A personal health abnormal behavior monitoring system integrating unsupervised learning

Abstract

The present invention provides a personal health abnormal behavior monitoring system that integrates unsupervised learning and relates to the field of data processing technology. The system is used to: cluster historical behavior data, divide the historical behavior data into multiple behavior labels, extract the feature vector of each behavior label, and establish an individual behavior physical model based on the feature vector; input target feature data into the individual behavior physical model; calculate the degree of deviation of the user's recent behavior; monitor and record the deviation to obtain an observation label group; determine whether the observation label group meets the conditions for absorption by the individual behavior physical model; if the result is negative, the behavior label is determined to be abnormal behavior; calculate the behavior inhibition factor based on the deviation degree and duration of the abnormal behavior, and dynamically adjust the individual behavior physical model based on the behavior inhibition factor to prevent abnormal behavior from being identified as normal behavior. The present invention can prevent the system from mistaking periodic abnormal behavior for normal behavior.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a personal health abnormal behavior monitoring system integrating unsupervised learning. Background Art

[0002] At present, personal health monitoring systems based on machine learning are gradually replacing traditional rule-based alarm methods. In particular, the method of identifying behavioral patterns through unsupervised learning and then using supervised learning to establish a personalized baseline model has become a common practice. This type of method usually first uses the K-means clustering algorithm to automatically identify behavioral labels such as bedtime, toilet frequency, and indoor activity interval distribution from behavioral data within 7-15 days, and then trains the model with supervised algorithms such as support vector machines to establish an individual's health behavior baseline. After inputting new data, the model realizes health alerts by comparing the deviation between current behavior and historical patterns. This type of method has good adaptability in building personalized models and is particularly suitable for long-term monitoring in chronic disease management and home care scenarios.

[0003] However, in practical applications, there are still key technical flaws. The model's lack of sensitivity to potential behavioral variations may prevent it from promptly identifying abnormal patterns that develop over multiple days. For example, in a nighttime behavioral monitoring scenario, if an elderly person recently experiences frequent toileting due to urinary problems, their behavior may have been within normal fluctuations in the previous few days, and the model would consider this as a reasonable part of a health baseline update. However, if this abnormality continues to evolve into a new behavioral habit, the model, lacking a mechanism for intentional recognition of abnormal patterns, may classify this behavior as normal, leading to missed reports. Current technology may not be able to clearly distinguish between deviant but reasonable behaviors, and there is a risk that abnormal behaviors will be absorbed by the model as the new normal. Summary of the Invention

[0004] The purpose of the present invention is to provide a personal health abnormal behavior monitoring system integrating unsupervised learning, aiming to solve the problems mentioned in the background technology.

[0005] In order to solve the above technical problems, the technical solutions of the present invention are as follows:

[0006] A personal health abnormal behavior monitoring system integrating unsupervised learning, the system comprising:

[0007] The unsupervised module is used to obtain historical behavior data of users and perform clustering processing on it, dividing the historical behavior data into multiple behavior labels to obtain behavior label data;

[0008] The individual model module is used to extract the feature vector of each behavior label based on the behavior label data, obtain sample feature data, and establish an individual behavior physical model based on it;

[0009] The target behavior module is used to obtain the user's recent target behavior data, cluster it, and extract features to obtain target feature data;

[0010] The deviation calculation module is used to input the target feature data into the individual behavior physical model, calculate the deviation degree of the user's recent behavior, and monitor and record the abnormal behavior labels whose deviation degree exceeds the preset deviation threshold to obtain an observation label group;

[0011] The anomaly detection module is used to determine whether the observed label group meets the conditions absorbed by the individual behavior physical model. If the result is no, the behavior label is determined to be abnormal behavior;

[0012] The behavior inhibition module is used to calculate the behavior inhibition factor according to the deviation degree and duration of abnormal behavior, and dynamically adjust the individual behavior physical model based on it to limit abnormal behavior from being identified as normal behavior.

[0013] Furthermore, the unsupervised module includes:

[0014] A time segmentation unit is used to segment and process the historical behavior data, splitting the historical behavior data into multiple time segments according to a preset time window to obtain a time segment data set;

[0015] A feature extraction unit is used to extract features from the time segment data set, extract the user's stay rate, spatial transfer frequency and related physiological values in different spatial areas within each time segment, and obtain a feature data set;

[0016] The clustering unit is used to perform clustering processing on the preliminary feature data set, classifying time segments with similar features into the same cluster to obtain clustering result data;

[0017] The behavior mapping unit is used to perform behavior analysis on each cluster in the clustering result data, identify the significant features of the cluster in spatial behavior and physiological parameters, and map the significant features into behavior labels to obtain behavior label data.

[0018] Furthermore, the behavior mapping unit includes:

[0019] A statistical analysis unit is used to perform statistics on the time segment data contained in each cluster in the clustering result data, and respectively calculate the spatial behavior parameters and physiological parameters of the users in each spatial area in the cluster to obtain a cluster characteristic index set;

[0020] The significant feature identification unit is used to standardize the spatial behavioral parameters and physiological parameters in the cluster feature index set, and screen out the spatial regions and physiological value intervals with significant features according to the preset behavioral feature threshold to obtain the significant feature set;

[0021] The label classification unit is used to match the set of significant features with the preset behavior pattern, select labels according to the spatial area and physiological value range, determine the behavior label type that best matches the cluster, and bind the behavior label type to the cluster to obtain behavior label data.

[0022] Furthermore, the individual model module includes:

[0023] The behavior decomposition unit is used to perform sample analysis on each behavior tag in the behavior tag data, extract the behavior occurrence time period, behavior duration, behavior spatial area, behavior occurrence frequency and behavior time concentration corresponding to the tag, and obtain the behavior feature set;

[0024] The behavior structure unit is used to construct ternary structure data for each behavior tag in the behavior feature unit set. The ternary structure data includes a time period field, a spatial region field, and a behavior feature field.

[0025] The behavior sequence unit is used to combine all the ternary structure data in chronological order, identify the user's behavior patterns in different time periods and different spatial areas, and obtain the behavior structure sequence;

[0026] The model building unit is used to define the ternary structure with repeated patterns as periodic behavior according to the behavior structure sequence, and convert it into a regular expression form to generate an individual behavior physical model.

[0027] Furthermore, the model building unit includes:

[0028] The candidate behavior unit is used to repeatedly match the ternary structure data in the behavior structure sequence, and classify the time periods and spatial regions with more than a preset overlap rate in multiple consecutive days and the ternary structures with behavior characteristics exceeding the preset overlap rate as the same candidate behavior segment, thereby obtaining a candidate behavior segment set;

[0029] a periodicity identification unit, configured to perform statistical calculations on the occurrence frequency and time concentration of each candidate behavior segment based on the candidate behavior segment set, and identify the behavior segment as a periodic behavior when its occurrence frequency is greater than a preset frequency threshold and its time concentration is less than a preset concentration threshold;

[0030] The rule generation unit is used to reconstruct the pattern of the time period field, spatial area field and behavior feature field in the periodic behavior, extract the average occurrence time, main area coordinates and stable behavior features, and obtain the structured behavior set;

[0031] The physical modeling unit is used to determine the user's behavior pattern in a specific time period and spatial area based on the structured behavior set, and obtain the individual behavior physical model.

[0032] Furthermore, the deviation calculation module includes:

[0033] A first calculation unit is used to compare the target feature data with the time field, space field, and behavior field of the corresponding behavior tag in the individual behavior physical model, calculate the difference between the behavior feature set, obtain the behavior deviation indicator set, and calculate the deviation degree based on the difference;

[0034] An anomaly identification unit is used to compare the degree of deviation with a preset deviation threshold, and when the degree of deviation is greater than the preset deviation threshold, determine that the behavior label corresponding to the target feature data is a suspicious behavior label;

[0035] The behavior recording unit is used to monitor suspicious behavior tags, record the first occurrence time, last occurrence time, consecutive occurrence times and spatial distribution range of suspicious behavior tags, and obtain an observation tag group.

