A data analysis based wearable device monitoring method

Abstract

The present application relates to the field of data analysis, and more particularly to a kind of data analysis-based wearable device monitoring method, comprising, periodically for each target monitoring data of target monitoring personnel Acquisition, determine key analysis data, and obtain the key monitoring video corresponding to each key analysis data;According to the analysis strategy of corresponding key analysis data determined by abnormal coincidence coefficient;If abnormal coincidence coefficient is greater than preset abnormal coincidence coefficient, the data change period of each type of regional data is detected;If abnormal coincidence coefficient is less than or equal to preset abnormal coincidence coefficient, the associated personnel search is carried out for key monitoring video, and the video information tracing mode is determined according to the interaction time length proportion of target monitoring personnel in key monitoring video and the number of associated personnel;Under the condition of information analysis completion, the early warning information of each key analysis data is sent to user, and the effectiveness of early warning information is improved.

Description

Technical Field

[0001] This invention relates to the field of data analysis, and more particularly to a method for monitoring wearable devices based on data analysis. Background Technology

[0002] Real-time monitoring of various vital signs is necessary to obtain abnormal physical data as promptly as possible. However, relying solely on the acquired vital sign data cannot effectively determine the cause of the abnormality. Judging the patient's condition based solely on vital sign data is prone to error and cannot effectively trace the reason for the patient's abnormal state. Collaborative monitoring using acquired videos can make an effective judgment on the patient's condition. However, existing wearable device monitoring methods often only extract monitoring videos based on the time corresponding to the abnormal state, which cannot guarantee the validity of the acquired monitoring videos. Therefore, how to obtain effective video segments based on the acquired data and video information is a problem that urgently needs to be solved by those skilled in the art.

[0003] Chinese Patent Publication No. CN117936073A discloses a remote real-time online monitoring and management system for human health information. The system includes a patient information acquisition module for acquiring the patient's medical information; a wearable data acquisition module for collecting vital sign data; a primary warning module for identifying abnormal patient conditions based on sensor data and generating a primary alarm message; a warning correction module for correcting the primary alarm message using behavioral patterns obtained from video data corresponding to the wearable data; an alarm module for sending notifications to relevant personnel based on the corrected alarm message; and a suggestion module for matching emergency response plans based on the alarm message and sending them to the patient and relevant personnel. However, this solution has the following problems: it fails to extract the effective portion of the video information based on the actual scene the patient is in, resulting in low effectiveness of the acquired video information and consequently low accuracy in judging abnormal data. Summary of the Invention

[0004] To address this, the present invention provides a wearable device monitoring method based on data analysis, which overcomes the problem in the prior art that fails to extract the effective portion of the video information based on the actual scene in which the patient is located within the acquired video information, resulting in low effectiveness of the acquired video information and consequently low accuracy in judging abnormal data.

[0005] To achieve the above objectives, the present invention provides a wearable device monitoring method based on data analysis, comprising:

[0006] Periodically collect various target monitoring data from target monitoring personnel, record target monitoring data whose monitoring values ​​are not within the corresponding preset data range or whose data change coefficient is greater than the preset data change coefficient as key analysis data, and obtain the key monitoring video corresponding to each key analysis data;

[0007] The analysis strategy for determining the corresponding key analytical data is based on the abnormal overlap coefficient.

[0008] If the abnormal overlap coefficient is greater than the preset abnormal overlap coefficient, the data change cycle of various regional data will be detected.

[0009] If the abnormal overlap coefficient is less than or equal to the preset abnormal overlap coefficient, a search for associated personnel is conducted for the key monitoring video. The video information tracing method is determined based on the proportion of interaction time of the target monitoring personnel in the key monitoring video and the number of associated personnel.

[0010] The video information tracing method includes extracting video segments corresponding to each associated person in the key monitoring video and setting an early warning priority coefficient, as well as dividing the key monitoring video and determining the video extraction method based on the historical abnormal repetition coefficient.

[0011] Once the information analysis is complete, early warning information for each key analysis data point will be sent to the user.

[0012] Furthermore, under key analysis conditions, the abnormal overlap coefficients of various key analysis data of the target monitoring personnel are detected, and the corresponding analysis strategy for key analysis data is determined based on the abnormal overlap coefficients.

[0013] The anomaly overlap coefficient of key analysis data is determined based on the reference anomaly difference among monitoring personnel in the relevant area;

[0014] The key analytical condition is that key analytical data exists in the target monitoring data of the target monitoring personnel.

[0015] Furthermore, for any key analytical data, if the abnormal overlap coefficient is greater than the preset abnormal overlap coefficient, various regional data of that key analytical data are obtained to determine the data change cycle of the corresponding regional data.

