Wearable device alarm method and system

Abstract

The invention provides a wearable device alarm method and system. The method comprises the following steps: receiving user monitoring state information, user motion information and user environment information; receiving user first-class attributes and user second-class attributes; obtaining a first monitoring attribute concentration interval based on the user one-class attribute; obtaining a monitoring attribute system error based on the second-class attribute of the user; according to the monitoring attribute system error, correcting the first monitoring attribute concentration interval to obtain a second monitoring attribute concentration interval; and calculating a deviation coefficient between the user monitoring state information and the interval in the second monitoring attribute set, and fusing the deviation coefficient and the user position to generate an alarm signal to execute alarm. The monitoring data is dynamically corrected by fusing the universal attributes and the personalized attributes of the user, the semantic abundance of the alarm information is improved, the alarm accuracy of the wearable device is improved, and therefore optimal scheduling of limited rescue resources is achieved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent alarm, and particularly to a wearable device alarm method and system. Background Art

[0002] Currently, the alarm mechanism of wearable devices is mainly based on preset threshold judgment, that is, when the monitored user status information exceeds a predetermined threshold range, an alarm is triggered. For example, when it is detected that the user's heart rate is too high or too low, blood pressure is abnormal, or the user has fallen, etc., the device will send an alarm signal according to the preset threshold and notify relevant personnel or institutions for rescue in combination with the user's location information. However, due to the large differences in physiological characteristics, health conditions, daily habits, etc. among different users, a unified threshold standard is difficult to meet individual needs, and this threshold judgment mechanism has obvious deficiencies, resulting in improper allocation of rescue resources. For example, for users who often engage in high-intensity exercise, an increase in heart rate may be a normal phenomenon; while for users with weak heart function, even a small change in heart rate may indicate a serious problem. Therefore, there is a technical problem in the existing wearable device alarm that, based on the comparison of predetermined thresholds, the semantic richness of the alarm information is insufficient, resulting in ineffective resource scheduling according to the needs of different users. Summary of the Invention

[0003] Aiming at the technical problem in the existing wearable device alarm that only based on the comparison of predetermined thresholds, the semantic richness of the alarm information is insufficient, resulting in ineffective resource scheduling according to the needs of different users, the present invention provides a wearable device alarm method and system to solve this problem.

[0004] The technical solution of the present invention to solve the above technical problems is as follows: In a first aspect, the present invention provides a wearable device alarm method, including: receiving user monitoring status information, user movement information, and user environment information; receiving user first-class attributes and user second-class attributes, where the user first-class attributes are general attributes predefined by an administrator, and the user second-class attributes are personalized attributes that can be optionally added by the user; based on the user first-class attributes, performing monitoring attribute sample statistics on the user movement information and the user environment information to obtain a first monitoring attribute concentration interval; based on the user second-class attributes, combining the user movement information and the user environment information, traversing the monitoring attribute set to perform monitoring attribute sample statistics to obtain a monitoring attribute system error; according to the monitoring attribute system error, correcting the first monitoring attribute concentration interval to obtain a second monitoring attribute concentration interval; calculating a deviation coefficient between the user monitoring status information and the second monitoring attribute concentration interval, and fusing the deviation coefficient and the user location to generate an alarm signal for alarm.

[0005] Optionally, calculating a deviation coefficient of the user monitoring status information from the middle interval of the second monitoring attributes includes: comparing the user monitoring status information with a monitoring status deviation vector matrix of the middle interval of the second monitoring attributes; analyzing a triggering probability of an abnormal event that satisfies the monitoring status deviation vector matrix; when the triggering probability of the abnormal event is less than or equal to a triggering probability threshold, returning to start an execution loop of the process; when the triggering probability of the abnormal event is greater than the triggering probability threshold, setting the triggering probability of the abnormal event as the deviation coefficient.

[0006] Optionally, analyzing the triggering probability of the abnormal event that satisfies the monitoring status deviation vector matrix includes: configuring a first monitoring status deviation vector record matrix to an Nth monitoring status deviation vector record matrix, where N is an integer and N≥50000; taking the user type-1 attributes as a constraint, collecting a plurality of user abnormal event triggering identifiers that satisfy the first monitoring status deviation vector record matrix, and the user abnormal event triggering identifiers are generated according to the decision result of the historical data at the rescue end. If it belongs to an abnormal event, the user abnormal event triggering identifier is equal to 1, otherwise, the user abnormal event triggering identifier is equal to 0; counting the ratio of the number of the user abnormal event triggering identifiers equal to 1 to the number of the plurality of user abnormal event triggering identifiers, and setting it as a first abnormal event triggering probability true value; until obtaining the Nth abnormal event triggering probability true value; using the first abnormal event triggering probability true value to the Nth abnormal event triggering probability true value as a supervision, and using the first monitoring status deviation vector record matrix to the Nth monitoring status deviation vector record matrix as inputs, configuring multiple groups of data to train an abnormal event triggering probability predictor, and analyzing the triggering probability of the abnormal event that satisfies the monitoring status deviation vector matrix.

[0007] Optionally, using the true values of the first abnormal event triggering probability to the true values of the Nth abnormal event triggering probability as supervision, and using the first monitoring status deviation vector record matrix to the Nth monitoring status deviation vector record matrix as input, configure multiple groups of data to train an abnormal event triggering probability predictor, including: Step 1: Set a training process quantity threshold, retrieve the multiple groups of data, and train the first abnormal event triggering probability predictor; Step 2: Extract the output error vector set of the first abnormal event triggering probability predictor; Step 3: When the output error vector variance of the output error vector set is greater than or equal to the variance threshold, return to Step 1 to execute the loop; Step 4: When the output error vector variance of the output error vector set is less than the variance threshold, statistically analyze the system error vector, and when the modulus value of the system error vector is greater than or equal to the convergence threshold, construct a first residual block, fuse it with the first abnormal event triggering probability predictor, generate a second abnormal event triggering probability predictor architecture, and return to Step 1 to execute the loop, where the second abnormal event triggering probability predictor architecture is equal to the sum of the outputs of the first abnormal event triggering probability predictor and the first residual block; Step 5: When the output error vector variance of the output error vector set is less than the variance threshold, statistically analyze the system error vector, and when the modulus value of the system error vector is less than the convergence threshold, set the first abnormal event triggering probability predictor as the abnormal event triggering probability predictor.

[0008] Optionally, based on the user's first-class attributes, perform monitoring attribute sample statistics on the user's motion information and the user's environment information to obtain the first monitoring attribute concentration interval, including: Based on the user's first-class attributes, according to the user's motion information and the user's environment information, and in combination with a predefined attribute deviation threshold, construct a first query constraint condition; Retrieve a first sample set that satisfies the first query constraint condition, where any sample in the first sample set includes user motion record data and user environment record data; Based on the user's first-class attributes, according to the user motion record data and the user environment record data, and in combination with a predefined attribute deviation threshold, construct a second query constraint condition; Retrieve a second sample set that satisfies the second query constraint condition; Statistically analyze the monitoring attribute concentration intervals of the first sample set and the second sample set, and set it as the first monitoring attribute concentration interval.

