Blood oxygen monitoring data error correction method

Abstract

The invention discloses a blood oxygen monitoring data error correction method, which mainly relates to the technical field of blood oxygen monitoring, and comprises the following steps: implementing data acquisition in a multi-dimensional way according to model information of a target blood oxygen monitor, and performing similar aggregation on acquisition results to obtain K implementation starting data sets; obtaining K scene description associated keyword sets; obtaining K scene error concentration degrees and K scene error concentration values; obtaining an inertial scene description associated keyword set and a conventional scene description associated keyword set; obtaining a matching result; and performing real-time error correction on the real-time blood oxygen monitoring data of the target blood oxygen monitor to obtain real-time corrected blood oxygen monitoring data. The technical problems that in the prior art, blood oxygen monitoring is affected by scenes, and monitoring data have large errors are solved, and the technical effect of improving the accuracy of the blood oxygen monitoring data is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of blood oxygen monitoring, and in particular to a method for correcting blood oxygen monitoring data errors. Background Art

[0002] A blood oxygen monitor is a device commonly used in the medical field that can monitor the blood oxygen concentration of the human body in real time. It is widely used in hospitals, homes, and clinical diagnosis. However, due to environmental factors, equipment performance, and the diversity of usage scenarios, existing blood oxygen monitors often have data deviations or errors, which affect the accuracy of the monitoring results. These errors may be caused by sensor instability, equipment failure, or external environment (such as temperature and humidity changes). In addition, most of the existing error correction methods for blood oxygen monitoring equipment rely on static calibration data and cannot effectively correct the monitoring results according to the actual application scenarios. Therefore, the accuracy and adaptability of error correction are poor.

[0003] The existing technology has a technical problem that the blood oxygen monitoring is affected by the scene and the monitoring data has large errors. Summary of the invention

[0004] The present application provides a method for correcting errors in blood oxygen monitoring data, which is used to solve the technical problem in the prior art that the monitoring data has large errors due to the influence of the scene when performing blood oxygen monitoring.

[0005] In view of the above problems, the present application provides a method for correcting blood oxygen monitoring data errors, the method comprising:

[0006] According to the model information of the target blood oxygen monitor, data collection is implemented through a multi-dimensional approach, and the collection results are aggregated for similarity to obtain K implementation and activation data sets, where K is an integer greater than or equal to 1; the K implementation and activation data sets are traversed to perform iterative association extraction of scene description keywords to obtain K scene description associated keyword sets; the K implementation and activation data sets are traversed to perform monitoring error extraction to obtain K monitoring error sets, and error distribution identification is performed on the K monitoring error sets to obtain K scene error concentrations and K scene error concentration values; the K scene error concentrations are binary classified using a preset concentration threshold to obtain an inertial scene description associated keyword set and a conventional scene description associated keyword set; the real-time application scenario information of the target blood oxygen monitor is obtained, and the real-time application scenario information is matched with the inertial scene description associated keyword set and the conventional scene description associated keyword set respectively to obtain a matching result; the real-time error correction of the real-time blood oxygen monitoring data of the target blood oxygen monitor is performed based on the matching result and the K scene error concentration values ​​to obtain real-time corrected blood oxygen monitoring data.

[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0008] The present application implements data collection through a multi-dimensional approach according to the model information of the target blood oxygen monitor, performs similar aggregation on the collection results, obtains K implementation and activation data sets, where K is an integer greater than or equal to 1, and then traverses the K implementation and activation data sets to perform iterative association extraction of scene description keywords to obtain K scene description associated keyword sets, traverses the K implementation and activation data sets to perform monitoring error extraction to obtain K monitoring error sets, and performs error distribution identification on the K monitoring error sets to obtain K scene error concentrations and K scene error concentration values, and then uses a preset concentration threshold to perform binary classification on the K scene error concentrations to obtain an inertial scene description associated keyword set and a conventional scene description associated keyword set, and then obtains real-time application scenario information of the target blood oxygen monitor, matches the real-time application scenario information with the inertial scene description associated keyword set and the conventional scene description associated keyword set respectively to obtain matching results, and performs real-time error correction on the real-time blood oxygen monitoring data of the target blood oxygen monitor based on the matching results and the K scene error concentration values ​​to obtain real-time corrected blood oxygen monitoring data. The technical effect of accurately correcting errors in blood oxygen monitoring data based on multidimensional data analysis and error distribution identification has been achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Attached Figure 1 A flow chart of a method for correcting blood oxygen monitoring data errors provided in an embodiment of the present application;

