlated at time-series nodes, so that IOP data at each time-series node corresponds to the corresponding pose impact state, forming time-series multimodal IOP impact data; through the time-series multimodal IOP impact data, multimodal impact analysis data corresponding to each IOP impact analysis data is extracted to obtain associated multimodal impact data; the IOP impact analysis data is clustered and combined, grouping IOP abnormal data with similar characteristics into one category to form IOP clustering combination information; based on the IOP clustering combination information, the corresponding associated multimodal impact data is synchronously clustered and combined to obtain multimodal clustering combination information; the two types of clustering combination information are integrated to form complete multimodal IOP dynamic monitoring data. By associating time-series nodes, precise matching of intraocular pressure (IOP) data and posture-related data is achieved, enabling the acquisition of the correlation between IOP abnormalities and head posture. Clustering is used to integrate the two types of data, classifying monitoring results with similar characteristics to facilitate rapid identification of typical IOP abnormality patterns and their corresponding causes. The output multimodal IOP dynamic monitoring data can comprehensively reflect the IOP change patterns and influencing factors.

[0042] According to one embodiment of the present invention, the system includes: The data acquisition module is used to determine the monitoring points and acquire multimodal data of the points by using eye sensors and pose sensors installed in the glasses, thereby obtaining multimodal data acquisition of the points. The impact analysis module is used to perform multimodal and intraocular pressure data impact analysis on the multimodal acquisition data of the location, and obtain multimodal impact analysis data and intraocular pressure impact analysis data; The monitoring module is used to perform multimodal intraocular pressure dynamic analysis based on multimodal impact analysis data and intraocular pressure impact analysis data, and obtain multimodal intraocular pressure dynamic monitoring data.

[0043] The working principle and technical effects of the above-mentioned technical solution are as follows: This method uses smart glasses to monitor the eyes and posture, integrating eye sensors and posture sensors to achieve multimodal monitoring and analysis. Monitoring points are analyzed and determined using sensors, and eye physiological data and head posture data are collected at each monitoring point to obtain a set of multimodal data. Based on the collected data, a dual-dimensional analysis of eye and posture is performed to analyze the interference patterns of posture and other multimodal data on intraocular pressure measurement, and to assess the abnormal characteristics of the intraocular pressure data itself. The two types of analysis results are then temporally correlated and fused to achieve multimodal data-driven dynamic intraocular pressure monitoring, outputting comprehensive monitoring results. This method avoids the problem of inaccurate and inflexible monitoring results caused by traditional intraocular pressure monitoring that typically relies on single sensor data. Through a closed-loop process of acquisition, analysis, and fusion, it achieves collaborative analysis of eye physiological information and head posture information, greatly reducing the impact of interference factors such as posture changes on the monitoring results and significantly improving the accuracy of the monitoring results.

[0044] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for monitoring dynamic intraocular pressure in smart glasses based on multimodal sensing, characterized in that, The method includes: S1. The monitoring points are determined and multimodal data of the points are collected by eye sensors and posture sensors installed in the glasses, and multimodal data of the points are obtained. S2. Perform multimodal and intraocular pressure data impact analysis on the multimodal acquisition data at the locations to obtain multimodal impact analysis data and intraocular pressure impact analysis data; S3. Based on the multimodal impact analysis data and the intraocular pressure impact analysis data, perform multimodal intraocular pressure dynamic analysis to obtain multimodal intraocular pressure dynamic monitoring data.

2. The intelligent intraocular pressure dynamic glasses monitoring method based on multimodal sensing according to claim 1, characterized in that, S1 includes: Eye sensors and pose sensors are set on glasses, and the eye sensors and pose sensors are combined to obtain a sensor group; Multimodal sensing data is acquired by collecting multimodal sensing data from the sensor array. Multiple monitoring points are generated based on the monitoring area of ​​the glasses. The multimodal sensor data is divided according to the monitoring point information to obtain multimodal data of the points.

3. The intelligent intraocular pressure dynamic glasses monitoring method based on multimodal sensing according to claim 2, characterized in that, The process involves generating multiple monitoring points based on the glasses' monitoring area, dividing the multimodal sensor data based on the monitoring point information, and obtaining multimodal data for each monitoring point, including: The sensor group acquires information about the monitoring area of ​​the glasses, and the monitoring area information of the glasses is used to identify regional features to obtain regional feature identification information. Multiple monitoring points are generated based on the regional feature identification information; Data features are extracted from the multimodal sensor data at each monitoring point to obtain point feature extraction data; Similarity calculations are performed on the extracted point feature data to obtain point feature similarity data; Based on the similarity data of the points, the point feature extraction data is analyzed to determine the representative points, and then the multimodal acquisition data of the representative points are obtained.