[0016] The warning priority coefficient for each region's data is set based on the correlation of abnormal changes in various regional data.

[0017] Furthermore, for any key analysis data, if the abnormal overlap coefficient is less than or equal to the preset abnormal overlap coefficient, the relevant personnel search is performed on the key monitoring video corresponding to the key analysis data, and the interactive video segments corresponding to each relevant personnel in the key monitoring video are extracted. The video information tracing method is determined based on the proportion of the interaction time of the target monitoring personnel in the key monitoring video and the number of relevant personnel.

[0018] The associated personnel corresponding to key monitoring videos are determined based on the personnel correlation coefficient between each monitoring personnel and the target monitoring personnel within the video monitoring period. Monitoring personnel whose personnel correlation coefficient is greater than the preset personnel correlation coefficient are recorded as associated personnel.

[0019] Furthermore, for a single key monitoring video, if the interaction time percentage is greater than the preset interaction time percentage or the number of associated personnel is greater than the preset number of associated personnel, the warning priority coefficient for the corresponding interactive video segment is set according to the interaction correlation coefficient between each interactive video segment and the corresponding key analysis data.

[0020] The warning priority coefficient and the interaction correlation coefficient are positively correlated.

[0021] Furthermore, for a single key monitoring video, if the interaction time percentage is less than or equal to the preset interaction time percentage and the number of associated personnel is less than or equal to the preset number of associated personnel, the key monitoring video is divided to obtain several video segments to be analyzed. The historical abnormal repetition coefficient of the key analysis data corresponding to the video segments to be analyzed is detected, so as to determine the video extraction strategy based on the historical abnormal repetition coefficient.

[0022] Furthermore, if the historical abnormal repetition coefficient is greater than the preset historical abnormal repetition coefficient, the warning priority coefficient is set according to the feature overlap of each video segment to be analyzed and the corresponding feature correlation coefficient.

[0023] The warning priority coefficient is positively correlated with the feature reference value.

[0024] Furthermore, if the historical abnormal repetition coefficient is less than or equal to the preset historical abnormal repetition coefficient, the key reference frame of the key monitoring video is set according to the relevant time of the key analysis data corresponding to the video segment to be analyzed, and the setting method is determined according to the distribution status of the key reference frame.

[0025] If there are key reference frames in a dense state, the adjacent video segments to be analyzed with key reference frames in a dense state are recorded as a warning video segment, and the warning priority coefficient of the warning video segment is determined according to the number of key reference frames contained in the warning video segment.

[0026] The warning priority coefficient is positively correlated with the number of key reference frames.

[0027] Furthermore, behavioral features are extracted for the video segments to be analyzed corresponding to each key reference frame that is in a scattered state, and the warning priority coefficient for the video segments to be analyzed is set according to the matching coefficient between each behavioral feature and the corresponding key reference frame.

[0028] The warning priority coefficient and the matching coefficient are positively correlated.

[0029] Furthermore, once the information analysis is complete, warning information for each key analysis data is sent to the user based on the regional data and the warning priority coefficient of the video segment.

[0030] The information analysis is completed when all key analysis data have been assigned a regional data warning priority coefficient, or when all corresponding key monitoring videos have been assigned a video segment warning priority coefficient.

[0031] Compared with the prior art, the beneficial effect of the present invention is that, in the technical solution of the present invention, a targeted and effective information extraction method is selected based on the abnormal overlap coefficient of key analysis data and the actual situation of the personnel associated in the key monitoring video. This makes the information extraction method more in line with the actual monitoring scenario, thereby ensuring the effectiveness of the early warning information extracted from the regional data of key analysis data and the collected video information. It avoids the inefficiency of information transmission and judgment of abnormal status of target monitoring personnel caused by too much invalid information in the sent early warning information.

[0032] Furthermore, in this invention, the corresponding analysis strategy is determined based on the abnormal overlap coefficient of each key analysis data. When the abnormal overlap coefficient of a certain key analysis data is large, it indicates that there are many monitoring personnel in the relevant area of ​​the target monitoring personnel with similar abnormal situations in the key analysis data. Therefore, various data in the area where the personnel are located are collected, and the area data most related to the abnormal state of the data is selected for early warning based on the correlation of regional anomalies. This invention improves the effectiveness of early warning information.

[0033] Furthermore, in this invention, when the abnormal overlap coefficient of a certain key analysis data is small, it indicates that there are no or few monitoring personnel in the relevant area of ​​the target monitoring personnel and there are similar abnormalities in the key analysis data. Therefore, specific analysis is performed on the key monitoring video, and the video information tracing method is determined according to the proportion of interaction time of the target monitoring personnel in the key monitoring video and the number of related personnel. This ensures that the method of extracting effective video segments is more in line with the actual situation of the target monitoring personnel, thereby ensuring the effectiveness of the extracted video segments.