[0036] Furthermore, the first computing unit includes:

[0037] The second calculation unit is configured to obtain a first deviation index item based on the product relationship between the behavior time period, behavior frequency, and behavior concentration of the user's recent behavior, combined with the inhibitory effect of behavior concentration on time-frequency offset; obtain a second deviation index item based on the interaction between the behavior duration and the spatial location of the user's recent behavior, combined with the adjustment function of spatial offset on duration fluctuation; and obtain a third deviation index item based on the joint relationship between the behavior time period, behavior concentration, and behavior duration of the user's recent behavior, combined with the degree of interference of frequency offset on this joint relationship;

[0038] The first deviation index item, the second deviation index item, and the third deviation index item are integrated to obtain the degree of deviation of the user's recent behavior from the individual behavior physical model.

[0039] Furthermore, the anomaly detection module includes:

[0040] A trend extraction unit is used to extract the continuous occurrence records of suspicious behavior tags in the observation tag group in the time dimension to obtain a trend sample set;

[0041] A trend analysis unit is used to calculate the average frequency and average deviation of suspicious behavior labels in continuous records based on the trend sample set to obtain a trend feature indicator set;

[0042] The judgment rule unit is used to judge whether the average frequency of the trend characteristic indicator set exceeds the frequency of behavior of the same type of label in the individual behavior physical model, and at the same time judge whether the average deviation degree exceeds the preset deviation threshold. If any judgment result is yes, it is determined that the behavior label does not meet the absorption condition;

[0043] The abnormality judgment unit is used to judge whether the suspicious behavior label has a trend reversal within a continuous time window. When there is no trend reversal, the suspicious behavior label is confirmed to be an abnormal behavior.

[0044] Furthermore, the behavior suppression module includes:

[0045] The inhibition factor unit is used to analyze the minimum confidence level that needs to be met when the abnormal behavior is absorbed into the individual behavior physical model based on the deviation degree and duration of the abnormal behavior, and obtain the behavior inhibition factor;

[0046] The absorption judgment unit is used to compare the behavior inhibition factor with the absorption trigger strength of the individual behavior physical model. When the behavior inhibition factor is greater than the absorption trigger strength, the abnormal behavior does not meet the absorption conditions;

[0047] An absorption adjustment unit, configured to dynamically restrict the behavior label associated with the abnormal behavior in the individual behavior physical model, thereby increasing the duration of the update window of the behavior label in the individual behavior physical model;

[0048] The behavior blocking unit is used to add a blocking status mark to the abnormal behavior label that does not meet the absorption conditions, and store the behavior label, its deviation degree, and behavior inhibition factor in the behavior adjustment data.

[0049] Furthermore, the inhibitory factor unit includes:

[0050] The suppression factor calculation unit is used to calculate the impact of changes in the time distribution and trust of abnormal behavior based on the logarithm of the time offset of the abnormal behavior and its coverage ratio in the individual behavior physical model, thereby obtaining the main control factor term; calculate the immediate interference capability of short-term abnormal behavior on the individual behavior physical model based on the duration difference of the abnormal behavior, thereby obtaining the linear adjustment factor term; calculate the overall suppression amplitude based on the ratio of the concentration offset value to the model coverage, and the ratio of the frequency offset value to the behavior duration, thereby obtaining the balance adjustment term;

[0051] The main control factor item, linear adjustment factor item and balance adjustment item are integrated to obtain the behavioral inhibition factor of abnormal behavior.

[0052] The above solution of the present invention includes at least the following beneficial effects:

[0053] The present invention establishes a linkage mechanism between the anomaly detection module and the behavior suppression module to achieve dynamic identification and screening of potential abnormal behaviors in the process of updating the individual behavior physical model, which can effectively prevent the technical defect of normalization of anomalies in traditional models, especially in scenarios where user behavior gradually deviates from the original rules, such as increased frequency of going to the toilet at night due to physiological health problems. The system can judge whether the abnormal behavior has the conditions to be absorbed by the model based on the degree of deviation and its duration. When it is determined that the absorption conditions are not met, the update logic of the behavior model will be intervened based on the behavior inhibition factor, dynamically limiting the update window time of the behavior label in the behavior model, and adding a blocking state to the behavior label to ensure that the anomaly will not be regarded as a new behavior baseline. This pre-blocking mechanism based on the behavior evolution trend effectively prevents behavioral deviations from penetrating into the model structure, maintains the long-term stability and predictive ability of the model, and has significant value in the practical application of individual health management.

[0054] The present invention determines the evolution trend of behavior and analyzes the trend characteristics of the continuity and deviation degree of the user's recent behavior by observing the tag group, so that the system has the ability to predict progressive anomalies in advance. Unlike the traditional model's outlier detection method based only on single-point offset, this system pays more attention to the evolution direction and stability of behavior in the time dimension. By extracting target behavior data and constructing a trend sample set, and further combining trend indicators such as average frequency and deviation degree, intervention can be made before the behavior has completely formed a significant deviation. The system can make judgments based on its continuous change trend and the comparison results with similar tags in the individual behavior model to achieve early warning. This trend judgment mechanism improves the stability and foresight of the system in long-term behavior monitoring, and helps to provide timely health intervention recommendations.

[0055] The present invention can comprehensively judge the deviation index from three aspects: behavior time period, spatial location and behavior concentration, and quantify the overall deviation. Different from the traditional single-dimensional deviation identification method such as only using activity frequency as the criterion, the system fully considers the multivariate interaction relationship between behavioral data and realizes the quantitative description of multi-dimensional differential behaviors. Through the deviation index set composed of these calculation results, the system can perform more precise identification of complex behavioral differences, thereby reducing the false alarm and missed alarm rates. It is particularly suitable for use in scenarios with strong behavioral concentration and high sporadicity, such as toileting and sleeping, effectively improving the system's recognition rate of atypical abnormal behaviors and enhancing the credibility and adaptability of actual health monitoring systems.

[0056] The present invention further generates an individual behavior physical model in the form of rule expression by constructing a behavioral feature set, a ternary structure sequence and a behavioral structure sequence. Compared with existing rule-based or statistical modeling methods, this solution breaks through the limitations of traditional single-variable modeling in design, introduces the ability to express complex features such as behavioral time period, spatial area, behavioral frequency, and concentration, and realizes the dynamic characterization of behavioral patterns. This modeling method has strong personalized expression capabilities and can adapt to behavioral differences between different users, ensuring that the model has high adaptability and interpretability. At the same time, it also provides structured support for multi-dimensional comparison, provides a data basis for system anomaly identification, and improves the long-term stability and behavior prediction capabilities of the model.

