Sensor data labeling method and tool based on wearable device and storage medium

Abstract

The invention discloses a sensor data annotation method and tool based on wearable equipment and a storage medium, and belongs to the technical field of sensor data annotation, the method comprises the following steps: obtaining to-be-annotated file data, preprocessing the received to-be-annotated file data, and marking the preprocessed to-be-annotated file data as target data; setting a corresponding display mode according to user requirements, and visually displaying the read data according to the display mode; establishing an annotation mode library and a feature library, obtaining annotation requirements of a user, and matching corresponding annotation modes and annotation models from the annotation mode library according to the annotation requirements; labeling the target data according to the matched labeling mode and labeling model; obtaining target annotation data; verifying the target annotation data, and correspondingly adjusting the target annotation data according to a verification result; the verified target annotation data is exported; and continuously optimizing, analyzing and adjusting the annotation pattern library.

Description

Technical Field

[0001] The present invention belongs to the technical field of sensor data annotation, and specifically relates to a sensor data annotation method, tool and storage medium based on a wearable device. Background Art

[0002] With the development of sensor technology, more and more industries rely on sensor data to work. Sensor data usually exists in digital form and needs to be labeled before it can be further analyzed and applied. However, for the processing and analysis of large amounts of sensor data, especially data labeling, traditional manual labeling methods have problems such as low efficiency, high cost, and insufficient accuracy. They are also prone to subjective errors, resulting in inaccurate labeling results. Therefore, a more intelligent and efficient data labeling tool and method is needed.

[0003] Based on this, the present invention provides a sensor data annotation method, tool and storage medium based on wearable devices. Through the combination of artificial intelligence technology and visualization, sensor data is automatically annotated, combined with manual verification, to provide training data for model construction efficiently and accurately. Summary of the invention

[0004] In order to solve the problems existing in the above scheme, the present invention provides a sensor data annotation method, tool and storage medium based on a wearable device.

[0005] The purpose of the present invention can be achieved through the following technical solutions: A sensor data annotation method based on a wearable device, the method comprising: Step 1: Obtain the file data to be labeled, preprocess the received file data to be labeled, and mark the preprocessed file data to be labeled as target data; Step 2: Set the corresponding display mode according to user needs, and visualize the read data according to the display mode; Step 3: Establish a labeling pattern library and a feature library, wherein the labeling pattern library is used to store various labeling patterns and the labeling models corresponding to each labeling pattern; the feature library is used to store various labeling features; obtain the user's labeling requirements, and match the corresponding labeling pattern and labeling model from the labeling pattern library according to the labeling requirements; Furthermore, the method of matching the corresponding annotation mode and the annotation model from the annotation mode library according to the annotation requirement includes: Analyze the annotation requirements and determine the characteristics of the requirements; generate test sets based on the characteristics of the requirements and set standard cases based on the test sets; Input the demand characteristics into the annotation mode library for matching, and obtain various candidate annotation modes that meet the demand characteristics; Use the test set to conduct simulation tests on each candidate annotation method to obtain the annotation accuracy rate and annotation consistency rate corresponding to each candidate annotation method; and set the annotation cases corresponding to each candidate annotation method. Evaluate each annotation case according to the standard case to obtain the case evaluation value corresponding to each candidate annotation method. Calculate the corresponding annotation value according to the formula PY = b1 × PA(s) × (b2 × QA + B3 × QB). In the formula: PY is the annotation value; b1, b2, and b3 are all proportionality coefficients, and the value range is 0 < b1 ≤ 1, 0 < b2 ≤ 1, 0 < b3 ≤ 1; PA(s) is the case evaluation value output by the case evaluation model; QA is the annotation accuracy rate; QB is the annotation consistency rate. Sort each candidate annotation method in descending order of the annotation value to obtain the first sequence, and insert the corresponding annotation cases at each candidate annotation method in the first sequence. Determine the target annotation method according to the first sequence, and identify the annotation pattern and annotation model corresponding to the target annotation method.