[0010] Attached Figure 2 A schematic diagram of a process for determining K scene error concentration neighborhoods in a blood oxygen monitoring data error correction method provided in an embodiment of the present application. DETAILED DESCRIPTION

[0011] The present invention will be further described below in conjunction with specific embodiments. It should be understood that these embodiments are only used to illustrate the present invention and are not intended to limit the scope of the present invention. In addition, it should be understood that after reading the content taught by the present invention, those skilled in the art can make various changes or modifications to the present invention, and these equivalent forms fall within the scope limited by the appended claims of the application equally.

[0012] Embodiment, as attached Figure 1 As shown, the present application provides a method for correcting blood oxygen monitoring data errors, wherein the method comprises:

[0013] S100: performing multi-dimensional data collection according to the model information of the target blood oxygen monitor, and performing similar aggregation on the collection results to obtain K implementation data sets, where K is an integer greater than or equal to 1;

[0014] In a possible embodiment, the target blood oxygen monitor is any monitoring device that needs to perform monitoring data error correction. Model information refers to the specific model of the blood oxygen monitor used, which usually includes information such as the technical specifications, hardware configuration, and software functions of the device. Model information is usually key information that identifies the type and version of the device.

[0015] The operation data of the blood oxygen monitor is collected from multiple dimensions and multiple implementation scenarios. For example, data may be collected from different sensors of the device, at different time periods, or under different environmental conditions. This data collection method can fully reflect the performance of the target blood oxygen monitor in different scenarios. By classifying and merging the collected data, data with similar characteristics are classified into the same category. This is a data preprocessing step, the purpose of which is to integrate similar data from different sources into a more manageable set.

[0016] Optionally, implementation data is collected through a multi-dimensional approach to obtain an initial implementation activation data set, an initial implementation activation data is randomly selected from the initial implementation activation data set as a first aggregation target, the cosine similarity between the initial implementation activation data set and the first aggregation target is traversally calculated, and the initial implementation activation data whose calculation result is greater than or equal to a preset cosine similarity threshold is added to the first implementation activation data set; the first implementation activation data set is eliminated from the initial implementation activation data set, and an initial implementation activation data is randomly selected from the remaining initial implementation activation data sets again, and similar aggregation of the same type is performed based on the same principle as that for obtaining the first implementation activation data set to obtain a second implementation activation data set; after multiple aggregations, until the implementation activation data sets are aggregated, the K implementation activation data sets are obtained.

[0017] Among them, K implementation activation data sets refer to K implementation activation data sets that represent the implementation status of the target blood oxygen monitor in different implementation scenarios after aggregating similar data of the same type, and K is an integer greater than or equal to 1. Each implementation activation data set may represent monitoring data in a specific scenario or a class of specific states. By obtaining the K implementation activation data sets, the technical effect of providing diversified and accurate initial data for subsequent error correction is achieved.

[0018] S200: traversing the K implementation activation data sets to perform iterative association extraction of scene description keywords to obtain K scene description associated keyword sets;

[0019] In an embodiment of the present application, a scene description keyword refers to a keyword used to describe a specific application scenario. Scene description keywords are usually defined based on different environmental conditions, device states, or monitoring purposes, and may include information such as temperature, humidity, sensor state, and measurement accuracy. Iterative association extraction refers to gradually extracting keywords associated with the scene description keywords and keywords of the implicit relationship between the data in the K implementation-enabling data sets, respectively. Each round of iterative extraction will further optimize and extract new associated keywords based on existing information or previously extracted content. The K scene description associated keyword sets refer to a set of scene description keywords extracted from the K implementation-enabling data sets. These keywords describe the environment and device characteristics related to blood oxygen monitoring data in a specific application scenario.

[0020] The above step S200 of the present application reveals the scene information implicit in the data by extracting keywords, achieving the technical effect of providing key data support for subsequent error correction and scene matching. At the same time, by iteratively extracting related keywords, the differences between different monitoring scenes can be more carefully identified, paving the way for improving the accuracy of scene matching.