[0057] The present invention introduces the calculation logic of behavioral inhibition factors, comprehensively judges the credibility of abnormal behaviors based on multiple factors such as the degree of deviation, time coverage, behavior frequency, and concentration changes, dynamically generates inhibition factors, and compares them with the absorption threshold of the behavioral model to form an absorption constraint mechanism. If the inhibition factor is greater than the model trigger strength, the system will prevent the model from being updated, thereby further enhancing the conservatism and controllability of the model. In the scenario of dynamically evolving health conditions, this mechanism can ensure that individual behavior models are not disturbed by short-term abnormal behaviors, effectively extend the model life cycle and improve prediction consistency, and has practical value in maintaining the accuracy of long-term monitoring systems, reflecting a high degree of engineering practicality. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 This is a flowchart of a personal health abnormal behavior monitoring system integrated with unsupervised learning provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0059] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0060] like Figure 1 As shown, an embodiment of the present invention proposes a personal health abnormal behavior monitoring system integrating unsupervised learning, the system comprising:

[0061] The unsupervised module is used to obtain historical behavior data of users and perform clustering processing on it, dividing the historical behavior data into multiple behavior labels to obtain behavior label data;

[0062] The individual model module is used to extract the feature vector of each behavior label based on the behavior label data, obtain sample feature data, and establish an individual behavior physical model based on it;

[0063] The target behavior module is used to obtain the user's recent target behavior data, cluster it, and extract features to obtain target feature data;

[0064] The deviation calculation module is used to input the target feature data into the individual behavior physical model, calculate the deviation degree of the user's recent behavior, and monitor and record the abnormal behavior labels whose deviation degree exceeds the preset deviation threshold to obtain an observation label group;

[0065] The anomaly detection module is used to determine whether the observed label group meets the conditions absorbed by the individual behavior physical model. If the result is no, the behavior label is determined to be abnormal behavior;

[0066] The behavior inhibition module is used to calculate the behavior inhibition factor according to the deviation degree and duration of abnormal behavior, and dynamically adjust the individual behavior physical model based on it to limit abnormal behavior from being identified as normal behavior.

[0067] In an embodiment of the present invention, an unsupervised module is used to obtain historical behavior data for a user and perform clustering processing thereon, dividing the historical behavior data into multiple behavior labels to obtain behavior label data. By introducing a joint analysis of spatial behavior parameters and physiological parameters, the clustering results are made more physiologically meaningful and behaviorally interpretable, thereby improving the accuracy of subsequent models and the stability of behavior recognition; an individual model module is used to extract the feature vector of each behavior label based on the behavior label data to obtain sample feature data, and to establish an individual behavior physical model based thereon. By means of structured modeling, it not only fully expresses the typical behavior pattern of the user, but also enhances the model's ability to recognize the sequence and regularity of behavior; a target behavior module is used to obtain the user's recent target behavior data, cluster and extract features thereon to obtain target feature data, thereby realizing the system's rapid response and feature extraction to recent behaviors, supporting continuous tracking of user status, and being able to detect abnormal fluctuations in behavior at the first time.

[0068] The deviation calculation module is used to input target feature data into the individual behavior physical model, calculate the degree of deviation of the user's recent behavior, and monitor and record abnormal behavior labels whose deviation exceeds the preset deviation threshold to obtain an observation label group. This enables the system to accurately respond to the mutual influence between different behavioral attributes, greatly reducing the missed and false alarm rates, and improving the accuracy and robustness of health anomaly detection. The anomaly detection module is used to determine whether the observation label group meets the conditions for absorption by the individual behavior physical model. If the result is not, the behavior label is determined to be abnormal behavior. By extracting and comparing trend features, it can distinguish between fluctuating behavior and gradual abnormal behavior, effectively improving the system's ability to judge the evolution trend of abnormalities. The behavior suppression module is used to calculate the behavior suppression factor based on the deviation degree and duration of abnormal behavior, and dynamically adjust the individual behavior physical model based on it to prevent abnormal behavior from being identified as normal behavior. By quantifying and dynamically adjusting the model absorption conditions, it can effectively control the evolution process of the model structure, prevent abnormal behavior from being incorrectly learned by the model due to frequent occurrence, and improve the safety and stability of the entire system during actual health monitoring.

[0069] The target behavior module is used to obtain the user's recent target behavior data, cluster it, and extract features to obtain target feature data, including:

[0070] The system first accesses the user's original behavioral data from the past 1 to 3 days from the data collection end. This data can be obtained through environmental sensors, wearable devices or smart terminals installed in the user's residence. Specifically, it includes the user's activity trajectory, spatial location, residence time, behavior frequency, physiological parameters such as heart rate and body temperature, and environmental parameters such as illumination and noise in different time periods. The system synchronizes and cleans the collected data in time, eliminates invalid values, missing fragments and abnormal noise, to ensure the reliability of subsequent analysis.

[0071] Subsequently, the system uses a preset time window strategy to divide the original data into time segments, usually in units of 30 minutes or 60 minutes, and cuts the data of a whole day into several continuous time segments. In each time segment, the system calculates the multi-dimensional features such as the user's activity stay ratio in each spatial area, spatial transfer frequency, and the average and fluctuation range of physiological parameters in the corresponding time period in the segment, and uniformly outputs vectorized feature representations based on the set feature template. Then, the system uses an unsupervised clustering algorithm consistent with historical data processing to cluster these feature vectors, and classify behavioral segments with similar features into the same cluster cluster, thereby forming new behavioral labels.

[0072] After clustering is completed, the system will conduct a preliminary analysis of the fragments within each cluster, identify their significant manifestations in time continuity, spatial concentration and physiological parameters, and confirm whether they belong to the evolutionary manifestation of a certain behavioral pattern. Finally, the system will generate a corresponding target behavior label for each cluster, and extract its central eigenvalue, standard deviation, spatial behavior hot zone, concentration fluctuation index and other data as target feature data. The target feature data will then be input into the deviation calculation module as the basic data source for comparison with the individual behavior physical model, thereby realizing the quantitative analysis and judgment of the differences between recent behavior and existing behavior models.

[0073] In a preferred embodiment of the present invention, the unsupervised module includes:

[0074] A time segmentation unit is used to segment and process the historical behavior data, splitting the historical behavior data into multiple time segments according to a preset time window to obtain a time segment data set;

[0075] A feature extraction unit is used to extract features from the time segment data set, extract the user's stay rate, spatial transfer frequency and related physiological values in different spatial areas within each time segment, and obtain a feature data set;

[0076] The clustering unit is used to perform clustering processing on the preliminary feature data set, classifying time segments with similar features into the same cluster to obtain clustering result data;

[0077] The behavior mapping unit is used to perform behavior analysis on each cluster in the clustering result data, identify the significant features of the cluster in spatial behavior and physiological parameters, and map the significant features into behavior labels to obtain behavior label data.

[0078] In an embodiment of the present invention, a time division unit is used to divide and process the historical behavior data, split the historical behavior data into multiple time segments according to a preset time window, obtain a time segment data set, and implement standardized preprocessing of unstructured continuous behavior data, so that complex time series are divided into behavior observation units of uniform granularity; a feature extraction unit is used to extract features from the time segment data set, extract the user's stay rate, spatial transfer frequency and associated physiological values in different spatial areas in each time segment, obtain a feature data set, realize the mapping from original perception data to structured behavior feature vectors, and solve the problems of fuzzy behavior representation and inconsistent data dimensions; a clustering unit is used to extract features from the time segment data set, extract the user's stay rate, spatial transfer frequency and associated physiological values in different spatial areas in each time segment, obtain a feature data set, and implement the mapping from original perception data to structured behavior feature vectors, so as to ..., obtain a feature data set, and realize the mapping from original perception data to structured behavior feature vectors, The system performs clustering on the preliminary feature data set, classifies time segments with similar features into the same cluster, and obtains clustering result data, thereby realizing the system's structured classification operation on non-labeled behavioral data, enabling the system to identify the intrinsic similarities and potential categories between behavioral patterns in an unsupervised situation; the behavior mapping unit is used to perform behavioral analysis on each cluster in the clustering result data, identify the significant features of the cluster in spatial behavior and physiological parameters, and map the significant features into behavioral labels to obtain behavioral label data, and semanticize the abstract clustering results, that is, to give the cluster corresponding to each time segment a clear behavioral meaning, so that the system can transition from low-level feature processing to high-level behavior understanding.