[0009] Optionally, based on the user's second-class attributes, in combination with the user's motion information and the user's environmental information, traverse the monitoring attribute set to perform monitoring attribute sample statistics, and obtain the monitoring attribute system error, including: based on the user's second-class attributes, based on the user's first-class attributes, in combination with the user's motion information and the user's environmental information, traverse the monitoring attribute set to perform monitoring attribute sample statistics, and obtain the middle interval of the third monitoring attribute set; calculate the first monitoring attribute intersection-union ratio of the middle interval of the third monitoring attribute set and the middle interval of the first monitoring attribute set, and configure the first monitoring attribute system error according to the first monitoring attribute intersection-union ratio; until the Mth monitoring attribute intersection-union ratio of the middle interval of the third monitoring attribute set and the middle interval of the first monitoring attribute set is calculated, and configure the Mth monitoring attribute system error according to the Mth monitoring attribute intersection-union ratio; add the first monitoring attribute system error to the Mth monitoring attribute system error into the monitoring attribute system error.

[0010] Optionally, configuring the first monitoring attribute system error according to the first monitoring attribute intersection-union ratio includes: when the first monitoring attribute intersection-union ratio is less than the intersection-union ratio threshold, calculate the upper interval boundary deviation vector set and the lower interval boundary deviation vector set of the first monitoring attribute; count the mode deviation vector of the upper interval boundary deviation vector set, and set it as the upper limit system error of the first monitoring attribute, count the mode deviation vector of the lower interval boundary deviation vector set, and set it as the lower limit system error of the first monitoring attribute; add the upper limit system error of the first monitoring attribute and the lower limit system error of the first monitoring attribute to the first monitoring attribute system error; when the first monitoring attribute intersection-union ratio is greater than or equal to the intersection-union ratio threshold, the first monitoring attribute system error is set to 0.

[0011] In a second aspect, the present invention provides a wearable device alarm system, including: an information receiving module, configured to receive user monitoring status information, user motion information, and user environmental information; a user attribute receiving module, configured to receive the user's first-class attributes and the user's second-class attributes, where the user's first-class attributes are general attributes predefined by an administrator, and the user's second-class attributes are personalized attributes that the user can optionally add; a general attribute analysis module, configured to perform monitoring attribute sample statistics on the user's motion information and the user's environmental information based on the user's first-class attributes, and obtain the middle interval of the first monitoring attribute set; a personalized attribute analysis module, configured to traverse the monitoring attribute set to perform monitoring attribute sample statistics in combination with the user's motion information and the user's environmental information based on the user's second-class attributes, and obtain the monitoring attribute system error; a middle interval correction module, configured to correct the middle interval of the first monitoring attribute set according to the monitoring attribute system error, and obtain the middle interval of the second monitoring attribute set; an alarm execution module, configured to calculate the deviation coefficient between the user monitoring status information and the middle interval of the second monitoring attribute set, and generate an alarm signal to execute an alarm by fusing the deviation coefficient and the user's location.

[0012] The beneficial effects of the present invention are as follows: By receiving user monitoring status information, user movement information, and user environment information, comprehensive data collection of users is achieved, ensuring a sufficient data basis for subsequent analysis. By receiving user first-class attributes and user second-class attributes, two types of attribute information are distinguished, laying a foundation for subsequent personalized analysis. Based on user first-class attributes, statistical analysis of monitoring attribute samples is performed on user movement information and user environment information to obtain the first monitoring attribute concentration interval. Preliminary statistical analysis of movement and environment information is carried out using general attributes to determine the normal state interval based on general attributes, providing a basic reference range for subsequent personalized correction. Based on user second-class attributes, combined with user movement information and user environment information, statistical analysis of monitoring attribute samples is carried out by traversing the monitoring attribute set to obtain the monitoring attribute system error. By introducing personalized attributes, the system error relative to the general model is calculated, and the monitoring deviation caused by individual differences is identified, providing a basis for subsequent precise correction. According to the monitoring attribute system error, the first monitoring attribute concentration interval is corrected to obtain the second monitoring attribute concentration interval, realizing the transformation from the general interval to the personalized interval, making the monitoring interval more in line with the individual characteristics of users and improving the accuracy of abnormal judgment. By calculating the deviation coefficient between the user monitoring status information and the second monitoring attribute concentration interval, an alarm signal is generated by fusing the deviation coefficient and the user location to execute the alarm. The introduction of the deviation coefficient enables the alarm information to have a degree of differentiation, and at the same time, combining the location information can help the rescue party determine the rescue priority and resource allocation strategy, realizing the improvement of the semantic richness of the alarm information.

[0013] Through the above technical solutions, the transformation from a threshold comparison to a dynamic alarm mechanism that integrates user individual characteristics is achieved, solving the problem of insufficient semantic richness of alarm information in the prior art and providing support for the optimal scheduling of limited rescue resources. Brief Description of the Drawings

[0014] Figure 1 It is a schematic flowchart of a wearable device alarm method provided by the present invention; Figure 2 It is a schematic structural diagram of a wearable device alarm system provided by the present invention.

[0015] In the drawings, the components represented by each reference numeral are as follows: Information receiving module 11, user attribute receiving module 12, general attribute analysis module 13, personalized attribute analysis module 14, concentration interval correction module 15, alarm execution module 16. Detailed Embodiments

[0016] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of protection of the present invention.

[0017] In the description of the present invention, the terms "first" and "second" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of the described features. In the description of the present invention, "a plurality" means two or more, unless otherwise specifically defined.

[0018] In the description of the present invention, the term "for example" is used to mean "serving as an example, illustration, or explanation". Any embodiment described as "for example" in the present invention is not necessarily construed as being more preferred or having more advantages than other embodiments. In order for any person skilled in the art to implement and use the present invention, the following description is given. In the following description, details are set forth for the purpose of explanation. It should be understood that those of ordinary skill in the art can recognize that the present invention can be implemented without using these specific details. In other instances, well-known structures and processes are not elaborated in detail to avoid unnecessary details from obscuring the description of the present invention. Therefore, the present invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope that conforms to the principles and features disclosed in the present invention.

[0019] Embodiment 1, as Figure 1 shown, the embodiment of the present invention provides a wearable device alarm method, which is applied to a wearable device and includes: S100: Receive user monitoring status information, user movement information, and user environment information.

[0020] Specifically, the wearable device collects and receives three types of information related to the user in real time through various built-in sensors, namely user monitoring status information, user movement information, and user environment information.

[0021] Among them, the user monitoring status information refers to the user's physiological parameter data collected by the wearable device, such as vital signs indicators like heart rate, blood pressure, body temperature, blood oxygen saturation, respiratory rate, etc. These data can reflect the user's current health status. The user movement information refers to the user's movement-related data collected by the wearable device through sensors such as accelerometers and gyroscopes, including but not limited to the user's walking steps, movement posture, movement speed, fall detection, stationary state recognition, etc. These data can reflect the user's movement state and potential abnormal behaviors. The user environment information refers to the relevant parameters of the user's environment collected by the wearable device through temperature and humidity sensors, light sensors, barometric pressure sensors, etc., including but not limited to environmental temperature, humidity, atmospheric pressure, environmental noise, light intensity, etc. These data can reflect the safety status and potential risks of the user's environment.

[0022] By receiving the user monitoring status information, user movement information, and user environment information, it provides the original data support for subsequent abnormal state analysis and intelligent alarm.

[0023] S200: Receive the user's first-class attributes and second-class attributes. Among them, the user's first-class attributes are general attributes predefined by the administrator, and the user's second-class attributes are personalized attributes that the user can optionally add.

[0024] Specifically, the wearable device receives two types of user attribute data, namely the user's first-class attributes and second-class attributes. These two types of attribute data constitute the user's personal profile and provide a personalized reference benchmark for subsequent monitoring and analysis.