[0006] Further, the method for evaluating each annotation case according to the standard case includes: Establish a case evaluation model, and the expression of the case evaluation model is: ; In the formula: s is the input data, and the input data includes the standard case and the annotation case; the output data is the case evaluation value; BA is the optimization value, and the value range of the optimization value is [0, 1). Input the standard case and each annotation case into the case evaluation model for analysis to obtain the corresponding case evaluation value.

[0007] Further, when the corresponding annotation pattern and annotation model cannot be matched from the annotation pattern library according to the annotation requirements, assist the user to establish the corresponding annotation pattern and annotation model.

[0008] Further, the method for assisting the user to establish the corresponding annotation pattern and annotation model includes: Step SA1: Calculate the similarity between the requirement characteristics and each annotation pattern, mark the annotation patterns with a similarity not lower than the threshold X1 as the reference patterns, identify the annotation models corresponding to the reference patterns, and obtain the annotation principles of the annotation models; determine the corresponding adjustment items according to the annotation principles, and adjust the annotation principles according to the adjustment items and the requirement characteristics to obtain the corrected principles. Step SA2: Conduct annotation simulation according to the corrected principle to form multiple simulation data; display the simulation data to the user, and the user makes adjustments according to the displayed simulation data, and adjusts the corrected principle according to the adjustment results of the simulation data to obtain a new corrected principle. Step SA3: loop step SA2 until there is no new correction principle; Step SA4: Establish a training adjustment model, obtain a training set corresponding to the annotation model, analyze the annotation principle, the training set and the correction principle through the training adjustment model, and obtain training adjustment data; when no training adjustment data is obtained, eliminate the corresponding reference model; Step SA5: Set a new annotation model and annotation mode that meet the required characteristics based on the training adjustment data and annotation model.

[0009] Step 4: Annotate the target data according to the matching annotation mode and annotation model; obtain the target annotation data; Step 5: Verify the target annotation data, and adjust the target annotation data accordingly according to the verification result; export the verified target annotation data; Step 6: Continuously optimize, analyze and adjust the annotation pattern library.

[0010] Furthermore, the method for continuously optimizing, analyzing and adjusting the annotation pattern library includes: Acquire material data in real time, establish a material evaluation model, analyze each material data through the material evaluation model, and determine the classification category corresponding to each material data; Classify the material data according to the classification categories corresponding to them to obtain corresponding first classification data, second classification data and third classification data; classify the first classification data, second classification data and third classification data again to obtain optimized data of each unit; and label the optimized data of each unit with corresponding classification category labels; The recognition unit optimizes the data volume corresponding to the data, and performs corresponding optimization processing when the data volume reaches a preset standard.

[0011] Furthermore, the expression of the material evaluation model is: ; In the formula: q is the input data, the input data is the material data; the output data is the category value.

[0012] A tool for implementing the above data annotation method, comprising a data import and file indexing module, a visual display module, a feature library management module, a pattern library management module, a statistics module and a label processing module; The data import and file indexing module is used to obtain the data to be annotated and pre-process the data to be annotated; The visual display module is used to visually display the read sensor data according to a preset hierarchical management method; The feature library management module is used to manage the feature library; The pattern library management module is used to manage the pattern library; The statistical module is used to display the statistical information of the tags in real time; The label processing module is used to process multiple types of annotations, set up a manual verification interface, and the user processes the label results based on the manual verification interface; and exports the corresponding data according to a preset format.

[0013] A storage medium stores a computer program, which implements the above-mentioned data labeling method when executed by a processor.

[0014] Compared with the prior art, the present invention has the following beneficial effects: The introduction of a hierarchical management mechanism allows each layer of data to be loaded and displayed separately or simultaneously, and users can selectively load different data layers as needed. At the same time, customized options are provided, such as adjusting the curve color, line type, and label color and shape, to meet the personalized needs of users, increasing the plasticity and convenience of the user experience. The introduction of interactive functions, such as dragging, zooming, and other operations, further enhances the user's ability to operate and control the data, making the annotation process more flexible and intuitive. Allowing users to select existing annotation modes or create new modes and rules according to specific data annotation needs provides great flexibility. Users can fine-tune according to the characteristics of the data and the annotation goals to ensure the accuracy and effectiveness of the annotation. This flexibility makes the system more suitable for a variety of different annotation tasks.