[0079] The clustering unit is used to perform clustering processing on the preliminary feature data set, classifying time segments with similar features into the same cluster, and obtaining clustering result data, which specifically includes:

[0080] The system first receives a feature data set generated by the feature extraction unit. The feature data set consists of feature vectors of multiple time segments. Each feature vector contains multi-dimensional information such as the user's stay rate in different spatial areas, spatial transfer frequency, and related physiological parameters within the time segment. In actual operation, the system first standardizes all feature vectors to eliminate the order of magnitude differences between different feature dimensions, avoiding the phenomenon in the subsequent clustering process where a certain type of indicator dominates the clustering distance due to its large numerical range.

[0081] The standardized feature data will be clustered, and K-means can be used as the clustering algorithm. The system divides the feature space according to the preset or dynamically adjusted cluster number K value. Before starting clustering, the system first uses the silhouette coefficient indicator to evaluate the stability of the clustering structure under different K values, and automatically selects the optimal K value. During the clustering process, the system continuously adjusts the category of each feature vector in an iterative manner to minimize the Euclidean distance between samples within the class and maximize the distance between samples between classes, thereby achieving the convergence judgment condition. In order to cope with scenarios with blurred boundaries and unclear behavioral transitions, the system can also introduce a GMM soft clustering strategy, so that some time segments can have membership in multiple cluster clusters at the same time, thereby enhancing the ability to recognize cross-behavior states.

[0082] After clustering is completed, the system will assign time segments with similar feature patterns to the same cluster cluster, and output clustering result data including the cluster cluster number, the corresponding time segment ID and its central feature vector. The clustering result is used as the input of the next stage to identify the specific behavior label type represented by each cluster cluster, thereby completing the transition from the original behavior segment to the semantic label. This process improves the system's ability to automatically identify user behavior patterns.

[0083] In a preferred embodiment of the present invention, the behavior mapping unit includes:

[0084] A statistical analysis unit is used to perform statistics on the time segment data contained in each cluster in the clustering result data, and respectively calculate the spatial behavior parameters and physiological parameters of the users in each spatial area in the cluster to obtain a cluster characteristic index set;

[0085] The significant feature identification unit is used to standardize the spatial behavioral parameters and physiological parameters in the cluster feature index set, and screen out the spatial regions and physiological value intervals with significant features according to the preset behavioral feature threshold to obtain the significant feature set;

[0086] The label classification unit is used to match the set of significant features with the preset behavior pattern, select labels according to the spatial area and physiological value range, determine the behavior label type that best matches the cluster, and bind the behavior label type to the cluster to obtain behavior label data.

[0087] In an embodiment of the present invention, a statistical analysis unit is used to perform statistics on the time segment data contained in each cluster in the clustering result data, and respectively calculate the spatial behavior parameters and physiological parameters of the users in each spatial area in the cluster to obtain a cluster feature index set. Through the coupled statistics of the spatial behavior parameters and the physiological parameters, the user's behavior state and physiological state changes in a specific environment are fully reflected; a significant feature identification unit is used to perform standardization processing on the spatial behavior parameters and physiological parameters in the cluster feature index set, and screen out spatial areas and physiological parameters with significant features according to a preset behavioral feature threshold. Physiological value interval, obtain the significant feature set, effectively eliminate redundant features, reduce noise interference, and improve the effectiveness and accuracy of behavioral feature extraction in the cluster; the label classification unit is used to match the significant feature set with the preset behavioral pattern, select labels according to the spatial area and physiological value interval, determine the behavioral label type that best matches the cluster, and bind the behavioral label type to the cluster to obtain behavioral label data. Through the label classification mechanism, the system can automatically identify the user's living status in a specific time period, improve the health monitoring system's ability to express behavioral semantics, and enhance the interpretability of subsequent abnormal behavior identification and intervention.

[0088] The label classification unit is used to match the significant feature set with the preset behavior pattern, select labels based on the spatial region and physiological value range, determine the behavior label type that best matches the cluster, and bind the behavior label type to the cluster to obtain behavior label data, which specifically includes:

[0089] First, a structured behavioral pattern library is constructed based on the system's long-term accumulated data on user behavior. Each behavioral pattern consists of a typical spatial area identifier, physiological parameter interval, and behavioral time period. As a benchmark reference for label determination, the label classification unit first receives the significant feature set output by the significant feature recognition unit during operation. This set usually includes the high-frequency activity behavior characteristics of users in the cluster in several spatial areas and the corresponding physiological indicator intervals.

[0090] The system then vectorizes each set of significant features and uses a matching algorithm to calculate the similarity between each set and each behavior pattern in the behavior pattern library. Specifically, a vector space algorithm can be used to calculate the Euclidean distance between the significant feature set and the preset pattern in the spatial dimension and the physiological parameter dimension to obtain a set of similarity scores. In the similarity score, the system sets a pre-screening threshold to eliminate candidate patterns with low matching, and then selects the behavior pattern with the highest similarity in the candidate set as the optimal matching result. If the significant feature set of a cluster is close to a preset behavior pattern in multiple dimensions, such as staying in the kitchen area for more than 15 minutes, with a heart rate fluctuating between 90-110, and the occurrence time is concentrated in the noon period, the system will match the cluster as a dining behavior label.

[0091] When the match is successful, the system will assign the label to the current cluster to realize the mapping from low-level behavioral data to high-level semantic labels. The label classification unit will bind the behavioral label to the cluster in the system to form a structured behavioral label data entry. The entry not only contains the label type, but also the spatial, physiological and temporal characteristics on which the label is based. Through this label classification process, the system can semanticize the original behavior clustering results, making the subsequent individual behavior physical model more interpretable, and also facilitating the system to display the behavior classification results to users or professionals in human-computer interaction or clinical analysis.

[0092] In a preferred embodiment of the present invention, the individual model module includes:

[0093] The behavior decomposition unit is used to perform sample analysis on each behavior tag in the behavior tag data, extract the behavior occurrence time period, behavior duration, behavior spatial area, behavior occurrence frequency and behavior time concentration corresponding to the tag, and obtain the behavior feature set;

[0094] The behavior structure unit is used to construct ternary structure data for each behavior tag in the behavior feature unit set. The ternary structure data includes a time period field, a spatial region field, and a behavior feature field.

[0095] The behavior sequence unit is used to combine all the ternary structure data in chronological order, identify the user's behavior patterns in different time periods and different spatial areas, and obtain the behavior structure sequence;

[0096] The model building unit is used to define the ternary structure with repeated patterns as periodic behavior according to the behavior structure sequence, and convert it into a regular expression form to generate an individual behavior physical model.

[0097] In an embodiment of the present invention, the behavior decomposition unit is used to perform sample analysis on each behavior label in the behavior label data, extract the behavior occurrence time period, behavior duration, behavior spatial area, behavior occurrence frequency and behavior time concentration corresponding to the label, and obtain a behavior feature set. The key feature dimensions are extracted through high-dimensional deconstruction, which helps to improve the accuracy and diversity of the model in the behavior feature representation and avoid the problem of fuzzy or overlapping behavior patterns caused by low-dimensional descriptions; the behavior structure unit is used to construct ternary structure data for each behavior label in the behavior feature unit set. The ternary structure data includes a time period field, a spatial area field and a behavior feature field, and converts the unstructured behavior feature results into a normalized structure to ensure The consistency in the expression of behavioral data facilitates subsequent processing; the behavioral sequence unit is used to combine all ternary structure data in chronological order, identify the user's behavioral patterns in different time periods and different spatial areas, obtain the behavioral structure sequence, organize discrete behavioral labels into continuous sequences, and explore the temporal laws between behaviors, providing a complete semantic basis for subsequent periodic identification; the model construction unit is used to define the ternary structure with repeated patterns as periodic behavior based on the behavioral structure sequence, and convert it into a rule expression form to generate an individual behavior physical model, and convert the abstract behavior chain into a periodic rule expression to realize the conversion from data to knowledge, so that the system has the long-term memory function of user-specific behavior.