[0025] Among them, the user's first-class attributes refer to the standardized basic user information predefined by the administrator, belonging to the category of general attributes. Such attributes usually include but are not limited to basic physiological characteristics such as the user's gender, age, height, weight, blood type, basal metabolic rate, etc. General attributes have objectivity and universality and are important bases for basic judgments. They can be used to divide users into different basic population categories and provide a basic reference standard for subsequent abnormal state judgments. The user's second-class attributes refer to the personalized information that the user independently chooses to input according to personal circumstances, belonging to the category of non-standardized personalized attributes. Such attributes usually include but are not limited to highly personalized information such as the user's eating habits, past medical history, drug allergy history, recent exercise data, living habits, etc. Personalized attributes can reflect the user's special circumstances and individual differences, help to more accurately identify the user's abnormal state, reduce the false alarm rate, and improve the accuracy and pertinence of the alarm.

[0026] By receiving the user's first-class attributes and second-class attributes, it is possible to construct a more accurate user monitoring mechanism on the basis of general standards, combined with the user's personalized characteristics, and provide comprehensive user characteristic support for subsequent abnormal state judgments and intelligent alarms.

[0027] S300: Based on the first type of user attributes, perform monitoring attribute sample statistics on the user's motion information and the user's environmental information to obtain the first monitoring attribute concentration interval.

[0028] Specifically, based on the aforementioned received first type of user attributes (i.e., the general attributes predefined by the administrator, such as basic information like the user's gender, age, height, weight, etc.), perform monitoring attribute sample statistical analysis on the user's motion information and the user's environmental information, so as to obtain the first monitoring attribute concentration interval.

[0029] First, use the first type of user attributes as the basic screening conditions to retrieve the user group with similar general attributes to the current user in the historical database. For example, for an adult male user aged 30 - 35, with a height of 175 - 180 cm and a weight of 70 - 75 kg, retrieve the historical data of male users with similar age ranges, height ranges, and weight intervals. Then, perform statistical analysis on the motion information and environmental information of the screened user group, and calculate the normal distribution intervals of various monitoring attributes, including but not limited to the daily average step range, normal activity intensity interval, heart rate change range, blood pressure fluctuation range, suitable environmental temperature interval, safe environmental humidity range, etc.

[0030] Through monitoring attribute sample statistics, obtain the monitoring attribute concentration interval based on general attributes, that is, the first monitoring attribute concentration interval. The first monitoring attribute concentration interval represents the reasonable fluctuation range of the user group with similar user motion information and user environmental information under normal circumstances, providing a preliminary reference standard for subsequent abnormal state judgment.

[0031] S400: Based on the second type of user attributes, combine the user's motion information and the user's environmental information, traverse the monitoring attribute set to perform monitoring attribute sample statistics, and obtain the monitoring attribute system error.

[0032] Specifically, based on the second type of user attributes (i.e., the personalized attributes that the user can optionally add, such as diet recipes, past medical history, recent exercise data, etc.), combine the user's motion information and the user's environmental information, comprehensively traverse and deeply statistically analyze the monitoring attribute set, so as to obtain the monitoring attribute system error.

[0033] Specifically, the system first takes the user's secondary attributes as additional constraint conditions and retrieves a specific user subgroup in the historical database that satisfies both the user's primary attributes (general attributes) and has similar secondary attributes (personalized attributes). For example, for a user with a history of mild hypertension, who has been doing low-intensity aerobic exercise all year round and has a low-salt and low-fat diet, a specific user group with similar health conditions, exercise habits, and eating habits is screened out. Then, statistical analysis is performed on the historical monitoring data of this specific user group to obtain the centralized interval of the monitoring attributes considering personalized factors, that is, the third centralized interval of the monitoring attributes. By comparing the differences between the third centralized interval of the monitoring attributes and the first centralized interval of the monitoring attributes, the systematic error of each monitoring attribute is calculated. The systematic error of the monitoring attribute reflects the correction amount of the user's personalized characteristics to the general monitoring standard. For example, for a user with a history of mild hypertension, the upper limit value of the normal blood pressure range may need to be slightly increased compared to the general standard; for a user who has been doing aerobic exercise for a long time, their normal resting heart rate may be lower than that of the average population of the same age.

[0034] Through personalized statistical analysis, the systematic error of the monitoring attribute based on the user's personalized characteristics is obtained, which will be used for subsequent personalized adjustment of the monitoring standard, thereby improving the accuracy and pertinence of the alarm and reducing false alarms or missed alarms caused by individual differences.

[0035] S500: According to the systematic error of the monitoring attribute, correct the first centralized interval of the monitoring attribute to obtain the second centralized interval of the monitoring attribute.

[0036] Specifically, the obtained systematic error of the monitoring attribute is used to precisely correct the established first centralized interval of the monitoring attribute, thereby obtaining a more personalized second centralized interval of the monitoring attribute, which can more accurately reflect the normal state range of a specific user.

[0037] Specifically, the corresponding systematic error of the monitoring attribute is applied to the upper and lower boundaries of the first centralized interval of the monitoring attribute respectively. For each monitoring attribute, add its upper boundary to the corresponding upper systematic error value and add its lower boundary to the corresponding lower systematic error value to generate new boundary values after personalized adjustment. For example, if a user's resting heart rate is usually 5 - 10 beats per minute lower than that of the same-age group due to long-term physical exercise, then the lower limit of the normal heart rate range of this user is correspondingly lowered by 5 - 10 beats per minute; if a user's normal body temperature is slightly higher than that of ordinary people due to specific physique, then the upper limit of the normal body temperature range of this user is appropriately increased.

[0038] Through the personalized correction process, the first monitoring attribute concentration interval (group statistical interval based on common attributes) is converted into the second monitoring attribute concentration interval (customized interval taking into account personalized factors), thereby realizing personalized and accurate monitoring based on group statistics, which not only avoids the misjudgment that may be caused by relying entirely on common standards, but also overcomes the problem of insufficient data that may be faced by pure individual monitoring. The second monitoring attribute concentration interval provides a more accurate reference standard for subsequent abnormal state judgment, improves the semantic richness of alarm information, and effectively improves the judgment accuracy and reliability of wearable device alarms.

[0039] S600: Calculate the deviation coefficient between the user monitoring state information and the second monitoring attribute concentration interval, fuse the deviation coefficient and the user position to generate an alarm signal and execute an alarm.

[0040] Specifically, the user monitoring status information collected in real time is compared and analyzed with the established second monitoring attribute concentration interval, the deviation coefficient is calculated, and combined with the user's location information, an intelligent alarm signal is generated and a precise alarm is executed.

[0041] First, the user's monitoring status information (including physiological indicators such as heart rate, blood pressure, body temperature, blood oxygen, and information such as posture and motion status) is compared item by item with the second monitoring attribute concentration interval, and the deviation degree of each monitoring attribute is quantified and calculated to calculate a comprehensive deviation coefficient. The deviation coefficient can quantitatively characterize the degree and urgency of the abnormality of the user's status. The larger the deviation coefficient, the farther the user's status deviates from the normal range, the higher the potential risk, and the higher the priority of rescue. Subsequently, the calculated deviation coefficient is fused with the real-time user location. The user location includes the user's geographic coordinates, the type of environment (such as indoors, outdoors, in waters, high altitude, etc.), and the surrounding facilities. By fusing the deviation coefficient and location information, an alarm signal containing multi-dimensional information such as the degree of abnormality, location accuracy, and environmental risks can be generated. Afterwards, an alarm is executed according to the fused alarm signal. For example, for minor abnormalities, only a reminder is sent to the user himself; for moderate abnormalities, the user's emergency contact is notified at the same time; and for severe abnormalities, an emergency help signal containing user status details and precise location will be automatically sent to the rescue center.