[0021] Further, traversing the K implementation activation data sets to perform iterative association extraction of scene description keywords to obtain K scene description associated keyword sets, step S200 of the embodiment of the present application further includes:

[0022] Extracting a first implementation activation data set from the K implementation activation data sets, and randomly extracting a first first implementation activation data from the first implementation activation data set;

[0023] Using a preset scene description keyword set as an index, performing scene description keyword similarity extraction on the first first implementation activation data to obtain a first extracted scene description keyword set, and adding text features corresponding to keywords in the first extracted scene description keyword set into a first iteration association vector;

[0024] Performing iterative association extraction of scene description words on the first implementation activation data set based on the first extracted scene description keyword set, the preset scene description keyword set and the first iterative association vector to obtain a first scene description association keyword set;

[0025] The K implementation data sets are combined with the preset scene description keyword set to perform scene description keyword iterative association extraction to obtain K scene description associated keyword sets.

[0026] Further, based on the first extracted scene description keyword set, the preset scene description keyword set and the first iterative association vector, the first implementation activation data set is subjected to iterative association extraction of scene description words to obtain a first scene description associated keyword set. Step S200 of the embodiment of the present application further includes:

[0027] Randomly extracting a second first implementation activation data from the first implementation activation data set again, performing scene description keyword iterative association extraction on the second first implementation activation data with the first iterative association vector, the preset scene description keyword set and the first extracted scene description keyword set as indexes to obtain a second extracted scene description keyword set, and performing union extraction on the text features corresponding to the keywords in the second extracted scene description keyword set and the text features in the first iterative association vector, and adding them into the second iterative association vector;

[0028] Based on the preset scene description keyword set, the first extracted scene description keyword set, the second extracted scene description keyword set and the second iterative association vector, the first implementation activation data set is subjected to iterative scene description keyword association extraction to obtain N extracted scene description keyword sets, wherein N is the number of first implementation activation data in the first implementation activation data set, and N is an integer greater than or equal to 1;

[0029] A union is obtained for the N extracted scene description keyword sets to obtain a first scene description associated keyword set.

[0030] In one embodiment, the preset scene description keyword set is a keyword set pre-defined or calibrated by those skilled in the art, which is usually descriptive information related to a specific application scenario and is used as a reference index for scene description keyword extraction.

[0031] The first implementation activation data set refers to a specific data set extracted from the K implementation activation data sets, and then the first first implementation activation data is randomly extracted from the first implementation activation data set as the starting point of this iteration. The first iteration association vector is a vector constructed based on the relationship between the current data set and the preset scene description keywords each time a keyword is extracted. This vector is used to represent the correlation between the currently extracted scene description words and the data text, further promoting the iterative process of keyword extraction.

[0032] The purpose of step S200 is to extract keywords related to the application scenario from K implementation-enabling data sets through an iterative process, so as to more accurately describe different monitoring scenarios. In this step, a first implementation-enabling data set is first extracted from the K implementation-enabling data sets, and then a piece of data is randomly selected from the set as the starting data (the first first implementation-enabling data). Next, by performing keyword extraction on the data, using a preset scene description keyword set as an index, a scene description keyword similar to the data is obtained. Optionally, the keywords extracted from the first first implementation-enabling data are calculated for cosine similarity with the preset scene description keyword set. When the calculation result meets the preset requirements, it is added to the first scene description associated keyword set. Then, the text features corresponding to these keywords are added to the first iterative association vector. Through this iterative process, the scene description keyword set can be further refined and optimized according to the similarity and correlation of the data.

[0033] Then, based on these initially extracted first scene description keyword set, the first iterative association vector and the preset scene description keyword set, the data set is continued to be iteratively extracted, and the scene description keyword iterative association extraction is performed on the second first implementation activation data to obtain the second extracted scene description keyword set, and the text features corresponding to the keywords in the second extracted scene description keyword set are extracted by union with the text features in the first iterative association vector and added to the second iterative association vector.