[0098] The behavior sequence unit is used to combine all ternary structure data in chronological order, identify user behavior patterns in different time periods and different spatial areas, and obtain a behavior structure sequence, which specifically includes:

[0099] The system first receives multiple ternary structure data output from the behavior structure unit. Each ternary structure includes the time period field, spatial area field and behavior feature field of the user behavior. The system sorts and processes these structure data according to the time information contained in the time period field to ensure that the arrangement of the behavior data conforms to the actual occurrence order of the user's daily activities. When sorting, the system uses the time period start time of all behavior tags as the main sequence index, arranges multiple behavior tags linearly, and constructs an initial behavior sequence covering the complete daily cycle.

[0100] After completing the time sorting, the system further identifies the continuity of spatial and behavioral characteristics between behaviors in the ordered sequence. Specifically, the system determines whether the time interval between two consecutive behaviors is less than the set behavior switching window threshold, for example, 10 minutes. If the interval is short and there is a logical connection between the spatial areas, such as from the bedroom to the bathroom, the two behaviors are considered to have a behavioral linkage relationship. The system combines such continuous behaviors into behavioral segments and records their order of appearance and spatial transfer paths in a day.

[0101] After all ternary structures are organized into intra-day behavior segments, the system counts the user's behavior sequences over multiple consecutive days and extracts stable and recurring behavior patterns. For example, if a user exhibits the sequence of "getting up at 07:00 - entering the bathroom at 07:10 - entering the kitchen at 07:25" on multiple weekdays, the system identifies this sequence as a typical morning behavior pattern and records it as part of the behavior structure sequence. The construction of the behavior structure sequence includes not only the structural properties of each behavior, but also its relative position and previous and next dependencies in the entire day's behavior, ultimately forming a set of sequence expression results with temporal sequence, spatial distribution, and behavior concentration, which are used to support subsequent cycle identification and individual model construction tasks. This process significantly enhances the system's ability to understand the user's complex behavior flow and provides sufficient context for identifying fine-grained behavior variations.

[0102] In a preferred embodiment of the present invention, the model building unit includes:

[0103] The candidate behavior unit is used to repeatedly match the ternary structure data in the behavior structure sequence, and classify the time periods and spatial regions with more than a preset overlap rate in multiple consecutive days and the ternary structures with behavior characteristics exceeding the preset overlap rate as the same candidate behavior segment, thereby obtaining a candidate behavior segment set;

[0104] a periodicity identification unit, configured to perform statistical calculations on the occurrence frequency and time concentration of each candidate behavior segment based on the candidate behavior segment set, and identify the behavior segment as a periodic behavior when its occurrence frequency is greater than a preset frequency threshold and its time concentration is less than a preset concentration threshold;

[0105] The rule generation unit is used to reconstruct the pattern of the time period field, spatial area field and behavior feature field in the periodic behavior, extract the average occurrence time, main area coordinates and stable behavior features, and obtain the structured behavior set;

[0106] The physical modeling unit is used to determine the user's behavior pattern in a specific time period and spatial area based on the structured behavior set, and obtain the individual behavior physical model.

[0107] In an embodiment of the present invention, the candidate behavior unit is used to perform repetitive matching on the ternary structure data in the behavior structure sequence, and classify the time periods and spatial regions with a preset overlap rate for multiple consecutive days and the ternary structures with behavior characteristics exceeding the preset overlap rate as the same candidate behavior segment, thereby obtaining a candidate behavior segment set. By identifying segments with high similarity in the behavior sequence, the stable behavior pattern in the user's daily life is determined; the period identification unit is used to perform statistical calculations on the occurrence frequency and time concentration of each candidate behavior segment based on the candidate behavior segment set. When its occurrence frequency is greater than the preset frequency threshold and its time concentration is less than the preset concentration threshold, the behavior segment is identified as a periodic behavior. By setting dual conditional restrictions of frequency and concentration, the system has It effectively avoids misjudgments caused by occasional behavioral fluctuations and improves the scientificity and stability of periodic behavior identification; the rule generation unit is used to reconstruct the patterns of the time period field, spatial area field and behavioral feature field in periodic behavior, extract the average occurrence time, main area coordinates and stable behavior characteristics, and obtain a structured behavior set, realizing the abstraction and standardization of periodic behavior, so that complex and changeable behavior sequences are converted into universal rule templates, which facilitates rapid system call and matching; the physical modeling unit is used to determine the user's behavior pattern in a specific time period and spatial area based on the structured behavior set, and obtain an individual behavior physical model. By constructing a complete and dynamically updated individual behavior physical model, it solves the problem that traditional models rely on static behavior templates and lack personalized expression.

[0108] The rule generation unit is used to reconstruct the time period field, spatial region field, and behavior feature field in the periodic behavior, extract the average occurrence time, main region coordinates, and stable behavior features, and obtain a structured behavior set, which specifically includes:

[0109] After receiving the periodic behavior output by the periodic identification unit, the system first parses the time period field in the periodic behavior. In multiple periodic behavior fragments, the system extracts the starting time and duration of all occurrence time periods, performs statistical averaging on the starting time, forms a unified average occurrence time parameter, and calculates the standard deviation of the fluctuation range of the duration. This value is set as the time tolerance for the time matching threshold setting in subsequent behavior judgment. In this process, the system uses a sliding window mechanism to densely sample the behavior features in different time periods to ensure that the average value is representative and avoid deviation from the user's true behavior concentration due to interference from extreme values.

[0110] Next, the system performs aggregation operations on the spatial area fields involved in the periodic behavior segments. By analyzing the spatial location information associated with different behavior segments, the system identifies the area with the highest frequency of occurrence as the main area coordinates of the periodic behavior. If the same periodic behavior may involve switching of multiple spatial areas or parallel operations, such as a short movement between the kitchen and the living room, the system will use the location density function to determine the core center of the behavioral activity and define the area as the main area of the behavior, rather than simply based on the number statistics, thereby enhancing the robustness of spatial recognition. In the process of determining the main area coordinates, the system also maps the sensor number or geographic tag to the spatial coordinate identifier to ensure the consistency of the model structure and the callability of subsequent algorithm modules.

[0111] In terms of behavioral feature fields, the system focuses on extracting behavioral stability parameters reflected in periodic behaviors, including but not limited to characteristic dimensions such as movement intensity, behavior frequency, and physiological indicators. For these numerical features, the system uses multiple rounds of aggregation and filtering algorithms to eliminate noise and occasional deviation data, retaining a high-confidence set of behavioral features, and then extracts the most representative and least variable stable parameters of the behavior in the periodic sequence through methods such as mean fusion or principal component analysis as behavioral stability features.

[0112] Finally, the average occurrence time, main area coordinates and stable behavior characteristics of the processing results of the above three dimensions are structured and encapsulated to form a structured behavior set in a unified format. This not only can standardize the expression of users' periodic behavior, but also provides matching templates and reference benchmarks for subsequent deviation identification, trend judgment and anomaly detection, thereby greatly improving the reliability and adaptability of the entire monitoring system in practical applications.

[0113] The physical modeling unit is used to determine the user's behavior pattern in a specific time period and spatial area based on the structured behavior set, and obtain an individual behavior physical model, which specifically includes:

[0114] After receiving the structured behavior set, the system first performs behavioral template processing on each structured rule. Based on the time period field in the structured behavior set, the system maps the average occurrence time and its tolerance range to a unified time axis to construct the user's daily behavior time frame. The system sorts all rule entries in chronological order to form a complete time period sequence, and builds a daily concentration map of user behavior based on this. This map can clearly describe the distribution of user activities in each time period of a day, and is an important basis for subsequent model matching and deviation calculations.