[0042] Through the intelligent alarm mechanism based on the fusion of deviation coefficient and location information, the problem of insufficient information semantic richness in the traditional alarm system is effectively solved, so that rescue resources can be prioritized according to the actual degree of urgency, thereby improving the alarm accuracy and efficiency of wearable devices.

[0043] Further, calculating the deviation coefficient between the user monitoring state information and the second monitoring attribute concentration interval includes: S610: Compare the user monitoring status information with the monitoring status deviation vector matrix of the second monitoring attribute concentration interval; S620: Analyze the triggering probability of abnormal events that meet the monitoring status deviation vector matrix; S630: When the triggering probability of the abnormal event is less than or equal to the triggering probability threshold, return to the start of the process to execute the loop; S640: When the triggering probability of the abnormal event is greater than the triggering probability threshold, set the triggering probability of the abnormal event as the deviation coefficient.

[0044] In a preferred implementation, first, compare the user monitoring status information with the second monitoring attribute concentration interval to generate a monitoring status deviation vector matrix. Specifically, compare each current physiological index (such as heart rate, blood pressure, body temperature, blood oxygen saturation, etc.), motion state parameter (such as posture, acceleration, angular velocity, etc.), and environmental parameter (such as environmental temperature, humidity, etc.) in the user monitoring status information with the upper and lower limit values of the corresponding second monitoring attribute concentration interval, calculate the deviation value and deviation direction of each index, and organize these multi-dimensional deviation data into a structured deviation vector matrix to obtain the monitoring status deviation vector matrix. This matrix not only contains the deviation degree of each index but also reflects the correlation and combination pattern between the indexes. Then, based on the obtained monitoring status deviation vector matrix, analyze the triggering probability of various abnormal events that the user may encounter. For example, match and calculate the similarity between the current monitoring status deviation vector matrix and the deviation vector matrix corresponding to the known abnormal event patterns in the historical data, so as to estimate the occurrence probability of various abnormal events (such as falling, drowning, heart attack, hypoxia, hypothermia, etc.) that the user's current state may cause. Through probability-based risk assessment, it is possible to make a pre-judgment at the early stage of the abnormal state without having to wait until the index seriously deviates to trigger an alarm.

[0045] Subsequently, compare the calculated triggering probability of the abnormal event with the preset triggering probability threshold. When the triggering probability of the abnormal event is less than or equal to the triggering probability threshold, it indicates that although the user's current state deviates, the risk level is controllable. Return to the starting point of the monitoring process and continue to execute the monitoring loop to continuously observe the user's state. When the triggering probability of the abnormal event is greater than the triggering probability threshold, it indicates that the user's current state deviation has reached the risk warning line. Directly set the triggering probability value of the abnormal event as the deviation coefficient, which is an important parameter for the subsequent alarm process. This deviation coefficient not only quantifies the degree of the user's state abnormality but also reflects the type and urgency of the potential risk event, providing a basis for the subsequent intelligent alarm.

[0046] By converting multi-dimensional monitoring data into deviation coefficients, the alarm signal can accurately reflect the nature and urgency of the abnormal user status, thus supporting the efficient scheduling and precise intervention of rescue resources.

[0047] Further, analyzing the abnormal event triggering probability that satisfies the monitoring status deviation vector matrix includes: S621: Configure the first monitoring status deviation vector record matrix up to the Nth monitoring status deviation vector record matrix, where N is an integer and N ≥ 50000; S622: Constrained by the user's first-class attributes, collect several user abnormal event triggering identifiers that satisfy the first monitoring status deviation vector record matrix. The user abnormal event triggering identifier is generated based on the decision result of historical data at the rescue end. If it belongs to an abnormal event, the user abnormal event triggering identifier is equal to 1; otherwise, the user abnormal event triggering identifier is equal to 0; S623: Statistically calculate the ratio of the number of user abnormal event triggering identifiers equal to 1 to the number of the several user abnormal event triggering identifiers, and set it as the first abnormal event triggering probability true value; S624: Until the Nth abnormal event triggering probability true value is obtained; S625: Using the first abnormal event triggering probability true value to the Nth abnormal event triggering probability true value as supervision, and using the first monitoring status deviation vector record matrix to the Nth monitoring status deviation vector record matrix as input, configure multiple sets of data to train the abnormal event triggering probability predictor, and analyze the abnormal event triggering probability that satisfies the monitoring status deviation vector matrix.

[0048] In a preferred embodiment, first, a large-scale historical monitoring status deviation vector record matrix dataset is configured, which includes the first monitoring status deviation vector record matrix to the Nth monitoring status deviation vector record matrix, where N is an integer not less than 50,000. These record matrices contain the monitoring data deviation patterns of a large number of historical users in various states and constitute the basic dataset for abnormal event probability analysis. Each monitoring status deviation vector record matrix is a multi-dimensional data structure, containing the deviation values of various physiological indicators, motion parameters, and environmental factors and their combined relationships. Setting N not less than 50,000 ensures the sufficiency of the samples and helps improve the accuracy of the abnormal event trigger probability. Then, taking a user's first-class attributes (i.e., general attributes such as gender, age, height, weight, etc.) as the constraint conditions, for the first monitoring status deviation vector record matrix, several user abnormal event trigger identifiers that match it are collected from the historical database. These trigger identifiers are binary markers generated based on the actual decision results of the rescue end in historical data: if the corresponding status is finally confirmed as an abnormal event (such as an actual occurrence of an emergency such as a fall, drowning, heart attack, etc.), the user abnormal event trigger identifier is set to 1; if the corresponding status is confirmed as a false alarm or normal fluctuation, the user abnormal event trigger identifier is set to 0. This marking method based on the actual rescue results ensures the authenticity and reliability of the data. Subsequently, a statistical analysis is performed on the collected user abnormal event trigger identifiers, and the ratio of the number of samples with the identifier value equal to 1 (i.e., the number of cases where an abnormal event actually occurred) to the total number of all collected samples is calculated, and this ratio is set as the true value of the first abnormal event trigger probability. The true value of the first abnormal event trigger probability reflects the historical frequency of the actual occurrence of abnormal events under a specific monitoring status deviation pattern.

[0049] Repeat the process of steps S622 and S623 for the second monitoring status deviation vector record matrix to the Nth monitoring status deviation vector record matrix, and calculate and obtain the true value of the second abnormal event trigger probability to the true value of the Nth abnormal event trigger probability in sequence. Subsequently, using the obtained true value of the first abnormal event trigger probability to the true value of the Nth abnormal event trigger probability as the supervision labels, and using the corresponding first monitoring status deviation vector record matrix to the Nth monitoring status deviation vector record matrix as the input features, multiple groups of training data are configured, and machine learning or deep learning algorithms are applied to train the abnormal event trigger probability predictor. After being fully trained, this predictor can accurately predict the probability of a new monitoring status deviation vector matrix leading to an abnormal event, thereby providing a basis for precise alarm.

[0050] By refining empirical rules from historical data and constructing an abnormal event trigger probability predictor, the risk assessment of abnormal events is based on a large amount of empirical data, improving the accuracy and reliability of the alarm.