[0015] By presetting annotation modes, data can be automatically annotated, significantly reducing the time and effort of manual annotation, thereby improving annotation efficiency. At the same time, the data visualization function enables users to intuitively observe data characteristics and annotation results, which helps to more quickly identify patterns and formulate annotation rules, as well as verify the accuracy of annotation results. Users can easily manage multiple annotation modes, including operations such as adding, deleting, modifying, and calling modes. It can store new features and new modes, and reuse existing features and annotation modes. As data annotation tasks continue, new annotation data and optimized models can continuously update and maintain the pattern library, providing more powerful support for future annotation tasks. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0017] Figure 1 The figure is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0018] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0019] like Figure 1 As shown, a sensor data labeling method based on a wearable device includes: Step 1: Import the file data to be labeled from the outside, perform preprocessing operations such as cleaning, denoising, and format conversion on the received file data to improve data quality, and mark the preprocessed file data to be labeled as target data; Step 2: Select the data to be visualized. You can choose three axes, total acceleration, acceleration vector and other characteristic data; adjust the curve color, line type, label color and shape according to needs; set the corresponding display mode according to user needs; and then display the data according to the set display mode; Step 3: Establish a labeling pattern library and a feature library. The labeling pattern library is used to store various labeling patterns and the labeling models corresponding to each labeling pattern. One labeling pattern can correspond to multiple labeling models; the feature library is used to store various labeling features; obtain the user's labeling requirements, and match the corresponding labeling pattern and labeling model from the labeling pattern library according to the labeling requirements; In the process of data annotation, the annotation rules or modes are pre-set. A single mode or a combination of multiple modes can be selected: if there is an annotation mode that matches the current data type and annotation requirements in the annotation mode library, the preset mode is directly called and the corresponding annotation model under the mode is selected; if there is no matching annotation mode available in the annotation mode library, the custom annotation mode process is entered. First, users can fine-tune and optimize the existing annotation mode to adapt to the current specific annotation requirements. Secondly, if the existing annotation mode cannot meet the requirements, users need to deeply analyze the data characteristics and select relevant annotation features from the feature library, modify the existing annotation features or create new annotation features. These annotation features should closely revolve around the core requirements of data annotation to ensure the accuracy and effectiveness of the annotation. Subsequently, users need to associate the selected annotation features with the corresponding annotation models or rules. This involves operations such as configuring parameters and adjusting rule weights to ensure that the newly defined annotation mode can accurately capture the key information in the data. After completing the configuration, users can choose to store the newly created annotation features and annotation modes in the feature library and annotation mode library for direct call in subsequent tasks.

[0020] The method for matching corresponding annotation patterns and annotation models from an annotation pattern library according to annotation requirements includes: Analyze the annotation requirements to determine the requirement characteristics, that is, identify and analyze the annotation requirements to determine what type of data needs to be annotated and what data to annotate, and then determine the requirement characteristics according to the data type and the characteristics of the annotated data; generate a test set according to the requirement characteristics, and the test set includes each data to be marked that meets the requirement characteristics and the corresponding correctly annotated annotation results; the user selects an annotation form that meets their requirements according to the test set or the requirement characteristics, which is regarded as a standard case; Input the requirement characteristics into the annotation pattern library for matching to obtain candidate annotation methods that meet the requirement characteristics, that is, the annotation methods for implementing the data annotation of the requirement characteristics, specifically the corresponding annotation patterns, annotation models or non-single combinations; Conduct simulation tests on each candidate annotation method through the test set, and count the corresponding annotation accuracy rate and annotation consistency rate according to the simulation test results, and directly conduct statistics according to the test results; select a correctly annotated annotation style as an annotation case; Evaluate each annotation case according to the standard case to determine whether the annotation case meets the requirements of the standard case, that is, whether the annotation form is the same, etc. If they are the same, it is regarded as meeting the requirements of the standard case, otherwise it does not meet the requirements of the standard case; based on this, establish a case evaluation model in combination with existing recognition and judgment technologies, and the expression of the case evaluation model is ; where: s is the input data, and the input data includes the standard case and the annotation case; the output data is the case evaluation value; BA is the optimization value, which is set according to the difficulty of adjusting from the annotation case to the standard case, and the value range is [0, 1). Obtain the possible adjustment work, sort each adjustment work in ascending order of difficulty, and the sorting can be based on the adjustment workload, adjustment duration, etc. In the case where no adjustment can be made, its optimization value is 0, and the others are distributed between 0 and 1. Set the corresponding optimization values according to the difficulty differences between them, and then perform matching. It can be set or verified by professionals; Input the standard case and each annotation case into the case evaluation model for analysis to obtain the corresponding case evaluation values; Calculate the corresponding annotation value according to the formula PY = b1 × PA(s) × (b2 × QA + B3 × QB); In the formula: PY is the annotation value; b1, b2, and b3 are all proportionality coefficients, and the value range is 0 < b1 ≤ 1, 0 < b2 ≤ 1, 0 < b3 ≤ 1; PA(s) is the case evaluation value output by the case evaluation model; QA is the annotation accuracy rate; QB is the annotation consistency rate.