[0034] Furthermore, in this invention, when the interaction duration percentage is greater than the preset interaction duration percentage or the number of associated personnel is greater than the preset number of associated personnel, the video segments corresponding to each associated person in the key monitoring video are analyzed, and the warning priority coefficients for the corresponding video segments are set according to the time correlation coefficient and the behavior correlation coefficient. When there are many interactions between the target monitoring personnel, the video segments corresponding to the associated personnel are analyzed to determine the video segments that should be given priority for warning. This can improve the efficiency of video analysis while ensuring the effectiveness of the video segments that should be given priority for warning.

[0035] Furthermore, in this invention, when the percentage of interaction time is less than or equal to a preset percentage of interaction time and the number of associated personnel is less than or equal to a preset number of associated personnel, the video extraction strategy is determined based on the historical abnormal repetition coefficient of the key analysis data. When the historical abnormal repetition coefficient is large, that is, when the abnormal state of this key analysis data occurs frequently and has regularity, the warning priority coefficient of the video segment to be analyzed is set according to the feature overlap and the corresponding feature correlation coefficient. When the historical abnormal repetition coefficient is small, that is, when this key analysis data does not have regularity, the settings are made for each video segment to be analyzed according to the specific distribution of the key reference frames, so that the method of extracting effective video segments is more in line with the actual situation, improving the efficiency of video analysis while ensuring the effectiveness of the video segments with priority warning. Attached Figure Description

[0036] Figure 1 This is a schematic diagram of the wearable device monitoring method based on data analysis according to the present invention;

[0037] Figure 2 This is a flowchart illustrating the analysis strategy determined based on the abnormal overlap coefficient in this invention.

[0038] Figure 3 This is a flowchart illustrating how the present invention determines the video information tracing method based on the percentage of interaction time of target monitoring personnel within key monitoring videos and the number of associated personnel;

[0039] Figure 4 This is a flowchart illustrating the video extraction strategy determined by the present invention based on historical abnormal repetition coefficients. Detailed Implementation

[0040] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0041] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0042] It should be noted that in the description of this invention, the terms "upper", "lower", "left", "right", "inner", "outer", etc., which indicate directions or positional relationships, are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and is not intended to indicate or imply that the device or element must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation of this invention.

[0043] Furthermore, it should be noted that, in the description of this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0044] Please see Figures 1 to 4 As shown, the present invention provides a wearable device monitoring method based on data analysis, comprising:

[0045] Periodically collect various target monitoring data from target monitoring personnel, record target monitoring data whose monitoring values ​​are not within the corresponding preset data range or whose data change coefficient is greater than the preset data change coefficient as key analysis data, and obtain the key monitoring video corresponding to each key analysis data;

[0046] The analysis strategy for determining the corresponding key analytical data is based on the abnormal overlap coefficient.

[0047] If the abnormal overlap coefficient is greater than the preset abnormal overlap coefficient, regional abnormality period detection is performed for various types of regional data.

[0048] If the abnormal overlap coefficient is less than or equal to the preset abnormal overlap coefficient, a search for associated personnel is conducted for the key monitoring video. The video information tracing method is determined based on the proportion of interaction time of the target monitoring personnel in the key monitoring video and the number of associated personnel.

[0049] The video information tracing method includes extracting video segments corresponding to each associated person in the key monitoring video and setting an early warning priority coefficient, as well as dividing the key monitoring video and determining the video extraction method based on the historical abnormal repetition coefficient.

[0050] Once the information analysis is complete, a set of early warning messages for each key analysis data point is sent to the user.

[0051] The target monitoring personnel are those currently analyzing various data and video information. These personnel use wearable devices to monitor the target data. The types of target monitoring data in this invention include, but are not limited to: blood pressure, heart rate, electrocardiogram, blood oxygen saturation, body temperature, heart rate variability, electromyography signals, blood volume pulse waves, and blood glucose. Furthermore, the wearable devices used in this invention must be able to acquire various regional data and monitoring video data of the target monitoring personnel's location. The categories of regional data include, but are not limited to, temperature, air pressure, and light intensity. This invention does not specifically limit the model of the wearable device used; users can adaptively set the specific model of the wearable device according to the actual scenario. The selected wearable device only needs to be able to acquire the target monitoring data, regional data, and monitoring video data required by the user.