[0034] Then, the first implementation activation data set is subjected to iterative scene description keyword association extraction based on the latest second iteration association vector, the preset scene description keyword set, the first extracted scene description keyword set, and the second extracted scene description keyword set, to obtain N extracted scene description keyword sets. The purpose of this process is to accurately extract keywords related to a specific scene through repeated iterations and optimizations, and these keywords will be used in subsequent error correction and data analysis. Then, the N extracted scene description keyword sets are combined to avoid keyword duplication, thereby obtaining the first scene description associated keyword set.

[0035] Finally, this extraction method is applied to all K implementation activation data sets, and the N extracted scene description keyword sets are unioned to obtain the first scene description associated keyword set, and finally K scene description associated keyword sets are obtained.

[0036] S300: traversing the K implementation activation data sets to extract monitoring errors, obtaining K monitoring error sets, and performing error distribution identification on the K monitoring error sets to obtain K scene error concentrations and K scene error concentration values;

[0037] Further, the K implementation activation data sets are traversed to extract monitoring errors, K monitoring error sets are obtained, and error distribution identification is performed on the K monitoring error sets to obtain K scene error concentrations and K scene error concentration values. Step S300 of the embodiment of the present application further includes:

[0038] Constructing an error centralized identification coordinate system, wherein the x-axis of the error centralized identification coordinate system is the identification number, and the y-axis is the device test error;

[0039] Identify the monitoring errors in the K monitoring error sets with identification numbers respectively, and input the identified K monitoring error sets into the error set identification coordinate system respectively to obtain K coordinate point sets, wherein each coordinate point corresponds to a monitoring error;

[0040] In the error concentrated identification coordinate system, performing concentrated neighborhood identification on the K coordinate point sets respectively to determine K scene error concentrated neighborhoods;

[0041] Obtaining K scene error concentrations according to the proportions of the data amounts in the neighborhoods of the K scene error sets to the data amounts in the K monitoring error sets;

[0042] Calculate the mean error of the neighborhood in the K scene error sets to obtain the K scene error set values.

[0043] Further, as attached Figure 2 As shown, in the error concentrated identification coordinate system, concentrated neighborhood identification is performed on the K coordinate point sets respectively to determine K scene error concentrated neighborhoods. Step S300 of the embodiment of the present application further includes:

[0044] A straight line passing through the origin of the error concentration identification coordinate system and parallel to the x-axis of the error concentration identification coordinate system is used as a starting straight line;

[0045] Based on the K coordinate point sets, constructing K starting concentrated neighborhoods of the starting straight line according to a preset neighborhood bandwidth;

[0046] Perform edge diffusion on the K starting concentrated neighborhoods according to the preset neighborhood bandwidth to obtain K diffuse concentrated neighborhoods;

[0047] It is determined whether the amount of data in the K diffuse concentrated neighborhoods is greater than or equal to the amount of data in the K starting concentrated neighborhoods. If so, edge diffusion is continued for the K diffuse concentrated neighborhoods according to the preset neighborhood bandwidth until the preset diffusion times are met, thereby obtaining the K scene error concentrated neighborhoods.

[0048] Further, if not, the K starting concentrated neighborhoods are used as the K scene error concentrated neighborhoods.

[0049] In one possible embodiment, the K implementation and activation data sets are retrieved with the monitoring error as an index to obtain the K monitoring error sets. Among them, monitoring error extraction refers to extracting errors caused by environmental changes, device status or other factors from the blood oxygen monitoring data. These errors may come from the instability of the device itself, the external environment or the sensor. The K scene error concentrations reflect the concentration of errors in the scenes corresponding to the K implementation and activation data sets. The higher the scene error concentration, the more likely it is that inertial errors will occur in the scene, that is, the environment has a greater impact on the target blood oxygen monitor in such a scene. The higher the scene error concentration, the less impact the environment has on the target blood oxygen monitor in the scene, and the errors are mainly accidental errors, so the distribution is relatively discrete. The K scene error concentration values ​​reflect the error values ​​in the K implementation and activation data sets that best represent the general level of monitoring errors.

[0050] In one embodiment, the error concentration identification coordinate system is a two-dimensional coordinate system for visualizing and analyzing the distribution of monitoring errors. The x-axis of the coordinate system represents the identification number, and the y-axis represents the device test error, with the purpose of helping to identify the concentrated area of ​​the error. In the error concentration identification coordinate system, performing concentrated neighborhood identification on the K coordinate point sets respectively means identifying the clustered areas of error data in the error concentration identification coordinate system. The errors in these areas have a high degree of concentration, that is, the error values ​​are relatively close, which may represent the specific performance of the device in a certain scenario.