[0021] Sort the candidate annotation methods in descending order of annotation value to obtain a first sequence, and insert corresponding annotation cases at each candidate annotation method in the first sequence; According to the candidate annotation method selected by the user in the first sequence, the selected candidate annotation method is marked as a target annotation method, and the annotation mode and the annotation model corresponding to the target annotation method are identified.

[0022] In one embodiment, if there is no matching annotation mode available in the annotation mode library, the user can only perform custom processing manually, which requires the user to have certain professional knowledge and spend a lot of effort to process, which is inefficient. Based on this, when the corresponding annotation mode and annotation model cannot be matched from the annotation mode library according to the annotation requirements, the user is assisted in establishing the corresponding annotation mode and annotation model; the detailed process is as follows: Calculate the similarity between the demand characteristics and each annotation mode, mark the annotation mode with a similarity not less than a threshold value X1 as a reference mode, identify the annotation model corresponding to the reference mode, obtain the annotation principle of the annotation model, such as how the input data is converted into output data, determine the corresponding adjustment items according to the annotation principle, that is, adjust the annotation principle to the part that needs to be adjusted when it meets the demand characteristics; adjust the annotation principle according to the adjustment items and the demand characteristics to obtain the correction principle, simulate the annotation according to the correction principle, and obtain simulation data, that is, adjust the annotation principle according to the demand characteristics to obtain the correction principle, and then perform annotation simulation according to the correction principle to form simulation data, and the simulation data is multiple; display the simulation data to the user, and the user adjusts according to the displayed simulation data, that is, adjusts it to meet the demand characteristics. The labeling effect that meets the needs is obtained; the correction principle is adjusted according to the adjustment process; the simulation data is generated again according to the adjusted correction principle, and so on, until the user does not adjust the simulation data; a training adjustment model is established based on a neural network, and a corresponding training set is established manually for training, and the training set includes input data and output data, and the input data is the original training set and the labeling principle and the correction principle of the labeling model; the output data is the training adjustment data after the original training set is adjusted according to the correction principle; the corresponding training adjustment data is obtained by analyzing the training adjustment model after successful training; in other embodiments, analysis can also be performed based on other existing technologies to determine the training adjustment data; the reference mode that cannot obtain the training adjustment data is eliminated; The annotation model is trained and adjusted through training adjustment data to obtain a new annotation model, which is marked as the initial model, and the original annotation model is still saved; the initial model is tested through the test set to calculate the corresponding annotation value; a new annotation model and annotation mode are set according to the annotation value; that is, the annotation model with the largest annotation value is selected, and then a new annotation mode is formed according to the standard model.

[0023] Step 4: Annotate the target data according to the matching annotation mode and annotation model; obtain the target annotation data; Step 5: Verify the target annotation data and make corresponding adjustments to the target annotation data based on the verification results; generally perform manual verification to ensure the accuracy and consistency of the annotation. If necessary, adjust or correct the annotation results. Export the verified target annotation data; When generating multiple labels, you can choose to save them in multiple columns or in one column. In this case, the tool will trigger conflict handling and result integration to obtain the final label result. Users can select the data range and the exported data format to save the generated labels to the storage medium.