[0115] Subsequently, the system spatially maps the spatial area fields in the structured behavior concentration, maps the main area coordinates with the physical location in the actual monitoring environment, and constructs a time-space two-dimensional behavior scenario model. If multiple behavior rules are continuous or overlapping, the system will represent them in the form of a graph structure, where each node represents a specific behavior scenario, such as kitchen activities within a certain time period, and the edge represents the transition relationship between behaviors, such as the movement path from the bedroom to the kitchen. On this basis, the system uses a graph traversal algorithm to identify users' common spatial movement paths and their time series patterns, thereby enhancing its understanding of dynamic behavior flows.

[0116] For the behavioral feature field, the system injects stable behavioral feature parameters into each time-space node, and performs cluster analysis based on the correlation between parameter features to identify typical behavioral pattern combinations. For example, moderate-intensity activities and high-frequency physiological parameter fluctuations within a certain time period can be associated with exercise behavior patterns. The system will mark them as specific behavioral labels and uniformly encode them. The system will also perform frequency statistics and concentration fitting on the combinations of these behavioral labels to construct multiple typical behavioral units, and on this basis, build a periodic behavior template library.

[0117] Finally, the above time series, spatial trajectories and behavioral parameters are integrated to form a unified individual behavior physical model. This individual behavior physical model is essentially a multidimensional vector structure with multiple dimensions such as time dimension, space dimension, behavior dimension and feature dimension. It is used to represent the standardized behavioral performance of users in specific time and space scenarios. This individual behavior physical model can not only be used for comparison and deviation judgment of target behaviors, but also supports trend prediction, anomaly assessment and dynamic model updating during the behavior evolution process, thereby achieving high reliability, strong adaptability and interpretability of individual behavior modeling in long-term monitoring scenarios.

[0118] In a preferred embodiment of the present invention, the deviation calculation module includes:

[0119] A first calculation unit is used to compare the target feature data with the time field, space field, and behavior field of the corresponding behavior tag in the individual behavior physical model, calculate the difference between the behavior feature set, obtain the behavior deviation indicator set, and calculate the deviation degree based on the difference;

[0120] An anomaly identification unit is used to compare the degree of deviation with a preset deviation threshold, and when the degree of deviation is greater than the preset deviation threshold, determine that the behavior label corresponding to the target feature data is a suspicious behavior label;

[0121] The behavior recording unit is used to monitor suspicious behavior tags, record the first occurrence time, last occurrence time, consecutive occurrence times and spatial distribution range of suspicious behavior tags, and obtain an observation tag group.

[0122] In an embodiment of the present invention, a first calculation unit is used to compare the target feature data with the time field, spatial field, and behavior field of the corresponding behavior label in the individual behavior physical model, calculate the difference between the behavior feature set, and obtain a behavior deviation index set. Based on the calculated deviation degree, it can concisely and accurately reflect whether the behavior has an essential deviation from the historical model, effectively enhancing the system's judgment efficiency and response speed to behavioral anomalies; the anomaly identification unit is used to compare the deviation degree with a preset deviation threshold. When the deviation degree is greater than the preset deviation threshold, the behavior label corresponding to the target feature data is determined to be a suspicious behavior label. It can quickly respond to behaviors with significant deviation characteristics and screen out high-risk behavior events, thereby achieving preliminary filtering and extraction from normal behavior to abnormal behavior; the behavior recording unit is used to monitor suspicious behavior labels, record the first occurrence time, last occurrence time, number of consecutive occurrences, and spatial distribution range of the suspicious behavior labels, and obtain an observation label group. It can systematically save the evolution path and spatiotemporal distribution of the target behavior, thereby providing solid data support for trend behavior analysis and abnormal evolution pattern recognition.

[0123] In a preferred embodiment of the present invention, the first calculation unit includes:

[0124] The second calculation unit is configured to obtain a first deviation index item based on the product relationship between the behavior time period, behavior frequency, and behavior concentration of the user's recent behavior, combined with the inhibitory effect of behavior concentration on time-frequency offset; obtain a second deviation index item based on the interaction between the behavior duration and the spatial location of the user's recent behavior, combined with the adjustment function of spatial offset on duration fluctuation; and obtain a third deviation index item based on the joint relationship between the behavior time period, behavior concentration, and behavior duration of the user's recent behavior, combined with the degree of interference of frequency offset on this joint relationship;

[0125] The first deviation index item, the second deviation index item, and the third deviation index item are integrated to obtain the degree of deviation of the user's recent behavior from the individual behavior physical model.

[0126] In an embodiment of the present invention, a first deviation index is derived based on the multiplicative relationship between the time period, frequency, and concentration of a user's recent behavior, combined with the inhibitory effect of behavior concentration on time-frequency offset. This effectively identifies concentration misalignment anomalies caused by physiological changes. A second deviation index is derived based on the interaction between the duration and spatial location of a user's recent behavior, combined with a regulation function of spatial offset on duration fluctuations. Through coupled analysis of behavior duration and spatial distribution, the system's precision in identifying spatial behavior anomalies is improved. A third deviation index is derived based on the joint relationship between the time period, concentration, and duration of a user's recent behavior, combined with the degree of interference of frequency offset on this joint relationship. This effectively reveals behavioral characteristics where concentration is stable but strong interference occurs, improving the timeliness and accuracy of the system in identifying sudden behavior. The first, second, and third deviation indexes are integrated to determine the degree of deviation of the user's recent behavior from the individual behavior physical model. This eliminates the judgment distortion caused by single-dimensional deviation and provides a stable and comprehensive anomaly assessment reference when multiple behavioral changes coexist.

[0127] The calculation formula for the degree of deviation is:

[0128] ,

[0129] in, is the degree of deviation of the user’s recent behavior, is the difference between the user's recent behavior and the behavior time period of the individual behavior physical model. , The time period during which the user's recent behavior occurred. is the behavior occurrence time period of the individual behavior physical model, is the difference between the user's recent behavior and the behavior frequency of the individual behavior physical model, , The frequency of the user's recent behavior. is the frequency of behavior occurrence of the individual behavior physical model, is the difference between the user's recent behavior and the behavior time concentration of the individual behavior physical model, , is the time concentration of the user's recent behavior, is the behavior time concentration of the individual behavior physical model, is the difference between the duration of the user's recent behavior and the behavior of the individual behavior physical model, , The duration of the user's recent behavior. is the duration of the behavior of the individual behavior physical model, is the difference between the coordinates of the behavior space area of the user's recent behavior and the individual behavior physical model, , is the coordinate of the behavioral space area of the user's recent behavior, is the coordinate of the behavior space region of the individual behavior physical model.

[0130] Among them, when calculating the degree of deviation of behavior, the difference of each behavior field is uniformly normalized, that is, the baseline value of the corresponding data in the individual behavior physical model is used for normalization, and the time, frequency, space and other differences between the current behavior and the individual behavior physical model are divided by the standard value of the field in the individual behavior physical model, thereby converting them into dimensionless offset ratios with consistent units and unified dimensions.

[0131] In a preferred embodiment of the present invention, the anomaly detection module includes:

[0132] A trend extraction unit is used to extract the continuous occurrence records of suspicious behavior tags in the observation tag group in the time dimension to obtain a trend sample set;

[0133] A trend analysis unit is used to calculate the average frequency and average deviation of suspicious behavior labels in continuous records based on the trend sample set to obtain a trend feature indicator set;

[0134] The judgment rule unit is used to judge whether the average frequency of the trend characteristic indicator set exceeds the frequency of behavior of the same type of label in the individual behavior physical model, and at the same time judge whether the average deviation degree exceeds the preset deviation threshold. If any judgment result is yes, it is determined that the behavior label does not meet the absorption condition;

[0135] The abnormality judgment unit is used to judge whether the suspicious behavior label has a trend reversal within a continuous time window. When there is no trend reversal, the suspicious behavior label is confirmed to be an abnormal behavior.