[0051] Optionally, first extract the monitoring error data from the K implementation data sets to obtain K monitoring error sets. These error data represent the error level in each scenario. Next, construct an error concentration identification coordinate system, where the x-axis is the test sequence number or time series, and the y-axis is the equipment test error. By inputting these monitoring error data into the coordinate system, K coordinate point sets can be obtained, and each coordinate point corresponds to a monitoring error.

[0052] Then, by performing "concentrated neighborhood identification" on these coordinate point sets, the concentrated areas of errors can be identified, which may represent the characteristics of errors in certain specific scenes. The error data of these concentrated neighborhoods are further analyzed, and the error mean of each concentrated neighborhood is calculated to obtain the error concentration value of each scene, and obtain the error concentration degrees and K scene error concentration values ​​of the K scenes.

[0053] A straight line passing through the origin of the error concentrated identification coordinate system and parallel to the x-axis of the error concentrated identification coordinate system is used as the starting straight line. Based on the K coordinate point sets, K starting concentrated neighborhoods of the starting straight line are constructed according to the preset neighborhood bandwidth. Among them, half of the preset neighborhood bandwidth is the distance from the neighborhood edge to the starting straight line or the previous neighborhood edge when a single neighborhood construction or neighborhood diffusion is performed in advance by a person skilled in the art. Among them, the distance from any monitoring error in the K starting concentrated neighborhoods to the starting straight line is less than or equal to half of the preset neighborhood bandwidth.

[0054] According to half of the preset neighborhood bandwidth, the upper edges of the neighborhoods of the K starting concentrated neighborhoods are edge diffused upward, and the lower edges of the neighborhoods are edge diffused downward, to obtain K diffuse concentrated neighborhoods. It is determined whether the amount of data in the K diffuse concentrated neighborhoods is greater than or equal to the amount of data in the K starting concentrated neighborhoods. If so, it indicates that the degree of data aggregation in the K diffuse concentrated neighborhoods is higher than that in the K starting concentrated neighborhoods. Then, the K diffuse concentrated neighborhoods are edge diffused according to the preset neighborhood bandwidth until the preset diffusion times (the maximum diffusion times pre-set by those skilled in the art) are met, to obtain the K scene error concentrated neighborhoods.

[0055] If not, it indicates that the data aggregation degree of the K starting concentrated neighborhoods is higher than that of the K diffuse concentrated neighborhoods. Then the K starting concentrated neighborhoods are used as the K scene error concentrated neighborhoods.

[0056] By accurately analyzing the distribution of error data, identifying and quantifying the error concentration areas in different scenarios, an accurate basis can be provided for subsequent real-time error correction. By calculating the error concentration and concentration value, the error characteristics of the device in different scenarios can be more effectively described, providing data support for subsequent scene differentiation.

[0057] S400: performing binary classification on the K scene error concentrations using a preset concentration threshold to obtain an inertial scene description associated keyword set and a conventional scene description associated keyword set;

[0058] S500: Acquire real-time application scenario information of the target blood oxygen monitor, and match the real-time application scenario information with the inertial scenario description associated keyword set and the conventional scenario description associated keyword set to obtain a matching result;

[0059] In one embodiment, the preset concentration threshold refers to a standard value pre-set by a person skilled in the art in error analysis to distinguish different degrees of error concentration. The threshold is used to determine whether the error concentration belongs to an "inertial scenario" or a "conventional scenario". When the error concentration exceeds the threshold, it may mean that the device is in an inertial scenario (i.e., the device is greatly affected by the environment of the application scenario), otherwise it is a conventional scenario (i.e., the device may be affected by other factors and the error is more dispersed).

[0060] The preset concentration threshold divides the error concentration of different scenes into two categories: inertial scenes and conventional scenes. Thus, the inertial scene description-associated keyword set and the conventional scene description-associated keyword set are obtained. The inertial scene description-associated keyword set refers to a set of keywords associated with scenes (inertial scenes) with high error concentration and stable consistency. Usually, these keywords may reflect the characteristics of the device under long-term stable use, such as factors such as constant ambient temperature and humidity. The conventional scene description-associated keyword set refers to a set of keywords associated with scenes (conventional scenes) with low error concentration and more dispersion. Conventional scenes may reflect the working state of the device under the influence of large external changes.