[0051] Further, with the true values of the first abnormal event triggering probability to the true values of the Nth abnormal event triggering probability as the supervision, and the first monitoring state deviation vector recording matrix to the Nth monitoring state deviation vector recording matrix as the input, configure multiple groups of data to train the abnormal event triggering probability predictor, including: Step 1: Set the threshold of the number of training processes, retrieve the multiple groups of data, and train the first abnormal event triggering probability predictor; Step 2: Extract the output error vector set of the first abnormal event triggering probability predictor; Step 3: When the variance of the output error vectors in the output error vector set is greater than or equal to the variance threshold, return to Step 1 to execute the loop; Step 4: When the variance of the output error vectors in the output error vector set is less than the variance threshold, calculate the system error vector, and when the norm of the system error vector is greater than or equal to the convergence threshold, construct the first residual block, fuse it with the first abnormal event triggering probability predictor to generate the second abnormal event triggering probability predictor architecture, and return to Step 1 to execute the loop, where the second abnormal event triggering probability predictor architecture is equal to the sum of the outputs of the first abnormal event triggering probability predictor and the first residual block; Step 5: When the variance of the output error vectors in the output error vector set is less than the variance threshold, calculate the system error vector, and when the norm of the system error vector is less than the convergence threshold, set the first abnormal event triggering probability predictor as the abnormal event triggering probability predictor.

[0052] In a preferred embodiment, to train the abnormal event triggering probability predictor, first, set the threshold of the number of training processes, which is used to control the number of training iterations to ensure that the training process is both sufficient and efficient. Subsequently, retrieve multiple groups of training data and use machine learning algorithms to train the first abnormal event triggering probability predictor. The first abnormal event triggering probability predictor is a model that can receive the monitoring state deviation vector matrix as the input and output the abnormal event triggering probability. The training process adopts the supervised learning method, and by continuously adjusting the model parameters, the error between the predicted output and the true probability value is minimized. Subsequently, evaluate the performance of the training result of the first abnormal event triggering probability predictor and extract its output error vector set. These error vectors reflect the deviation between the predicted value and the true value of the first abnormal event triggering probability predictor and are indicators for evaluating the model performance and guiding further optimization. Then, calculate the variance of the output error vector set and compare it with the preset variance threshold. When the output error vector variance is greater than or equal to the variance threshold, it indicates that the stability of the prediction result of the first abnormal event triggering probability predictor is insufficient and there is a large random fluctuation, and further optimization is required. At this time, return to Step 1 to re-execute the training loop to improve the stability and consistency of the first abnormal event triggering probability predictor.

[0053] When the variance of the output error vector set is less than the variance threshold, it indicates that the prediction result of the first abnormal event trigger probability predictor already has good stability, but the systematic component of the prediction deviation of the first abnormal event trigger probability predictor still needs to be further evaluated. Then calculate the average value of the error vectors in the output error vector set to obtain the system error vector, and calculate its modulus value and compare it with the preset convergence threshold. When the modulus value of the system error vector is greater than or equal to the convergence threshold, it indicates that there is an obvious systematic deviation in the first abnormal event trigger probability predictor and structural adjustment is required. At this time, based on the current system error vector, a special first residual block is constructed and fused with the first abnormal event trigger probability predictor to generate an improved second abnormal event trigger probability predictor architecture. The output of this architecture is equal to the sum of the output of the first predictor and the output of the first residual block. Through this residual learning mechanism, the systematic deviation is effectively compensated and the prediction accuracy is improved. After obtaining the second abnormal event trigger probability predictor architecture, return to step one and re-execute the training process using the second abnormal event trigger probability predictor architecture. When the variance of the output error vector set is less than the variance threshold and the modulus value of the system error vector is less than the convergence threshold, it indicates that the first abnormal event trigger probability predictor has met the requirements of both stability and accuracy, and the training process can be terminated. At this time, the first abnormal event trigger probability predictor is determined as the final abnormal event trigger probability predictor for subsequent abnormal event trigger probability evaluation.

[0054] Through adaptive training and optimization, an abnormal event trigger probability predictor with both high accuracy and high stability can be constructed, providing reliable decision support for the intelligent alarm of wearable devices and improving the accuracy and timeliness of abnormal state recognition.

[0055] Furthermore, based on the user's first type of attributes, monitor attribute sample statistics are performed on the user's motion information and the user's environment information to obtain the first monitoring attribute concentration interval, including: S310: Based on the user's first type of attributes, according to the user's motion information and the user's environment information, combined with the predefined attribute deviation threshold, construct the first query constraint condition; S320: Retrieve the first sample set that satisfies the first query constraint condition, where any sample in the first sample set includes user motion record data and user environment record data; S330: Based on the user's first type of attributes, according to the user's motion record data and the user's environment record data, combined with the predefined attribute deviation threshold, construct the second query constraint condition; S340: Retrieve the second sample set that satisfies the second query constraint condition; S350: Statistically analyze the monitoring attribute concentration intervals of the first sample set and the second sample set, and set it as the first monitoring attribute concentration interval.

[0056] In a feasible implementation, first, based on the user's first-class attributes (i.e., general attributes predefined by the administrator, such as basic information like gender, age, height, weight, etc.), combined with the currently collected user movement information and user environment information, and referring to the predefined attribute deviation threshold, construct the first query constraint condition. The first query constraint condition is a set of structured query rules used to locate user groups with similar basic characteristics and environmental conditions to the current user in the historical database. For example, for a 35-year-old male user with a height of 175 cm and a weight of 70 kg, construct the following query constraint conditions: gender = male, age range = 30 - 40 years old, height range = 170 - 180 cm, weight range = 65 - 75 kg, and further refine the query conditions by combining the current user's movement state (such as stationary, walking, running, etc.) and environmental conditions (such as indoor, temperature range, humidity range, etc.). Then, use the constructed first query constraint condition to perform a retrieval operation in the historical database to obtain the first sample set that meets the conditions. Each sample in the first sample set contains two parts of data, namely user movement record data and user environment record data. Among them, the user movement record data includes, but is not limited to, parameters such as the movement state, posture changes, and activity intensity of historical users; the user environment record data includes, but is not limited to, parameters such as the environmental temperature, humidity, air pressure, and light where historical users are located.

[0057] Then, further based on the user's first-class attributes, combined with the retrieved user movement record data and user environment record data, and referring to the predefined attribute deviation threshold, construct the second query constraint condition, which can achieve sample expansion, effectively increase the breadth of data collection, and thus be able to capture a more comprehensive reference sample. Next, use the constructed second query constraint condition to perform a second-round retrieval operation in the historical database to obtain the second sample set that meets the new conditions. The second sample set is a supplementary set of similar samples to the first sample set. Through this multi-level query strategy, the coverage range of the reference sample is expanded, the robustness of statistical analysis is enhanced, and the data basis for establishing the first monitoring attribute concentration interval is ensured to be more sufficient and comprehensive. After that, perform comprehensive statistical analysis on the obtained first sample set and the obtained second sample set, calculate the distribution characteristics of each monitoring attribute, including, but not limited to, statistical quantities such as mean, standard deviation, and quantile, so as to determine the normal fluctuation interval of each monitoring attribute, set it as the first monitoring attribute concentration interval, and use it as the basic reference standard for subsequent personalized adjustment.

[0058] Through refined data retrieval and statistical analysis, a monitoring benchmark interval that highly matches the basic characteristics of the current user can be constructed based on a large amount of historical data, laying a foundation for subsequent personalized optimization and abnormal state judgment.