[0052] This invention utilizes several historical monitoring records. Each historical monitoring record contains basic information about the monitoring personnel (such as age, gender, and medical history), the data change coefficient of the target monitoring data, the monitoring data difference value, the abnormal overlap coefficient, the personnel correlation coefficient, the historical repetition coefficient, and the feature similarity from at least one monitoring record for the monitoring personnel. Furthermore, each historical monitoring record has a corresponding qualified mark, which records whether the accuracy of the abnormality judgment result meets the user's requirements.

[0053] This invention employs a cyclical data monitoring period, the duration of which can be determined by the user. The higher the user's requirement for the timeliness of the monitoring results from the target personnel, the shorter the data monitoring period. One data monitoring period is provided, with a duration of 10 seconds. At the end of each data monitoring period, the monitoring data for each target are detected to obtain the corresponding monitoring values. Each target monitoring data item in this invention has a corresponding preset data range. The maximum and minimum values ​​of the preset data range can be set by the user based on reference records. These reference records are historical monitoring records of personnel with the same age, gender, and medical history as the target personnel. For single-target monitoring data, the maximum value of the monitoring value in historical monitoring records without regional data and video segment extraction is recorded as the maximum value of the preset data range, and the minimum value is recorded as the minimum value of the preset data range. How to set the preset data range is easily understood by those skilled in the art and will not be elaborated here.

[0054] For any target monitoring data, the data variation coefficient is the sum of the monitoring data fluctuation value and the monitoring data difference value. The monitoring data difference value is the absolute value of the difference between the monitoring value obtained in the current data detection period and the previous data detection period. The monitoring data fluctuation value = the sum of the absolute values ​​of the differences between each monitoring value and the average monitoring value within the fluctuation detection period / the number of times the monitoring data for this target is detected within the fluctuation detection period. The average monitoring value is the average value of each monitoring data within the fluctuation detection period. The duration of the fluctuation detection period can be determined by the user according to the actual work scenario. The user's monitoring results for the target monitoring personnel... The higher the accuracy requirement, the longer the fluctuation detection period. A fluctuation detection period of 5 minutes is provided. The value of the preset data change coefficient can be determined by the user based on the actual work scenario. For example, the user can set it based on historical monitoring records. The higher the accuracy requirement for the monitoring results of the target monitoring personnel, the smaller the preset data change coefficient. A method for determining the preset data change coefficient is provided, where the maximum value of the data change coefficient in historical monitoring records without regional data and video segment extraction is recorded as the preset data change coefficient.

[0055] For a single key monitoring data point, key monitoring videos are obtained based on key anomaly moments. If the monitored value is not within the corresponding preset data range, the moment when the monitored value is detected is recorded as the key anomaly moment. If the data change coefficient is greater than the preset data change coefficient, the first moment when the monitored data difference value is detected to be greater than the preset monitored data difference value is recorded as the key anomaly moment. The key anomaly moment is set as the cutoff moment for the key monitoring video, and the key monitoring video is obtained according to the preset video extraction duration. The values ​​of the preset monitored data difference value and the preset video extraction duration can be determined by the user according to the actual work scenario. For example, the user can set them based on historical monitoring records. The higher the user's requirement for the accuracy of the monitoring results of the target monitoring personnel, the smaller the value of the preset monitored data difference value and the larger the value of the preset video extraction duration. A method for determining the value of the preset monitored data difference value is provided, which records the maximum value of the monitored data difference value in historical monitoring records where no regional data and video segment extraction have been performed as the preset monitored data difference value. A method for determining the value of the preset video extraction duration is provided, which records the average duration of each key monitoring video in historical monitoring records that meet the user's accuracy requirements for the monitoring results of the target monitoring personnel as the preset video extraction duration.

[0056] Specifically, under key analysis conditions, the abnormal overlap coefficients of various key analysis data of the target monitoring personnel are detected, and the corresponding analysis strategy for key analysis data is determined based on the abnormal overlap coefficients.

[0057] The anomaly overlap coefficient of key analysis data is determined based on the reference anomaly difference among monitoring personnel in the relevant area;

[0058] The key analytical condition is that key analytical data exists in the target monitoring data of the target monitoring personnel.

[0059] Wherein, for any key analysis data, the anomaly overlap coefficient is the average of the reference anomaly differences of each monitoring person in the relevant area. For a single monitoring person, the reference anomaly difference is the sum of the products of the monitoring difference value, the anomaly time difference value, and the corresponding influence coefficient. The monitoring difference value is the absolute value of the difference between the monitoring value of the key analysis data of the monitoring person and the target monitoring person. The duration difference value is the interval between the key anomaly time of the key analysis data of the monitoring person and the target monitoring person. The relevant area is the set of locations of monitoring persons within a building space whose regional data are consistent with those obtained by the target monitoring person.