[0136] In an embodiment of the present invention, a trend extraction unit is used to extract the continuous occurrence records of suspicious behavior labels in the observation label group in the time dimension to obtain a trend sample set, which can screen out trend behaviors with time evolution characteristics from a large number of suspicious behavior labels, avoiding misjudgment of short-term sporadic behaviors; a trend analysis unit is used to calculate the average frequency and average deviation degree of suspicious behavior labels in continuous records based on the trend sample set to obtain a trend feature index set, and convert the trend behavior from perceptual observation into a set of indicators that can be processed numerically through quantitative operations, so that the subsequent judgment logic has a quantitative basis; a judgment rule unit is used to judge whether the average frequency in the trend feature index set exceeds The frequency of behavior of similar labels in the individual behavior physical model is obtained, and at the same time, it is judged whether the average deviation degree exceeds the preset deviation threshold. When any judgment result is yes, it is judged that the behavior label does not meet the absorption condition. Through the double-threshold judgment mechanism, it avoids misidentifying behaviors that temporarily rise due to environmental changes as abnormalities, and can also identify trend behaviors that change slowly but deviate from stability; the abnormality judgment unit is used to judge whether the suspicious behavior label has a trend reversal within a continuous time window. When there is no trend reversal, the suspicious behavior label is confirmed to be abnormal behavior, which effectively increases the system's buffer for abnormal behavior judgment, avoids false alarms caused by premature abnormality due to temporary deviation, and improves judgment stability.

[0137] The abnormality judgment unit is used to judge whether the suspicious behavior label has a trend reversal within a continuous time window. If there is no trend reversal, the suspicious behavior label is confirmed to be abnormal behavior. Specifically, it includes:

[0138] After receiving the result from the judgment rule unit, that is, a suspicious behavior label is judged to not meet the absorption conditions, the trend reversal analysis phase of the label is entered. This operation process relies on systematic tracking and statistical analysis of the evolution trend of the behavior in a continuous time window. The goal is to determine whether the behavior has a self-correction trend, that is, whether its behavioral performance will return to the normal fluctuation range of the individual behavior model in the subsequent time period, so as to avoid misjudgment due to short-term behavioral disturbances.

[0139] The system first sets a trend monitoring time window, which generally covers several consecutive days after the user behavior occurs, such as 7 days, 10 days or 14 days. In this window, all records of the suspicious behavior label are extracted, and a trend trajectory is constructed in chronological order. For each record, the system obtains its corresponding behavior frequency and deviation degree value, and compares and analyzes it with the average value in the trend sample set. If in this window, the system monitors that the frequency of the behavior shows a continuous downward trend, for example, the behavior frequency decreases day by day within three days, or is lower than 50% of the trend average frequency at most time points, the system believes that the intensity of the behavior is weakening; at the same time, if the degree of behavior deviation also shows a downward trend in subsequent records, especially if it is lower than the deviation threshold set by the individual behavior physical model for several consecutive times, it indicates that the behavior deviation state has gradually returned to the model baseline range.

[0140] Furthermore, the system performs trend fitting on the frequency curve and deviation curve, calculating their slope and fluctuation range. If the overall trend slope is negative and the fluctuation amplitude is lower than the set variation tolerance, the behavior is determined to have experienced a trend reversal. At this point, the system retains the label in an observation state and does not perform abnormal processing. However, if any of the above trends do not meet the reversal criteria, the system confirms that the behavior has maintained a high frequency or sustained deviation for a certain period of time and has not shown a self-repair trend, then the behavior label is identified as abnormal behavior, marked as an object requiring suppression or intervention, and passed to the behavior suppression module for subsequent processing. This operational strategy ensures that the system has a certain degree of flexible judgment space in the abnormal behavior identification process, avoids mistakenly classifying behavioral fluctuations as abnormal, and improves the judgment stability of the health monitoring system and the reliability of clinical intervention.

[0141] In a preferred embodiment of the present invention, the behavior suppression module includes:

[0142] The inhibition factor unit is used to analyze the minimum confidence level that needs to be met when the abnormal behavior is absorbed into the individual behavior physical model based on the deviation degree and duration of the abnormal behavior, and obtain the behavior inhibition factor;

[0143] The absorption judgment unit is used to compare the behavior inhibition factor with the absorption trigger strength of the individual behavior physical model. When the behavior inhibition factor is greater than the absorption trigger strength, the abnormal behavior does not meet the absorption conditions;

[0144] An absorption adjustment unit, configured to dynamically restrict the behavior label associated with the abnormal behavior in the individual behavior physical model, thereby increasing the duration of the update window of the behavior label in the individual behavior physical model;

[0145] The behavior blocking unit is used to add a blocking status mark to the abnormal behavior label that does not meet the absorption conditions, and store the behavior label, its deviation degree, and behavior inhibition factor in the behavior adjustment data.

[0146] In an embodiment of the present invention, the inhibition factor unit is used to analyze the minimum confidence strength that must be met when the abnormal behavior is absorbed into the individual behavior physical model based on the deviation degree and duration of the abnormal behavior, and obtain the behavior inhibition factor, thereby achieving a quantitative expression of the confidence level of the abnormal behavior, so that the system can make a judgment based on quantitative indicators on whether the behavior should be absorbed by the model; the absorption judgment unit is used to compare the behavior inhibition factor with the absorption trigger strength of the individual behavior physical model. When the behavior inhibition factor is greater than the absorption trigger strength, the abnormal behavior does not meet the absorption conditions, ensuring that only behaviors that meet the requirements of small deviation and stability will have a structural impact on the individual behavior physical model. The absorption adjustment unit is used to dynamically restrict the behavior labels related to the abnormal behavior in the individual behavior physical model, increase the time length of the update window of the behavior label in the individual behavior physical model, and increase the observation period of behavior evolution by limiting the model's premature acceptance of potential abnormal behaviors, reducing the probability of false absorption while retaining tolerance for sudden behaviors; the behavior blocking unit is used to add a blocking status mark to the abnormal behavior label that does not meet the absorption conditions, and store the behavior label and its deviation degree and behavior inhibition factor in the behavior regulation data, building a blacklist mechanism for abnormal behavior for the system to ensure that the behavior will not repeatedly trigger model false updates in the short term.

[0147] The absorption adjustment unit is used to dynamically limit the behavior label related to the abnormal behavior in the individual behavior physical model, thereby increasing the duration of the update window of the behavior label in the individual behavior physical model. Specifically, the absorption adjustment unit includes:

[0148] The system will first retrieve the behavior structure unit in the individual behavior physical model that has a structural mapping relationship with the behavior label based on the abnormal behavior label identified by the previous module and combined with the current status information of the label, such as the frequency of behavior occurrence, degree of deviation, duration, etc. The behavior structure unit usually contains ternary field information, namely the time period field, spatial area field and behavior feature field. After establishing a correspondence between the ternary structure and the historical behavior sequence in the model, the system will determine whether the label has the identity of a candidate segment with periodic behavior characteristics, and check its behavior update window parameters formed during the model construction phase.

[0149] Next, the system will compare the inhibitory factor strength previously calculated by the inhibitory factor unit with the default absorption update strategy of the current behavior label in the individual model; if the inhibitory factor is greater than the model absorption trigger threshold, the system will execute the window parameter adjustment mechanism, that is, extend the observation period of the behavior label in the individual model. The observation period is defined by the system as the time range in which the label must meet the "low deviation-high persistence-stable concentration" conditions. In actual operation, the system will overwrite the update time threshold of the behavior label in the behavior model. For example, if the original update period is 3 days, it will be adjusted to 7 days, 10 days or even longer. The specific extension value is dynamically calculated in proportion to the inhibitory factor strength. The higher the inhibitory factor, the longer the window, thereby reducing the risk of it being quickly updated to the model structure.