[0059] Furthermore, based on the user's secondary attributes, combined with the user's movement information and the user's environmental information, traverse the monitoring attribute set for monitoring attribute sample statistics to obtain the monitoring attribute systematic error, including: S410: Based on the user's secondary attributes, based on the user's primary attributes, combined with the user's movement information and the user's environmental information, traverse the monitoring attribute set for monitoring attribute sample statistics to obtain the third monitoring attribute set middle interval; S420: Calculate the first monitoring attribute intersection-union ratio of the third monitoring attribute set middle interval and the first monitoring attribute set middle interval, and configure the first monitoring attribute systematic error according to the first monitoring attribute intersection-union ratio; S430: Until calculating the Mth monitoring attribute intersection-union ratio of the third monitoring attribute set middle interval and the first monitoring attribute set middle interval, and configure the Mth monitoring attribute systematic error according to the Mth monitoring attribute intersection-union ratio; S440: Add the first monitoring attribute systematic error until the Mth monitoring attribute systematic error to the monitoring attribute systematic error.

[0060] In an alternative embodiment, based on the user's secondary attributes (i.e., personalized attributes that the user can optionally add, such as eating habits, past medical history, recent exercise data, etc.) and the user's primary attributes (general attributes), combined with the user's movement information and the user's environmental information, comprehensively traverse and deeply statistically analyze the monitoring attribute set. Through the statistical analysis with double constraints, obtain the third monitoring attribute set middle interval that takes into account the user's personalized characteristics. The third monitoring attribute set middle interval reflects the normal fluctuation range of the user's various monitoring indicators under specific personalized conditions. Then, calculate the intersection-union ratio of the first monitoring attribute between the third monitoring attribute set middle interval and the first monitoring attribute set middle interval, that is, the first monitoring attribute intersection-union ratio. Among them, the first monitoring attribute is one of the monitoring attributes in the monitoring attribute set, and the monitoring attribute set includes all attributes in the user's secondary attributes and the user's primary attributes. The intersection-union ratio is an index to measure the similarity of two intervals, and the calculation method is the ratio of the intersection to the union of the two intervals. The higher the intersection-union ratio value, the greater the overlap degree of the two intervals, and the smaller the impact of the user's personalized characteristics on the general monitoring standard; the lower the intersection-union ratio value, the greater the adjustment requirement of the personalized factor on the monitoring standard. According to the calculated first monitoring attribute intersection-union ratio, configure the first monitoring attribute systematic error, which reflects the adjustment amount that needs to be made to the first monitoring attribute set middle interval.

[0061] Repeat the calculation of the intersection-over-union ratio and the system error configuration in S420, and process the second monitoring attribute until the Mth monitoring attribute (M is the total number of attributes in the monitoring attribute set) in sequence, obtaining the system error of the first monitoring attribute until the system error of the Mth monitoring attribute, comprehensively quantifying the influence degree of user personalization characteristics on various monitoring indicators. After that, add all the obtained system errors of the first monitoring attribute until the system error of the Mth monitoring attribute into the monitoring attribute system error, which contains the personalized adjustment parameters for each monitoring attribute and will be used later to correct the interval in the first monitoring attribute set, so as to generate a more personalized interval in the second monitoring attribute set.

[0062] Through the intersection-over-union ratio analysis and the system error configuration, the conversion from the general monitoring standard to the personalized monitoring standard is realized, providing a customized monitoring benchmark for the intelligent alarm of wearable devices, and improving the accuracy and pertinence of abnormal state judgment.

[0063] Furthermore, configuring the system error of the first monitoring attribute according to the intersection-over-union ratio of the first monitoring attribute includes: S421: When the intersection-over-union ratio of the first monitoring attribute is less than the intersection-over-union ratio threshold, calculate the upper-boundary deviation vector set and the lower-boundary deviation vector set of the first monitoring attribute interval; S422: Statistically analyze the mode deviation vector of the upper-boundary deviation vector set, which is set as the upper system error of the first monitoring attribute, and statistically analyze the mode deviation vector of the lower-boundary deviation vector set, which is set as the lower system error of the first monitoring attribute; S423: Add the upper system error of the first monitoring attribute and the lower system error of the first monitoring attribute into the system error of the first monitoring attribute; S424: When the intersection-over-union ratio of the first monitoring attribute is greater than or equal to the intersection-over-union ratio threshold, the system error of the first monitoring attribute is set to 0.

[0064] In a preferred embodiment, taking the configuration of the system error of the first monitoring attribute according to the intersection-over-union ratio of the first monitoring attribute as an example, the configuration process of the monitoring attribute system error is described. First, judge the size relationship between the intersection-over-union ratio of the first monitoring attribute and the intersection-over-union ratio threshold. When the intersection-over-union ratio of the first monitoring attribute is less than the preset intersection-over-union ratio threshold, it indicates that there are significant differences between the obtained interval in the third monitoring attribute set and the obtained interval in the first monitoring attribute set, and systematic adjustment is required. At this time, calculate the upper-boundary deviation vector set and the lower-boundary deviation vector set of the first monitoring attribute interval. Specifically, compare the upper and lower limit values of the third monitoring attribute set and the first monitoring attribute set on the first monitoring attribute, and generate a deviation vector set representing these differences. These deviation vectors reflect the adjustment requirements of the monitoring interval boundary caused by personalized factors.

[0065] Subsequently, statistical analysis is performed on the obtained set of upper-bound deviation vectors of the interval to identify the deviation vector with the highest frequency of occurrence (i.e., the mode deviation vector), and it is set as the upper-limit systematic error of the first monitoring attribute. Similarly, similar statistical analysis is performed on the set of lower-bound deviation vectors of the interval to identify the mode deviation vector, and it is set as the lower-limit systematic error of the first monitoring attribute. By using the mode as the systematic error, the interference of extreme values is effectively avoided, ensuring statistical stability and representativeness. Then, the determined upper-limit systematic error and lower-limit systematic error of the first monitoring attribute are added to the systematic error of the first monitoring attribute.

[0066] When the intersection-over-union ratio of the first monitoring attribute is greater than or equal to the intersection-over-union ratio threshold, it indicates that the interval in the obtained third monitoring attribute set is highly consistent with the interval in the obtained first monitoring attribute set, and the influence of personalized features on this monitoring attribute is negligible. At this time, directly set the systematic error of the first monitoring attribute to 0, that is, do not make any adjustment to the interval boundary of this monitoring attribute, and retain its original value in the general standard.

[0067] Through the systematic error configuration based on the intersection-over-union ratio, differential processing of different monitoring attributes is achieved: for attributes significantly affected by personalized factors, precise quantitative boundary adjustment is performed; for attributes slightly affected by personalized factors, the general standard remains unchanged. This flexible adjustment mechanism not only avoids unnecessary complexity but also ensures the accuracy and effectiveness of personalized adjustment, providing a monitoring standard that combines generality and personalization for the intelligent alarm system of wearable devices.