[0060] Specifically, for any key analytical data, if the abnormal overlap coefficient is greater than the preset abnormal overlap coefficient, various regional data of the key analytical data are obtained to determine the data change cycle of the corresponding regional data.

[0061] The warning priority coefficient for each region's data is set based on the correlation of abnormal changes in various regional data.

[0062] The value of the preset abnormal overlap coefficient can be determined by the user according to the actual working scenario. For example, the user can set it according to historical monitoring records. A method for determining the value of the preset abnormal overlap coefficient is provided, in which historical monitoring records based on regional data as early warning information are recorded as overlap coefficient reference records, and the minimum value of the preset abnormal overlap coefficient in the overlap coefficient reference records that meets the user's requirements for the accuracy of the monitoring results of the target monitoring personnel is recorded as the preset abnormal overlap coefficient.

[0063] This invention employs a cyclical regional data monitoring cycle. The duration of the regional data monitoring cycle can be determined by the user. The higher the user's requirement for the timeliness of the monitoring results from the target monitoring personnel, the shorter the duration of the regional data monitoring cycle. One regional data monitoring cycle duration is provided, which is 1 minute. At the end of each regional data monitoring cycle, the values ​​of various types of regional data are detected. For any category of regional data, the starting time of the data change cycle is the moment when the first regional data difference value is detected to be greater than the preset regional data difference value within the time period corresponding to the key monitoring cycle. The regional data difference value is the absolute value of the difference between the values ​​obtained in the current regional data detection cycle and the previous regional data detection cycle. The abnormal change correlation = the overlap time between the data change cycle and the abnormal monitoring cycle / the duration of the key monitoring video.

[0064] Specifically, for any key analysis data, if the abnormal overlap coefficient is less than or equal to the preset abnormal overlap coefficient, the relevant personnel search is performed on the key monitoring video corresponding to the key analysis data, and the interactive video segments corresponding to each relevant personnel in the key monitoring video are extracted. The video information tracing method is determined based on the proportion of the interaction time of the target monitoring personnel in the key monitoring video and the number of relevant personnel.

[0065] The associated personnel corresponding to key monitoring videos are determined based on the personnel correlation coefficient between each monitoring personnel and the target monitoring personnel within the video monitoring period. Monitoring personnel whose personnel correlation coefficient is greater than the preset personnel correlation coefficient are recorded as associated personnel.

[0066] The personnel correlation coefficient is determined based on the influence parameters corresponding to the interaction characteristics between the monitoring personnel and the target monitoring personnel, as well as the influence parameters corresponding to the existing correlation relationships. The categories of interaction characteristics include, but are not limited to, drug distribution, drug injection, and food distribution. The categories of correlation relationships include, but are not limited to, doctor-patient relationships and nursing relationships. Users can set the values ​​of the influence parameters corresponding to the interaction characteristics and correlation relationships according to the actual work scenario. This invention provides interaction characteristics and correlation relationships for determining the personnel correlation coefficient. The interaction characteristics include drug distribution, drug injection, and food distribution; the correlation relationships include doctor-patient relationships and nursing relationships. It provides values ​​for the influence parameters corresponding to drug distribution, drug injection, and food distribution: the influence parameter for drug distribution is 0.3; the influence parameter for drug injection is 0.3; the influence parameter for food distribution is 0.4; and it provides values ​​for the influence parameters corresponding to doctor-patient relationships and nursing relationships: the influence parameter for doctor-patient relationships is 0.5; the influence parameter for nursing relationships is 0.5.

[0067] The value of the preset personnel correlation coefficient can be determined by the user based on the actual work scenario. For example, the user can set it based on historical monitoring records. The higher the user's requirement for the accuracy of the monitoring results of the target monitoring personnel, the larger the value of the preset personnel correlation coefficient. A method for determining the value of the preset personnel correlation coefficient is provided, which is the average value of the personnel correlation coefficients of each associated person in the historical monitoring records that meet the user's accuracy requirements for the monitoring results of the target monitoring personnel. In this invention, the video segment is a portion of the video extracted from the key monitoring video. The interaction duration percentage = the sum of the video durations corresponding to each interaction video segment / the video duration of the key monitoring video. For a single associated person, the start time of the corresponding interaction video segment is the moment when the associated person first appears in the key monitoring video, and the end time of the interaction video is the moment when the associated person last appears. How to extract the interaction video segment for each associated person is a content that is easy for those skilled in the art to understand, and will not be elaborated here.

[0068] Specifically, for a single key monitoring video, if the interaction time percentage is greater than the preset interaction time percentage or the number of associated personnel is greater than the preset number of associated personnel, the warning priority coefficient for the corresponding interactive video segment is set according to the interaction correlation coefficient between each interactive video segment and the corresponding key analysis data.