[0061] The real-time application scenario information refers to the scenario data of the target blood oxygen monitor in the current actual operation. These data may include environmental parameters (such as temperature, humidity), usage mode, and other real-time factors related to the blood oxygen monitoring process.

[0062] The real-time application scenario information is compared with the preset keyword sets of the inertial scenario description and the conventional scenario description to determine which keyword sets are more compatible with the current scenario, thereby determining the application environment of the device. Optionally, the similarity between the keywords in the real-time application scenario information and the inertial scenario description associated keyword set and the conventional scenario description associated keyword set is determined using a cosine similarity calculation formula or a Euclidean distance calculation formula, and the inertial scenario description associated keyword or the conventional scenario description associated keyword corresponding to the maximum similarity is used as a matching result.

[0063] Through real-time application scenario data and error concentration analysis, the error correction strategy of the monitoring data is dynamically adjusted. If the device is in an "inertial scenario", the error is relatively consistent, and the system can adopt a more accurate correction method; if the device is in a "conventional scenario", it is necessary to consider a wider range of factors to adjust the correction strategy, thereby improving the accuracy of the monitoring data.

[0064] S600: Performing real-time error correction on the real-time blood oxygen monitoring data of the target blood oxygen monitor based on the matching result and the K scene error concentration values ​​to obtain real-time corrected blood oxygen monitoring data.

[0065] Further, based on the matching result and the K scene error concentration values, real-time error correction is performed on the real-time blood oxygen monitoring data of the target blood oxygen monitor to obtain real-time corrected blood oxygen monitoring data. Step S600 of the embodiment of the present application further includes:

[0066] When the matching result is a successful match with the inertial scene description associated keyword set and the matched inertial scene description associated keyword is obtained, mapping and extracting the matched inertial scene description associated keyword and the K scene error concentrated values ​​to obtain the matched inertial scene error concentrated value;

[0067] Calculating a device correction coefficient based on the matched inertial scene error concentration value and the range of the target blood oxygen monitor to obtain a device correction coefficient;

[0068] The real-time blood oxygen monitoring data is corrected using the device correction coefficient to obtain the real-time corrected blood oxygen monitoring data.

[0069] Furthermore, when the matching result is a successful match with a set of keywords associated with a conventional scene description and a matching keyword associated with a conventional scene description is obtained, two neighboring blood oxygen monitoring data adjacent to the real-time blood oxygen monitoring data are obtained, and the average of the two neighboring blood oxygen monitoring data and the real-time blood oxygen monitoring data is calculated to obtain the real-time corrected blood oxygen monitoring data.

[0070] In a possible embodiment, real-time error correction refers to analyzing the blood oxygen data collected by the target blood oxygen monitor in real time, and correcting it according to the scene error concentration, concentration value and other information to ensure the accuracy of the blood oxygen data. The real-time corrected blood oxygen monitoring data is the error-corrected blood oxygen monitoring data. By comparing and adjusting with other scene data or neighboring data, more accurate real-time monitoring data is obtained.

[0071] Optionally, when the matching result is a successful match between the real-time application scenario information and the set of keywords associated with the inertial scenario description, the error concentration value related to the scenario is extracted from the K scene error concentration values ​​based on the matching result, and the device correction coefficient is calculated based on these concentration values ​​and the range of the device. The device correction coefficient is an important parameter used to adjust the blood oxygen monitoring data. Through the correction coefficient, the real-time blood oxygen data can be accurately corrected to obtain more accurate real-time blood oxygen monitoring data. Preferably, the ratio of the matching inertial scene error concentration value to the range of the target blood oxygen monitor is used as the device correction coefficient. Then, the real-time blood oxygen monitoring data is corrected using the device correction coefficient to obtain the real-time corrected blood oxygen monitoring data.

[0072] Optionally, if the real-time scene matches a set of keywords associated with a regular scene description, a different correction method will be used. In this case, two neighboring blood oxygen monitoring data adjacent to the real-time blood oxygen monitoring data are obtained, and the current blood oxygen monitoring data is corrected by calculating the mean of the two neighboring blood oxygen monitoring data and the current data. The purpose of this method is to correct the current data by smoothing the fluctuations between neighboring data, thereby reducing the impact of errors caused by environmental changes.