[0068] Embodiment 2, as Figure 2 shown, based on the same inventive concept as the wearable device alarm method provided in Embodiment 1, the present invention embodiment also provides a wearable device alarm system, including: An information receiving module 11, configured to receive user monitoring status information, user movement information, and user environment information; A user attribute receiving module 12, configured to receive user type-one attributes and user type-two attributes, where the user type-one attributes are general attributes predefined by an administrator, and the user type-two attributes are personalized attributes optionally added by the user; A general attribute analysis module 13, configured to perform monitoring attribute sample statistics on the user movement information and the user environment information based on the user type-one attributes to obtain a first monitoring attribute set interval; A personalized attribute analysis module 14, configured to perform monitoring attribute sample statistics by traversing the monitoring attribute set based on the user type-two attributes in combination with the user movement information and the user environment information to obtain a monitoring attribute systematic error; The centralized interval correction module 15 is used to correct the first monitored attribute centralized interval according to the system error of the monitored attribute, and obtain the second monitored attribute centralized interval; The alarm execution module 16 is used to calculate the deviation coefficient between the user monitoring status information and the second monitored attribute centralized interval, and generate an alarm signal to execute the alarm by fusing the deviation coefficient and the user location.

[0069] Furthermore, the alarm execution module 16 includes the following execution steps: Compare the user monitoring status information with the monitoring status deviation vector matrix of the second monitored attribute centralized interval; Analyze the triggering probability of abnormal events that meet the monitoring status deviation vector matrix; When the triggering probability of the abnormal event is less than or equal to the triggering probability threshold, return to the start of the process to execute the loop; When the triggering probability of the abnormal event is greater than the triggering probability threshold, set the triggering probability of the abnormal event as the deviation coefficient.

[0070] Furthermore, the alarm execution module 16 also includes the following execution steps: Configure the first monitoring status deviation vector record matrix to the Nth monitoring status deviation vector record matrix, where N is an integer and N≥50000; Constrained by the user's first-class attributes, collect several user abnormal event trigger identifiers that meet the first monitoring status deviation vector record matrix. The user abnormal event trigger identifier is generated according to the decision result of the historical data at the rescue end. If it belongs to an abnormal event, the user abnormal event trigger identifier is equal to 1; otherwise, the user abnormal event trigger identifier is equal to 0; Count the number of user abnormal event trigger identifiers equal to 1, and calculate the ratio to the number of several user abnormal event trigger identifiers, which is set as the true value of the first abnormal event triggering probability; Until the true value of the Nth abnormal event triggering probability is obtained; Using the true value of the first abnormal event triggering probability to the true value of the Nth abnormal event triggering probability as supervision, and using the first monitoring status deviation vector record matrix to the Nth monitoring status deviation vector record matrix as input, configure multiple sets of data to train the abnormal event triggering probability predictor, and analyze the triggering probability of abnormal events that meet the monitoring status deviation vector matrix.

[0071] Furthermore, the alarm execution module 16 also includes the following execution steps: Step 1: Set the threshold of the number of training processes, retrieve the multiple sets of data, and train the first abnormal event triggering probability predictor; Step 2: Extract the output error vector set of the first abnormal event triggering probability predictor; Step 3: When the output error vector variance of the output error vector set is greater than or equal to the variance threshold, return to Step 1 to execute the loop; Step 4: When the output error vector variance of the output error vector set is less than the variance threshold, statistically analyze the system error vector, and when the system error vector modulus is greater than or equal to the convergence threshold, construct a first residual block, fuse it with the first abnormal event triggering probability predictor to generate a second abnormal event triggering probability predictor architecture, and return to Step 1 to execute the loop, where the second abnormal event triggering probability predictor architecture is equal to the sum of the output of the first abnormal event triggering probability predictor and the first residual block; Step 5: When the output error vector variance of the output error vector set is less than the variance threshold, statistically analyze the system error vector, and when the system error vector modulus is less than the convergence threshold, set the first abnormal event triggering probability predictor as the abnormal event triggering probability predictor.

[0072] Furthermore, the general attribute analysis module 13 includes the following execution steps: Based on the user's first-class attributes, according to the user's motion information and the user's environment information, combined with the predefined attribute deviation threshold, construct the first query constraint condition; Retrieve the first sample set that satisfies the first query constraint condition, where any sample in the first sample set includes user motion record data and user environment record data; Based on the user's first-class attributes, according to the user's motion record data and the user's environment record data, combined with the predefined attribute deviation threshold, construct the second query constraint condition; Retrieve the second sample set that satisfies the second query constraint condition; Statistically analyze the monitoring attribute concentration intervals of the first sample set and the second sample set, and set them as the first monitoring attribute concentration intervals.

[0073] Furthermore, the personalized attribute analysis module 14 includes the following execution steps: Based on the user's second-class attributes, based on the user's first-class attributes, combined with the user's motion information and the user's environment information, traverse the monitoring attribute set to perform monitoring attribute sample statistics to obtain the third monitoring attribute concentration interval; Calculate the first monitoring attribute intersection-union ratio of the third monitoring attribute concentration interval and the first monitoring attribute concentration interval, and configure the first monitoring attribute system error according to the first monitoring attribute intersection-union ratio; Until the Mth monitoring attribute intersection-union ratio of the third monitoring attribute concentration interval and the first monitoring attribute concentration interval is calculated, and the Mth monitoring attribute system error is configured according to the Mth monitoring attribute intersection-union ratio; Add the first monitoring attribute systematic error to the Mth monitoring attribute systematic error into the monitoring attribute systematic error.

[0074] Furthermore, the personalized attribute analysis module 14 further includes the following execution steps: When the intersection over union of the first monitoring attribute is less than the intersection over union threshold, calculate the upper interval boundary deviation vector set and the lower interval boundary deviation vector set of the first monitoring attribute; Statistically analyze the mode deviation vector of the upper interval boundary deviation vector set, denoted as the upper systematic error of the first monitoring attribute, and statistically analyze the mode deviation vector of the lower interval boundary deviation vector set, denoted as the lower systematic error of the first monitoring attribute; Add the upper systematic error of the first monitoring attribute and the lower systematic error of the first monitoring attribute to the systematic error of the first monitoring attribute; When the intersection over union of the first monitoring attribute is greater than or equal to the intersection over union threshold, the systematic error of the first monitoring attribute is set to 0.

[0075] It should be noted that in the above embodiments, the descriptions of the various embodiments have their own emphases. For the parts not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0076] Those skilled in the art should understand 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 aspects. 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 code.

[0077] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to the embodiments of the present invention. It should be understood that each flow and / or block in the flowchart and / or block diagram, and the combination of flows and / or blocks in the flowchart and / or block diagram, can be realized by computer program instructions. These computer program instructions can be provided to the processors of general-purpose computers, special-purpose computers, embedded computers, or other programmable data processing devices to generate a machine, so that the instructions executed by the processors of the computer or other programmable data processing devices generate means for realizing the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0078] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to work in a particular manner, such that the instructions stored in the computer-readable memory produce a manufacture including an instruction device that implements the functions specified in one or more of the procedures Figure 1 or blocks Figure 1 specified in one or more of the procedures and / or blocks.

[0079] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus, such that a series of operational steps are performed on the computer or other programmable apparatus to produce a computer-implemented process, whereby the instructions executed on the computer or other programmable apparatus provide steps for implementing the functions specified in one or more of the procedures Figure 1 or blocks Figure 1 specified in one or more of the blocks.

[0080] Although the preferred embodiments of the present invention have been described, additional changes and modifications can be made to these embodiments by those skilled in the art once they learn of the basic inventive concept.

[0081] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the present invention and its equivalent technologies, the present invention is also intended to include these modifications and variations.