[0069] The warning priority coefficient and the interaction correlation coefficient are positively correlated.

[0070] The user can determine the values ​​of the preset interaction duration percentage and the preset number of associated personnel based on the actual work scenario. For example, the user can set them based on historical monitoring records. One method for determining the preset interaction duration percentage is to record the historical monitoring records that are set according to the warning priority coefficient of the corresponding video segment based on the time correlation coefficient and the behavior correlation coefficient as the interaction reference record. The minimum value of the interaction duration percentage of each key monitoring video in the interaction reference record that meets the user's requirements for the accuracy of the monitoring results of the target monitoring personnel is recorded as the preset interaction duration percentage. Another method for determining the preset number of associated personnel is to record the minimum value of the number of associated personnel in each key monitoring video in the interaction reference record that meets the user's requirements for the accuracy of the monitoring results of the target monitoring personnel as the preset number of associated personnel.

[0071] The interaction correlation coefficient is the sum of the products of each relevant interaction behavior and its corresponding time matching coefficient. Behavioral features are extracted for each interactive video segment to determine the category of the interaction behavior. For a key analysis data point, if a certain interaction behavior identified in the interactive video segment exists in the anomaly causes in its historical monitoring records, that behavioral feature is recorded as a relevant interaction behavior. The time matching coefficient corresponding to each relevant interaction behavior is negatively correlated with the behavior time difference value. The behavior time difference value is the absolute value of the difference between the relevant interval duration and the reference interval duration. The relevant interval duration is the interval duration between the time corresponding to the relevant interaction behavior and the key anomaly time. The reference interval duration is the average interval duration between the time corresponding to the relevant interaction behavior and the key anomaly time in each historical monitoring record.

[0072] Specifically, for a single key monitoring video, if the interaction time percentage is less than or equal to the preset interaction time percentage and the number of associated personnel is less than or equal to the preset number of associated personnel, the key monitoring video is divided to obtain several video segments to be analyzed. The historical abnormal repetition coefficient of the key analysis data corresponding to the video segments to be analyzed is detected, so as to determine the video extraction strategy based on the historical abnormal repetition coefficient.

[0073] The system divides key monitoring videos into several video segments of equal length for analysis. Users can determine the number of segments based on their actual work scenario. For example, users can set the number of segments based on historical monitoring records. The higher the accuracy requirements for the monitoring results of the target personnel, the larger the number of video segments to be analyzed. One possible value for the number of video segments is 10. For single-item target monitoring data, the historical anomaly repetition coefficient is calculated as: the number of times the target monitoring data shows anomalies / the total number of data monitoring cycles. If the target monitoring data shows anomalies, it is recorded as key analysis data. The total number of data monitoring cycles is the number of data monitoring cycles that exist from the first monitoring value of the target monitoring personnel when the data was first obtained in the historical monitoring records until the current moment.

[0074] Specifically, if the historical abnormal repetition coefficient is greater than the preset historical abnormal repetition coefficient, the warning priority coefficient is set according to the feature overlap of each video segment to be analyzed and the corresponding feature correlation coefficient.

[0075] The warning priority coefficient is positively correlated with the feature reference value.

[0076] The value of the preset historical anomaly repetition coefficient can be determined by the user according to the actual work scenario. For example, the user can set it according to historical monitoring records. A method for setting the value of the preset historical anomaly repetition coefficient is provided, in which historical monitoring records that are set according to feature overlap and corresponding feature correlation coefficient for early warning priority coefficient are recorded as repetition reference records, and the minimum value of the historical repetition coefficient in the repetition reference records that meets the user's requirements for the accuracy of the monitoring results of the target monitoring personnel is recorded as the preset historical anomaly repetition coefficient.

[0077] For a single key monitoring data point, historical monitoring records contain video segments identified as being associated with the data anomaly at each occurrence, along with corresponding anomaly causes. Behavioral characteristics of the target monitoring personnel and their associated personnel are extracted. For a single video segment to be analyzed, the feature reference value is the sum of the products of each feature overlap and its corresponding feature correlation coefficient. The feature overlap is defined as the number of overlapping behavioral features divided by the number of behavioral features extracted from the video segment to be analyzed. Overlapping behavioral features are those whose feature similarity to any behavioral feature in any video segment associated with the data anomaly is greater than a preset feature similarity. How to extract behavioral features from video segments and how to determine the feature similarity between behavioral features are easily understood by those skilled in the art, and the specific execution process is not limited here and will not be elaborated upon. The preset feature similarity... The value can be determined by the user based on the actual work scenario. For example, the user can set it based on historical monitoring records. The higher the user's requirement for the accuracy of the monitoring results of the target monitoring personnel, the larger the value of the preset feature similarity. A method for setting the preset feature similarity is provided, which is the average value of the feature similarity of each overlapping behavioral feature in the historical monitoring records that meets the user's accuracy requirements for the monitoring results of the target monitoring personnel. For a single feature overlap, it is determined based on the proportion of abnormal causes corresponding to the video segment that determines the feature overlap in the historical monitoring records. The feature correlation coefficient corresponding to the feature overlap is positively correlated with the proportion of abnormal causes in the historical monitoring records. The proportion of abnormal causes = the number of times the abnormal cause corresponding to the video segment that determines the feature overlap occurs in each historical monitoring record / the number of times the target monitoring data corresponding to the video segment that determines the feature overlap is recorded as key analysis data.