4. The intelligent intraocular pressure dynamic glasses monitoring method based on multimodal sensing according to claim 3, characterized in that, The step of analyzing and determining representative points based on similarity data of point features to obtain multimodal acquisition data of representative points includes: Compare the location feature similarity data with the location feature similarity threshold to obtain location feature similarity comparison data; Multiple similar monitoring points are determined based on the comparison data of similarity of location features, and the similar monitoring points are combined to obtain a similar point combination; Obtain the average value of similar point feature data for similar point combinations to obtain the mean value of similar point combinations; Obtain the absolute value of the difference between the feature similarity data of each location and the mean of the combination of similar locations, and obtain the feature difference coefficient of the location; The similarity difference sequence is obtained by sorting the feature difference coefficients of multiple similar point combinations from smallest to largest. The monitoring point corresponding to the feature difference coefficient of the point ranked first in the similarity difference sequence is selected as the representative monitoring point; The multimodal acquisition data representing the monitoring points is called point multimodal acquisition data.

5. The intelligent intraocular pressure dynamic glasses monitoring method based on multimodal sensing according to claim 1, characterized in that, S2 includes: The multimodal acquisition data of the points is divided according to the preset time series information to obtain the time series point information; Based on the time-series location information, acquire the time-series multimodal acquisition data of each time-series node of multiple representative monitoring points; Multimodal impact analysis was performed based on time-series multimodal acquisition data to obtain multimodal impact analysis data; Intraocular pressure (IOP) impact analysis was performed based on time-series multimodal acquisition data to obtain IOP impact analysis data.

6. The intelligent intraocular pressure dynamic glasses monitoring method based on multimodal sensing according to claim 5, characterized in that, The step of performing multimodal impact analysis based on time-series multimodal acquisition data to obtain multimodal impact analysis data includes: Multiple preset pose data are extracted from time-series multimodal acquisition data to obtain multiple pose extraction data; The extracted pose data is compared with preset pose data to obtain multiple pose comparison data. Based on the comparison data of various poses, normal pose extraction data and abnormal pose extraction data are determined; Determine the timing information of the influence of normal pose and the timing information of the influence of abnormal pose based on the extracted data of normal pose and abnormal pose. The timing information of the influence of normal pose and the timing information of the influence of abnormal pose constitute the multimodal influence analysis data.

7. The intelligent intraocular pressure dynamic glasses monitoring method based on multimodal sensing according to claim 6, characterized in that, Based on the aforementioned multiple pose comparison data, normal pose extraction data and abnormal pose extraction data are determined, including: Based on multiple pose comparison data, the difference between the extracted pose data and the preset pose data is obtained to obtain the pose difference data for each pose. The pose difference data is compared with a preset pose difference threshold to obtain the pose difference comparison result; When the pose difference data is greater than the preset pose difference threshold, the pose extraction data corresponding to the pose difference data is determined to be abnormal pose extraction data. When the pose difference data is less than or equal to the preset pose difference threshold, the pose extraction data corresponding to the pose difference data is determined to be normal pose extraction data.

8. The intelligent intraocular pressure dynamic glasses monitoring method based on multimodal sensing according to claim 5, characterized in that, The step of performing intraocular pressure (IOP) impact analysis based on time-series multimodal acquisition data to obtain IOP impact analysis data includes: Intraocular pressure (IOP) data was extracted from time-series multimodal acquisition data to obtain IOP extracted data. The extracted intraocular pressure data is compared with a preset intraocular pressure range to obtain the intraocular pressure comparison result; Based on the intraocular pressure comparison results, intraocular pressure abnormality is determined from the extracted intraocular pressure data to obtain abnormal intraocular pressure determination data. The abnormal intraocular pressure determination data is the intraocular pressure impact analysis data.

9. The intelligent intraocular pressure dynamic glasses monitoring method based on multimodal sensing according to claim 1, characterized in that, S3 includes: Based on preset time-series information, the multimodal impact analysis data and intraocular pressure impact analysis data are correlated with time-series nodes to obtain time-series multimodal intraocular pressure impact data; Based on the time-series multimodal intraocular pressure (IOP) impact data, obtain the multimodal impact analysis data corresponding to the IOP impact analysis data, and obtain the associated multimodal impact data; Clustering and combining data from the analysis of intraocular pressure impact yielded intraocular pressure clustering and combination information; Clustering and combining of associated multimodal impact data based on intraocular pressure clustering combination information yields multimodal clustering combination information; The intraocular pressure clustering information and the multimodal clustering information are combined to obtain multimodal intraocular pressure dynamic monitoring data.

10. A smart intraocular pressure dynamic glasses monitoring system based on multimodal sensing, characterized in that, The system includes: The data acquisition module is used to determine the monitoring points and acquire multimodal data of the points by using eye sensors and pose sensors installed in the glasses, thereby obtaining multimodal data acquisition of the points. The impact analysis module is used to perform multimodal and intraocular pressure data impact analysis on the multimodal acquisition data of the location, and obtain multimodal impact analysis data and intraocular pressure impact analysis data; The monitoring module is used to perform multimodal intraocular pressure dynamic analysis based on multimodal impact analysis data and intraocular pressure impact analysis data, and obtain multimodal intraocular pressure dynamic monitoring data.

Images