[0073] The purpose of these two correction processes is to flexibly select the appropriate correction method according to different scenarios and data characteristics. In stable inertial scenarios, the error concentration value and correction coefficient are used for more accurate correction; in conventional scenarios with large fluctuations, smooth correction is performed through neighboring data. This dynamic correction method helps to improve the accuracy and reliability of blood oxygen monitoring data.

[0074] In summary, the embodiments of the present application have at least the following technical effects:

[0075] This application collects data through multi-dimensional channels according to the model information of the target blood oxygen monitor, and performs similarity aggregation to obtain K implementation and activation data sets. Then, these data sets are traversed to perform iterative association extraction of scene description keywords to obtain K scene description associated keyword sets. Then, the monitoring errors in the data set are extracted, and K scene error concentrations and K scene error concentration values ​​are obtained through error distribution identification. Based on the preset concentration threshold, the K scene error concentrations are classified into two categories to distinguish inertial scenes from conventional scenes. By obtaining the application scenario information of the target blood oxygen monitor in real time, it is matched with the inertial scene description associated keyword set and the conventional scene description associated keyword set to obtain a matching result. Finally, based on the matching results and the K scene error concentration values, the real-time blood oxygen monitoring data is error corrected to obtain real-time corrected blood oxygen monitoring data. The technical effect of improving the accuracy of blood oxygen monitoring data is achieved.

[0076] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description and does not represent the advantages and disadvantages of the embodiments. And the above-mentioned specific embodiments of this specification are described. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0077] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application should be included in the protection scope of the present application.

[0078] This specification and drawings are merely exemplary illustrations of the present application and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. A blood oxygen monitoring data error correction method, characterized in that: The method comprises: Perform multi-dimensional data collection according to the model information of the target blood oxygen monitor, aggregate the collection results of the same type and similarity, and obtain K implementation data sets, where K is an integer greater than or equal to 1; Traversing the K implementation activation data sets to perform iterative association extraction of scene description keywords to obtain K scene description associated keyword sets; Traversing the K implementation activation data sets to extract monitoring errors, obtain K monitoring error sets, and perform error distribution identification on the K monitoring error sets to obtain K scene error concentrations and K scene error concentration values; Using a preset concentration threshold, the K scene error concentrations are classified into two categories to obtain an inertial scene description associated keyword set and a conventional scene description associated keyword set; Acquire real-time application scenario information of the target blood oxygen monitor, and match the real-time application scenario information with the inertial scenario description associated keyword set and the conventional scenario description associated keyword set to obtain a matching result; Based on the matching result and the K scene error concentration values, real-time error correction is performed on the real-time blood oxygen monitoring data of the target blood oxygen monitor to obtain real-time corrected blood oxygen monitoring data.

2. A blood oxygen monitoring data error correction method as claimed in claim 1, characterized in that: Traversing the K implementation activation data sets to perform iterative association extraction of scene description keywords, and obtaining K scene description associated keyword sets, including: Extracting a first implementation activation data set from the K implementation activation data sets, and randomly extracting a first first implementation activation data from the first implementation activation data set; Using a preset scene description keyword set as an index, performing scene description keyword similarity extraction on the first first implementation activation data to obtain a first extracted scene description keyword set, and adding text features corresponding to keywords in the first extracted scene description keyword set into a first iteration association vector; Performing iterative association extraction of scene description words on the first implementation activation data set based on the first extracted scene description keyword set, the preset scene description keyword set and the first iterative association vector to obtain a first scene description association keyword set; The K implementation data sets are combined with the preset scene description keyword set to perform scene description keyword iterative association extraction to obtain K scene description associated keyword sets.