[0024] Step 6: Continuously optimize, analyze and adjust the annotation pattern library.

[0025] As the data labeling task continues, new labeled data is collected, including original data, modified labels, and labels are regularly checked and updated to optimize existing models, create new models, or adjust labeling strategies.

[0026] When a sufficient amount of new annotated data is collected, it can be used to train new models or to fine-tune models in existing modes, and be fully trained and verified. When it is verified to be effective and ensures consistency and compatibility with the existing mode library, it is added to the newly created mode or existing mode in the mode library. The new mode needs to contain information such as the name, description, parameters, and associated models of the mode.

[0027] In one embodiment, the annotation pattern library may be continuously optimized, analyzed and adjusted in the following manner, including: The target annotation data adjusted after verification is integrated with the original data into material data, and the material data is evaluated to determine whether the adjustment is based on the original benchmark or re-annotation; that is, in order to realize the annotation of the data, whether a new annotation model or annotation mode needs to be established; according to the difference before and after the adjustment, it can be identified, therefore, the material data is divided into three classification categories, namely the first category, the second category and the third category. The first category indicates that it is sufficient to optimize and adjust the existing annotation mode or annotation model; the second category indicates that a new annotation mode needs to be established; the third category indicates that a new annotation model needs to be established.

[0028] Based on this, a corresponding material evaluation model is established. The expression of the material evaluation model is: ; In the formula: q is the input data, and the input data is the material data; Analyze each material data through the material evaluation model to determine the classification category corresponding to each material data; Classify the material data according to the classification categories corresponding to them, and obtain the first classification data, second classification data and third classification data corresponding to the first category, the second category and the third category respectively; classify the material data in the first classification data, the second classification data and the third classification data according to the annotation model or annotation mode, and obtain the optimized data of each unit; that is, classify them according to which annotation model or annotation mode they correspond to, and mark the corresponding classification category label for each unit optimized data; identify the data volume corresponding to the unit optimized data, that is, the data volume is the number of material data corresponding to the unit optimized data; when the data volume reaches the preset standard, that is, the data volume exceeds the preset quantity; perform corresponding optimization processing, such as optimizing the original annotation mode or annotation model, establishing a new original annotation mode or annotation model, and specifically setting the corresponding quantity standard according to the required data volume.

[0029] A tool for implementing the above data labeling method, the device comprising: Data import and file indexing module: This module can import and read the data to be annotated through file import, network interface, etc. It has a file type filtering function to ensure that only user-defined file types can be selected, and verify whether the data format of the selected file is correct to reduce problems caused by selecting the wrong file type. At the same time, an indexing function for files in folders has been added, allowing users to quickly switch files in the same folder and reduce the time of manual navigation. When the user selects a file in a non-specified folder, the previous / next buttons will be disabled to prevent users from accidentally operating in the wrong folder.

[0030] Visual display module: Visualize the sensor data. Introduce a hierarchical management mechanism to make the visualization of data clearer and more flexible. Use the two-dimensional coordinate axis, grid lines, etc. as the bottom layer and the data layer as the upper layer. Each layer of data can be loaded and displayed separately or simultaneously. Users can selectively load different data layers as needed. Provide customized options, such as allowing users to adjust curve color, line type, and label color and shape to meet user personalized needs. Introduce interactive functions, such as dragging, zooming, and other operations.

[0031] Feature library management module: This module allows users to create, store and call features to support data analysis and model building. Users can edit existing features or add new features to the feature library, and store features in modules.

[0032] Pattern library management module: This module allows users to define and manage multiple annotation modes, and multiple associated models or rules can be configured in each mode. Each mode contains information such as name, description, parameters and associated models. Users can add, delete and modify models. Configured modes can be directly called for use.

[0033] Statistics module: Real-time presentation of tag statistics, including tag quantity, frequency distribution, and related statistical data.

[0034] Tag processing module: For multi-type annotation, it has the functions of conflict handling and result integration. It provides a manual verification interface to allow users to modify the labeling results. It also provides one-click cancellation of all labels, undo and redo functions to help users quickly restore to the previous state when operating errors occur. Finally, the labels corresponding to the three-axis positions are generated and saved to the storage medium, supporting data export in various formats.