[0078] Specifically, if the historical abnormal repetition coefficient is less than or equal to the preset historical abnormal repetition coefficient, the key reference frame of the key monitoring video is set according to the relevant time of the key analysis data corresponding to the video segment to be analyzed, and the setting method is determined according to the distribution status of the key reference frame.

[0079] If there are key reference frames in a dense state, the adjacent video segments to be analyzed with key reference frames in a dense state are recorded as a warning video segment, and the warning priority coefficient of the warning video segment is determined according to the number of key reference frames contained in the warning video segment.

[0080] The warning priority coefficient is positively correlated with the number of key reference frames.

[0081] Specifically, for any key analysis data, the start time of the finally determined video segment in each historical monitoring record is recorded as the relevant time. The video frames corresponding to each relevant time in the key monitoring video corresponding to the key analysis data are recorded as key reference frames. The distribution state of the key reference frames is determined according to the interval between each key reference frame and its adjacent key reference frames. The distribution state includes a dense state and a discrete state. If the interval between a key reference frame and its adjacent key reference frames is less than or equal to a preset interval, the key reference frame and its adjacent key reference frames are determined to be in a dense state. If the interval between a key reference frame and its adjacent key reference frames is less than or equal to the preset interval, the key reference frame is determined to be in a discrete state. For a single key reference frame, key reference frames that have no other key reference frames between them are recorded as adjacent key reference frames. The value of the preset interval can be determined by the user according to the actual working scenario. One preset interval value is provided, which is 5% of the duration of the key monitoring video in which the key reference frame is located.

[0082] Specifically, behavioral features are extracted for the video segments to be analyzed corresponding to each key reference frame that is in a scattered state, and the warning priority coefficient for the video segments to be analyzed is set according to the matching coefficient between each behavioral feature and the corresponding key reference frame.

[0083] The warning priority coefficient and the matching coefficient are positively correlated.

[0084] Specifically, for a single video segment to be analyzed, the matching coefficient is the sum of the matching reference values ​​of the video segment to be analyzed and the video segments corresponding to each key reference frame. For any video segment corresponding to a key reference frame, the matching reference value = the number of overlapping behavioral features of the video segment to be analyzed / the number of behavioral features extracted from the video segments corresponding to the key reference frames.

[0085] Specifically, once the information analysis is complete, warning information for each key analysis data is sent to the user based on the regional data and the warning priority coefficient of the video segment.

[0086] The information analysis is completed when all key analysis data have been assigned a regional data warning priority coefficient, or when all corresponding key monitoring videos have been assigned a video segment warning priority coefficient.

[0087] Specifically, warning information is sent to users based on the warning priority coefficients of data from each region, each warning video segment, and each video segment to be analyzed. The higher the warning priority coefficient of the region data, the warning video segment, and the video segment to be analyzed, the earlier the warning information is sent to the user.

[0088] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

[0089] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A data analysis based wearable device monitoring method, characterized by, The method comprises the following steps: Periodically monitoring the target monitoring data of the target monitoring personnel, recording the target monitoring data whose monitoring value is not in the corresponding preset data range or whose data change coefficient is greater than the preset data change coefficient as key analysis data, and obtaining the key monitoring video corresponding to each key analysis data; According to the abnormal coincidence coefficient, determine the analysis strategy of the corresponding key analysis data; If the abnormal coincidence coefficient is greater than the preset abnormal coincidence coefficient, detect the data change period of each type of regional data; If the abnormal coincidence coefficient is less than or equal to the preset abnormal coincidence coefficient, search for associated personnel in the key monitoring video, and determine the video information tracing method according to the interaction time length proportion of the target monitoring personnel in the key monitoring video and the number of associated personnel; The video information tracing method comprises extracting each video segment corresponding to each associated personnel in the key monitoring video and setting a warning priority coefficient, and dividing the key monitoring video and determining the video extraction method according to the historical abnormal repetition coefficient; Under the condition of information analysis, send the warning information of each key analysis data to the user; For any one key analysis data, the abnormal coincidence coefficient is the average value of the reference abnormal difference of each monitoring personnel in the related area, and for a single monitoring personnel, the reference abnormal difference is the sum of the monitoring difference value and the abnormal time difference value multiplied by the corresponding influence coefficient.