3. A blood oxygen monitoring data error correction method as claimed in claim 2, characterized in that: The first implementation activation data set is subjected to iterative scene description word association extraction based on the first extracted scene description keyword set, the preset scene description keyword set and the first iterative association vector to obtain a first scene description associated keyword set, including: Randomly extracting a second first implementation activation data from the first implementation activation data set again, performing scene description keyword iterative association extraction on the second first implementation activation data with the first iterative association vector, the preset scene description keyword set and the first extracted scene description keyword set as indexes to obtain a second extracted scene description keyword set, and performing union extraction on the text features corresponding to the keywords in the second extracted scene description keyword set and the text features in the first iterative association vector, and adding them into the second iterative association vector; Based on the preset scene description keyword set, the first extracted scene description keyword set, the second extracted scene description keyword set and the second iterative association vector, the first implementation activation data set is subjected to iterative scene description keyword association extraction to obtain N extracted scene description keyword sets, wherein N is the number of first implementation activation data in the first implementation activation data set, and N is an integer greater than or equal to 1; A union is obtained for the N extracted scene description keyword sets to obtain a first scene description associated keyword set.

4. The method for correcting blood oxygen monitoring data errors according to claim 1, characterized in that: Traversing the K implementation activation data sets to extract monitoring errors, obtaining K monitoring error sets, and performing error distribution identification on the K monitoring error sets to obtain K scene error concentrations and K scene error concentration values, including: Constructing an error centralized identification coordinate system, wherein the x-axis of the error centralized identification coordinate system is the identification number, and the y-axis is the device test error; Identify the monitoring errors in the K monitoring error sets with identification numbers respectively, and input the identified K monitoring error sets into the error set identification coordinate system respectively to obtain K coordinate point sets, wherein each coordinate point corresponds to a monitoring error; In the error concentrated identification coordinate system, performing concentrated neighborhood identification on the K coordinate point sets respectively to determine K scene error concentrated neighborhoods; Obtaining K scene error concentrations according to the proportions of the data amounts in the neighborhoods of the K scene error sets to the data amounts in the K monitoring error sets; Calculate the mean error of the neighborhood in the K scene error sets to obtain the K scene error set values.

5. A blood oxygen monitoring data error correction method as claimed in claim 4, characterized in that: In the error concentrated identification coordinate system, concentrated neighborhood identification is performed on the K coordinate point sets respectively to determine K scene error concentrated neighborhoods, including: A straight line passing through the origin of the error concentration identification coordinate system and parallel to the x-axis of the error concentration identification coordinate system is used as a starting straight line; Based on the K coordinate point sets, constructing K starting concentrated neighborhoods of the starting straight line according to a preset neighborhood bandwidth; Perform edge diffusion on the K starting concentrated neighborhoods according to the preset neighborhood bandwidth to obtain K diffuse concentrated neighborhoods; It is determined whether the amount of data in the K diffuse concentrated neighborhoods is greater than or equal to the amount of data in the K starting concentrated neighborhoods. If so, edge diffusion is continued for the K diffuse concentrated neighborhoods according to the preset neighborhood bandwidth until the preset diffusion times are met, thereby obtaining the K scene error concentrated neighborhoods.

6. A blood oxygen monitoring data error correction method as claimed in claim 5, characterized in that: If not, the K starting concentrated neighborhoods are used as the K scene error concentrated neighborhoods.

7. A blood oxygen monitoring data error correction method as claimed in claim 1, characterized in that: Performing real-time error correction on the real-time blood oxygen monitoring data of the target blood oxygen monitor based on the matching result and the K scene error concentration values ​​to obtain real-time corrected blood oxygen monitoring data, including: When the matching result is a successful match with the inertial scene description associated keyword set and the matched inertial scene description associated keyword is obtained, mapping and extracting the matched inertial scene description associated keyword and the K scene error concentrated values ​​to obtain the matched inertial scene error concentrated value; Calculating a device correction coefficient based on the matched inertial scene error concentration value and the range of the target blood oxygen monitor to obtain a device correction coefficient; The real-time blood oxygen monitoring data is corrected using the device correction coefficient to obtain the real-time corrected blood oxygen monitoring data.

8. The method for correcting blood oxygen monitoring data errors according to claim 1, characterized in that: When the matching result is a successful match with the set of keywords associated with the general scene description, and the matching keywords associated with the general scene description are obtained, two neighboring blood oxygen monitoring data adjacent to the real-time blood oxygen monitoring data are obtained, and the average of the two neighboring blood oxygen monitoring data and the real-time blood oxygen monitoring data is calculated to obtain the real-time corrected blood oxygen monitoring data.

Images