Claims

1. A method for a wearable device to give an alarm, characterized in that, Applied to wearable devices, including: Receiving user monitoring status information, user movement information, and user environment information; Receiving user first-class attributes and user second-class attributes, where the user first-class attributes are general attributes predefined by the administrator, and the user second-class attributes are personalized attributes optionally added by the user; Based on the user first-class attributes, performing monitoring attribute sample statistics on the user movement information and the user environment information to obtain a first monitoring attribute central interval; Based on the user second-class attributes, combining the user movement information and the user environment information, traversing the monitoring attribute set for monitoring attribute sample statistics to obtain a monitoring attribute system error; According to the monitoring attribute system error, correcting the first monitoring attribute central interval to obtain a second monitoring attribute central interval; Calculating a deviation coefficient between the user monitoring status information and the second monitoring attribute central interval, fusing the deviation coefficient and the user location to generate an alarm signal for alarming.

2. The method according to claim 1, wherein Calculating the deviation coefficient between the user monitoring status information and the second monitoring attribute central interval includes: Comparing the monitoring status deviation vector matrix of the user monitoring status information and the second monitoring attribute central interval; Analyzing the abnormal event trigger probability that satisfies the monitoring status deviation vector matrix; When the abnormal event trigger probability is less than or equal to the trigger probability threshold, return to the process start to execute the loop; When the abnormal event trigger probability is greater than the trigger probability threshold, set the abnormal event trigger probability as the deviation coefficient.

3. The method according to claim 2, wherein Analyzing the abnormal event trigger probability that satisfies the monitoring status deviation vector matrix includes: Configuring a first monitoring status deviation vector record matrix to the Nth monitoring status deviation vector record matrix, where N is an integer and N≥50000; Constrained by the user first-class attributes, collecting several user abnormal event trigger identifiers that satisfy the first monitoring status deviation vector record matrix. The user abnormal event trigger identifier is generated based on the decision result of the historical data at the rescue end. If it belongs to an abnormal event, the user abnormal event trigger identifier is equal to 1, otherwise, the user abnormal event trigger identifier is equal to 0; Counting the number of user abnormal event trigger identifiers equal to 1 and the ratio to the number of several user abnormal event trigger identifiers, and setting it as the first abnormal event trigger probability true value; Until the Nth abnormal event trigger probability true value is obtained; Using the first abnormal event trigger probability true value to the Nth abnormal event trigger probability true value as supervision, and using the first monitoring status deviation vector record matrix to the Nth monitoring status deviation vector record matrix as input, configuring multiple groups of data to train an abnormal event trigger probability predictor to analyze the abnormal event trigger probability that satisfies the monitoring status deviation vector matrix.

4. The method according to claim 3, wherein Using the first abnormal event trigger probability true value to the Nth abnormal event trigger probability true value as supervision, and using the first monitoring status deviation vector record matrix to the Nth monitoring status deviation vector record matrix as input, configuring multiple groups of data to train an abnormal event trigger probability predictor, including: Step 1: Set the threshold for the number of training processes, retrieve the multiple sets of data, and train the first abnormal event triggering probability predictor; Step 2: Extract the output error vector set of the first abnormal event triggering probability predictor; Step 3: When the variance of the output error vectors in the output error vector set is greater than or equal to the variance threshold, return to Step 1 to execute the loop; Step 4: When the variance of the output error vectors in the output error vector set is less than the variance threshold, calculate the system error vector, and when the norm of the system error vector is greater than or equal to the convergence threshold, construct the first residual block, fuse it with the first abnormal event triggering probability predictor to generate the architecture of the second abnormal event triggering probability predictor, and return to Step 1 to execute the loop, where the architecture of the second abnormal event triggering probability predictor is equal to the sum of the outputs of the first abnormal event triggering probability predictor and the first residual block; Step 5: When the variance of the output error vectors in the output error vector set is less than the variance threshold, calculate the system error vector, and when the norm of the system error vector is less than the convergence threshold, set the first abnormal event triggering probability predictor as the abnormal event triggering probability predictor.

5. The method according to claim 1, wherein Based on the user's first type of attributes, perform monitoring attribute sample statistics on the user's motion information and the user's environment information to obtain the first monitoring attribute concentration interval, including: Based on the user's first type of attributes, according to the user's motion information and the user's environment information, and in combination with the predefined attribute deviation threshold, construct the first query constraint condition; Retrieve the first sample set that satisfies the first query constraint condition, where any sample in the first sample set includes user motion record data and user environment record data; Based on the user's first type of attributes, according to the user's motion record data and the user's environment record data, and in combination with the predefined attribute deviation threshold, construct the second query constraint condition; Retrieve the second sample set that satisfies the second query constraint condition; Statistically calculate the monitoring attribute concentration intervals of the first sample set and the second sample set, and set them as the first monitoring attribute concentration interval.

6. The method according to claim 5, wherein Based on the user's second type of attributes, in combination with the user's motion information and the user's environment information, traverse the monitoring attribute set to perform monitoring attribute sample statistics to obtain the monitoring attribute system error, including: Based on the user's second type of attributes, based on the user's first type of attributes, in combination with the user's motion information and the user's environment information, traverse the monitoring attribute set to perform monitoring attribute sample statistics to obtain the third monitoring attribute concentration interval; Calculate the first monitoring attribute intersection and union ratio of the third monitoring attribute concentration interval and the first monitoring attribute concentration interval, and configure the first monitoring attribute system error according to the first monitoring attribute intersection and union ratio; Until calculating the Mth monitoring attribute intersection and union ratio of the third monitoring attribute concentration interval and the first monitoring attribute concentration interval, and configure the Mth monitoring attribute system error according to the Mth monitoring attribute intersection and union ratio; Add the first monitoring attribute system error to the Mth monitoring attribute system error into the monitoring attribute system error.

7. The method according to claim 6, characterized in that Configuring the first monitoring attribute system error according to the intersection over union of the first monitoring attribute includes: When the intersection over union of the first monitoring attribute is less than the intersection over union threshold, calculating the upper interval boundary deviation vector set and the lower interval boundary deviation vector set of the first monitoring attribute; Counting the mode deviation vector of the upper interval boundary deviation vector set, which is set as the upper system error of the first monitoring attribute, and counting the mode deviation vector of the lower interval boundary deviation vector set, which is set as the lower system error of the first monitoring attribute; Adding the upper system error of the first monitoring attribute and the lower system error of the first monitoring attribute to the first monitoring attribute system error; When the intersection over union of the first monitoring attribute is greater than or equal to the intersection over union threshold, the first monitoring attribute system error is set to 0.

8. A wearable device alarm system, characterized in that, A method for alarming a wearable device according to any one of claims 1-7, applied to a wearable device, includes: An information receiving module, configured to receive user monitoring status information, user movement information, and user environment information; A user attribute receiving module, configured to receive user first-class attributes and user second-class attributes, where the user first-class attributes are general attributes predefined by an administrator, and the user second-class attributes are personalized attributes optionally added by the user; A general attribute analysis module, configured to perform monitoring attribute sample statistics on the user movement information and the user environment information based on the user first-class attributes to obtain a first monitoring attribute concentration interval; A personalized attribute analysis module, configured to perform monitoring attribute sample statistics by traversing the monitoring attribute set based on the user second-class attributes in combination with the user movement information and the user environment information to obtain a monitoring attribute system error; A concentration interval correction module, configured to correct the first monitoring attribute concentration interval according to the monitoring attribute system error to obtain a second monitoring attribute concentration interval; An alarm execution module, configured to calculate a deviation coefficient between the user monitoring status information and the second monitoring attribute concentration interval, and generate an alarm signal for execution based on the fusion of the deviation coefficient and the user location.

Images