[0035] A storage medium stores a computer program, which implements the above-mentioned data labeling method when executed by a processor.

[0036] The introduction of a hierarchical management mechanism allows each layer of data to be loaded and displayed separately or simultaneously, and users can selectively load different data layers as needed. At the same time, customized options are provided, such as adjusting the curve color, line type, and label color and shape, to meet the personalized needs of users, increasing the plasticity and convenience of the user experience. The introduction of interactive functions, such as dragging, zooming, and other operations, further enhances the user's ability to operate and control the data, making the annotation process more flexible and intuitive. Allowing users to select existing annotation modes or create new modes and rules according to specific data annotation needs provides great flexibility. Users can fine-tune according to the characteristics of the data and the annotation goals to ensure the accuracy and effectiveness of the annotation. This flexibility makes the system more suitable for a variety of different annotation tasks.

[0037] By presetting annotation modes, data can be automatically annotated, significantly reducing the time and effort of manual annotation, thereby improving annotation efficiency. At the same time, the data visualization function enables users to intuitively observe data characteristics and annotation results, which helps to more quickly identify patterns and formulate annotation rules, as well as verify the accuracy of annotation results. Users can easily manage multiple annotation modes, including operations such as adding, deleting, modifying, and calling modes. It can store new features and new modes, and reuse existing features and annotation modes. As data annotation tasks continue, new annotation data and optimized models can continuously update and maintain the pattern library, providing more powerful support for future annotation tasks.

[0038] The above formulas are all calculated by removing dimensions and taking numerical values. The formula is a formula that is closest to the actual situation obtained by collecting a large amount of data and performing software simulation. The preset parameters and preset thresholds in the formula are set by technical personnel in this field according to actual conditions or obtained by simulating a large amount of data.

[0039] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, device, article or method. In the absence of further restrictions, an element defined by the sentence "includes a ..." does not exclude the presence of other identical elements in the process, device, article or method including the element.

[0040] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, is also included in the patent protection scope of the present invention. It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

[0041] For the convenience of description, the above devices are described in various units according to their functions. Of course, when implementing the present application, the functions of each unit can be implemented in one or more software and / or hardware. It should be understood by those skilled in the art that the embodiments of the present invention can be provided as methods, systems or computer program products. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0042] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems) and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 The functionality specified in one or more boxes.

[0043] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A function specified in a process or multiple boxes.

[0044] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

Claims

1. A sensor data annotation method based on wearable devices, characterized in that: The method includes: Obtain the data of the file to be annotated, preprocess the received data of the file to be annotated, and mark the preprocessed data of the file to be annotated as target data; Set the corresponding display mode according to the user's requirements, and visually display the read data according to the display mode; Establish an annotation mode library and a feature library. The annotation mode library is used to store various annotation modes and the corresponding annotation models for each annotation mode; the feature library is used to store various annotation features; obtain the user's annotation requirements, and match the corresponding annotation mode and annotation model from the annotation mode library according to the annotation requirements; Annotate the target data according to the matched annotation mode and annotation model; obtain the target annotation data; Verify the target annotation data, and make corresponding adjustments to the target annotation data according to the verification result; export the verified target annotation data; Continuously optimize, analyze and adjust the annotation mode library.

2. The sensor data annotation method based on wearable devices according to claim 1 is characterized in that: The method for matching the corresponding annotation mode and annotation model from the annotation mode library according to the annotation requirements includes: Analyze the annotation requirements to determine the requirement features; generate a test set according to the requirement features, and set a standard case according to the test set; Input the requirement features into the annotation mode library for matching to obtain each candidate annotation method that meets the requirement features; Conduct simulation tests on each candidate annotation method through the test set to obtain the annotation accuracy rate and annotation consistency rate corresponding to each candidate annotation method; and set the annotation cases corresponding to each candidate annotation method; Evaluate each annotation case according to the standard case to obtain the case evaluation value corresponding to each candidate annotation method; Calculate the corresponding annotation value according to the formula PY = b1 × PA(s) × (b2 × QA + B3 × QB); In the formula: PY is the annotation value; b1, b2, and b3 are all proportionality coefficients, and the value range is 0 < b1 ≤ 1, 0 < b2 ≤ 1, 0 < b3 ≤ 1; PA(s) is the case evaluation value output by the case evaluation model; QA is the annotation accuracy rate; QB is the annotation consistency rate; Sort each candidate annotation method in descending order of the annotation value to obtain the first sequence, and insert the corresponding annotation cases at each candidate annotation method in the first sequence; Determine the target annotation method according to the first sequence, and identify the annotation mode and annotation model corresponding to the target annotation method.