2. The data analysis based wearable device monitoring method of claim 1, wherein, Under the key analysis condition, detect the abnormal coincidence coefficient of each key analysis data of the target monitoring personnel, and determine the analysis strategy of the corresponding key analysis data according to the abnormal coincidence coefficient; The abnormal coincidence coefficient of the key analysis data is determined according to the reference abnormal difference of each monitoring personnel in the related area; The key analysis condition is that there is key analysis data in each target monitoring data of the target monitoring personnel. 3.The data analysis based wearable device monitoring method of claim 2, wherein, For any one key analysis data, if the abnormal coincidence coefficient is greater than the preset abnormal coincidence coefficient, obtain each type of regional data of the key analysis data to determine the data change period of the corresponding regional data; According to the abnormal change correlation of each type of regional data, set the warning priority coefficient of each regional data. 4.The data analysis based wearable device monitoring method of claim 3, wherein, For any one key analysis data, if the abnormal coincidence coefficient is less than or equal to the preset abnormal coincidence coefficient, search for associated personnel in the key monitoring video corresponding to the key analysis data, and extract each interaction video segment corresponding to each associated personnel in the key monitoring video, and determine the video information tracing method according to the interaction time length proportion of the target monitoring personnel in the key monitoring video and the number of associated personnel; The associated personnel corresponding to the key monitoring video is determined according to the personnel association coefficient of each monitoring personnel and the target monitoring personnel in the video monitoring period, and the monitoring personnel whose personnel association coefficient is greater than the preset personnel association coefficient is recorded as the associated personnel. 5.The data analysis based wearable device monitoring method of claim 4, wherein, For a single key monitoring video, if the interaction time length proportion is greater than the preset interaction time length proportion or the number of associated personnel is greater than the preset number of associated personnel, set the warning priority coefficient of the corresponding interaction video segment according to the interaction correlation coefficient of each interaction video segment and the corresponding key analysis data; The early warning priority coefficient and the cross-correlation coefficient are in a positive correlation. 6.The data analysis based wearable device monitoring method of claim 5, wherein, For a single key monitoring video, if the interaction time length proportion is less than or equal to a preset interaction time length proportion and the number of associated personnel is less than or equal to a preset number of associated personnel, the key monitoring video is divided to obtain a plurality of video passages to be analyzed, and the historical abnormal repetition coefficient of the key analysis data corresponding to the video passages to be analyzed is detected to determine a video extraction strategy according to the historical abnormal repetition coefficient. 7.The data analysis based wearable device monitoring method of claim 6, wherein, If the historical abnormal repetition coefficient is greater than a preset historical abnormal repetition coefficient, the early warning priority coefficient is set according to the feature coincidence degree and the corresponding feature correlation coefficient of each video passage to be analyzed; The early warning priority coefficient and the feature reference value are in a positive correlation. 8.The data analysis based wearable device monitoring method of claim 7, wherein, If the historical abnormal repetition coefficient is less than or equal to a preset historical abnormal repetition coefficient, the key reference frame of the key monitoring video is set according to the relevant time of the key analysis data corresponding to the video passages to be analyzed, and the setting mode is determined according to the distribution state of the key reference frame; If there is a key reference frame in a dense state, the adjacent video passages to be analyzed of the key reference frame in the dense state are recorded as an early warning video passage, and the early warning priority coefficient of the early warning video passage is determined according to the number of key reference frames included in the early warning video passage; The early warning priority coefficient and the number of key reference frames are in a positive correlation. 9.The data analysis based wearable device monitoring method of claim 8, wherein, The behavior features of the video passages to be analyzed corresponding to each key reference frame in a dispersed state are extracted, and the early warning priority coefficient of the video passage to be analyzed is set according to the matching coefficient of each behavior feature and the corresponding key reference frame; The early warning priority coefficient and the matching coefficient are in a positive correlation. 10.The data analysis based wearable device monitoring method of claim 9, wherein, Under the condition that the information analysis is completed, the early warning information of each key analysis data is sent to the user according to the early warning priority coefficient of the video passage and the regional data; The information analysis completion condition is that each key analysis data completes the setting of the early warning priority coefficient of the regional data, or the corresponding key monitoring video completes the setting of the early warning priority coefficient of the video passage.

Images