3. The sensor data annotation method based on wearable devices according to claim 2 is characterized in that: The method for evaluating each annotation case according to the standard case includes: Establish a case evaluation model, and the expression of the case evaluation model is: ； In the formula: s is the input data, and the input data includes the standard case and the annotation case; the output data is the case evaluation value; BA is the optimization value, and the value range of the optimization value is [0, 1); Input the standard case and each annotation case into the case evaluation model for analysis to obtain the corresponding case evaluation value.

4. The sensor data annotation method based on wearable devices according to claim 2 is characterized in that: When the corresponding annotation mode and annotation model cannot be matched from the annotation mode library according to the annotation requirements, assist the user in establishing the corresponding annotation mode and annotation model.

5. The sensor data annotation method based on wearable devices according to claim 4 is characterized in that: The method for assisting the user in establishing the corresponding annotation mode and annotation model includes: Step SA1: Calculate the similarity between the demand characteristics and each annotation mode, mark the annotation mode with a similarity not less than the threshold value X1 as the reference mode, identify the annotation model corresponding to the reference mode, and obtain the annotation principle of the annotation model; determine the corresponding adjustment item according to the annotation principle, adjust the annotation principle according to the adjustment item and the demand characteristics, and obtain the correction principle; Step SA2: perform labeling simulation according to the correction principle to form multiple simulation data; display the simulation data to the user, the user makes adjustments according to the displayed simulation data, and adjusts the correction principle according to the adjustment result of the simulation data to obtain a new correction principle; Step SA3: loop step SA2 until there is no new correction principle; Step SA4: Establish a training adjustment model, obtain a training set corresponding to the annotation model, analyze the annotation principle, the training set and the correction principle through the training adjustment model, and obtain training adjustment data; when no training adjustment data is obtained, eliminate the corresponding reference model; Step SA5: Set a new annotation model and annotation mode that meet the required characteristics based on the training adjustment data and annotation model.

6. The sensor data annotation method based on wearable devices according to claim 1, characterized in that: The methods for continuous optimization, analysis and adjustment of the annotation pattern library include: Acquire material data in real time, establish a material evaluation model, analyze each material data through the material evaluation model, and determine the classification category corresponding to each material data; Classify the material data according to the classification categories corresponding to them to obtain corresponding first classification data, second classification data and third classification data; classify the first classification data, second classification data and third classification data again to obtain optimized data of each unit; and label the optimized data of each unit with corresponding classification category labels; The recognition unit optimizes the data volume corresponding to the data, and performs corresponding optimization processing when the data volume reaches a preset standard.

7. The sensor data annotation method based on wearable devices according to claim 6 is characterized in that: The expression of the material evaluation model is: ； In the formula: q is the input data, the input data is the material data; the output data is the category value.

8. A sensor data annotation tool based on wearable devices, characterized by: A sensor data annotation method based on a wearable device according to any one of claims 1 to 7 is implemented, comprising a data import and file indexing module, a visualization display module, a feature library management module, a pattern library management module, a statistics module and a label processing module; The data import and file indexing module is used to obtain the data to be annotated and pre-process the data to be annotated; The visual display module is used to visually display the read sensor data according to a preset hierarchical management method; The feature library management module is used to manage the feature library; The pattern library management module is used to manage the pattern library; The statistical module is used to display the statistical information of the tags in real time; The label processing module is used to process multiple types of annotations, set up a manual verification interface, and the user processes the label results based on the manual verification interface; and exports the corresponding data according to a preset format.

9. A storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by a processor, the sensor data labeling method based on a wearable device is implemented as described in any one of claims 1 to 7.

Images