[0150] While the update window is extended, the system will also synchronously reduce the active weight of the label in the model structure sequence, that is, by adding a dynamic scaling factor to the behavior frequency item and concentration distribution item of the label, so that its weight in the next round of periodic behavior recognition is reduced, avoiding the update mechanism from being accidentally triggered due to short-term frequent occurrences. The entire process does not modify the original label content, but rather realizes the regulation of the behavior absorption path through dynamic parameter restrictions, thereby forming a gentle and continuous behavior structure restriction method, which not only retains the possibility of the label continuing to be observed, but also effectively curbs its potential risk of misleading the evolution of the model, ensuring the long-term stability of the individual behavior model and the accuracy of behavior discrimination.

[0151] In a preferred embodiment of the present invention, the inhibitory factor unit comprises:

[0152] The suppression factor calculation unit is used to calculate the impact of changes in the time distribution and trust of abnormal behavior based on the logarithm of the time offset of the abnormal behavior and its coverage ratio in the individual behavior physical model, thereby obtaining the main control factor term; calculate the immediate interference capability of short-term abnormal behavior on the individual behavior physical model based on the duration difference of the abnormal behavior, thereby obtaining the linear adjustment factor term; calculate the overall suppression amplitude based on the ratio of the concentration offset value to the model coverage, and the ratio of the frequency offset value to the behavior duration, thereby obtaining the balance adjustment term;

[0153] The main control factor item, linear adjustment factor item and balance adjustment item are integrated to obtain the behavioral inhibition factor of abnormal behavior.

[0154] In an embodiment of the present invention, the inhibition factor calculation unit is used to calculate the impact of changes in the time distribution and trust of abnormal behavior based on the logarithmic value of the time offset of the abnormal behavior and its coverage ratio in the individual behavior physical model, and obtain a main control factor item, which can avoid the problem of imbalance in inhibition strength caused by extreme time differences and improve the accuracy of distinguishing time-sensitive abnormal behaviors; according to the duration difference of the abnormal behavior, the immediate interference ability of short-term abnormal behavior on the individual behavior physical model is calculated to obtain a linear adjustment factor item, and the time difference is converted into a quantitative indicator of interference ability, which can achieve effective evaluation of short-term or long-term abnormalities; according to the ratio of the concentration offset value to the model coverage, and the ratio of the frequency offset value to the behavior duration, the overall inhibition amplitude is calculated to obtain a balance adjustment item, and a quantitative deviation ratio is formed by the concentration and frequency, so as to accurately evaluate the impact of abnormal behavior on the stability of the model structure; the main control factor item, the linear adjustment factor item and the balance adjustment item are integrated to obtain the behavioral inhibition factor of the abnormal behavior, and a unified evaluation scale is established in the behavior space, time dimension and concentration characteristics, overcoming the problem of excessive reliance on single-dimensional data in traditional systems.

[0155] The calculation formula of the behavior inhibition factor is:

[0156] ,

[0157] in, A behavioral inhibitor of abnormal behavior. is the degree of deviation of abnormal behavior, is the coverage of behavior labels of the same type as abnormal behavior in the individual behavior physical model, is the duration of abnormal behavior, is the coefficient.

[0158] in, 、 、 and The data is the same as the calculation formula of the deviation degree. is the difference between the time period of abnormal user behavior and the time period of individual behavior physical model. , The time period during which the user's abnormal behavior occurs. is the behavior occurrence time period of the individual behavior physical model, is the difference between the frequency of abnormal user behavior and the behavior of the individual behavior physical model, , The frequency of abnormal user behavior. is the frequency of behavior occurrence of the individual behavior physical model, is the difference between the behavior time concentration of the user's abnormal behavior and the individual behavior physical model, , is the time concentration of abnormal user behavior, is the behavior time concentration of the individual behavior physical model, is the difference between the duration of the user's abnormal behavior and the duration of the individual behavior physical model. , The duration of the user's abnormal behavior. is the duration of the behavior of the individual behavior physical model.

[0159] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. A personal health abnormal behavior monitoring system integrating unsupervised learning, characterized by: The system comprises: The unsupervised module is used to obtain historical behavior data of users and perform clustering processing on it, dividing the historical behavior data into multiple behavior labels to obtain behavior label data; The individual model module is used to extract the feature vector of each behavior label based on the behavior label data, obtain sample feature data, and establish an individual behavior physical model based on it; The target behavior module is used to obtain the user's recent target behavior data, cluster it, and extract features to obtain target feature data; The deviation calculation module is used to input the target feature data into the individual behavior physical model, calculate the deviation degree of the user's recent behavior, and monitor and record the abnormal behavior labels whose deviation degree exceeds the preset deviation threshold to obtain an observation label group; The anomaly detection module is used to determine whether the observed label group meets the conditions absorbed by the individual behavior physical model. If the result is no, the behavior label is determined to be abnormal behavior; The behavior inhibition module is used to calculate the behavior inhibition factor based on the deviation degree and duration of abnormal behavior, and dynamically adjust the individual behavior physical model based on it to limit abnormal behavior from being identified as normal behavior; The behavior inhibition module includes: The inhibition factor unit is used to analyze the minimum confidence level that needs to be met when the abnormal behavior is absorbed into the individual behavior physical model based on the deviation degree and duration of the abnormal behavior, and obtain the behavior inhibition factor; The absorption judgment unit is used to compare the behavior inhibition factor with the absorption trigger strength of the individual behavior physical model. When the behavior inhibition factor is greater than the absorption trigger strength, the abnormal behavior does not meet the absorption conditions; An absorption adjustment unit, configured to dynamically restrict the behavior label associated with the abnormal behavior in the individual behavior physical model, thereby increasing the duration of the update window of the behavior label in the individual behavior physical model; A behavior blocking unit is used to add a blocking status mark to abnormal behavior labels that do not meet the absorption conditions, and store the behavior label, its deviation degree, and behavior inhibition factor in the behavior regulation data; The inhibitory factor unit comprises: The suppression factor calculation unit is used to calculate the impact of changes in the time distribution and trust of abnormal behavior based on the logarithm of the time offset of the abnormal behavior and its coverage ratio in the individual behavior physical model, thereby obtaining the main control factor term; calculate the immediate interference capability of short-term abnormal behavior on the individual behavior physical model based on the duration difference of the abnormal behavior, thereby obtaining the linear adjustment factor term; calculate the overall suppression amplitude based on the ratio of the concentration offset value to the model coverage, and the ratio of the frequency offset value to the behavior duration, thereby obtaining the balance adjustment term; The main control factor item, linear adjustment factor item and balance adjustment item are integrated to obtain the behavioral inhibition factor of abnormal behavior.

2. The personal health abnormal behavior monitoring system integrating unsupervised learning according to claim 1 is characterized in that: The unsupervised module includes: A time segmentation unit is used to segment and process the historical behavior data, splitting the historical behavior data into multiple time segments according to a preset time window to obtain a time segment data set; A feature extraction unit is used to extract features from the time segment data set, extract the user's stay rate, spatial transfer frequency and related physiological values in different spatial areas within each time segment, and obtain a feature data set; The clustering unit is used to perform clustering processing on the preliminary feature data set, classifying time segments with similar features into the same cluster to obtain clustering result data; The behavior mapping unit is used to perform behavior analysis on each cluster in the clustering result data, identify the significant features of the cluster in spatial behavior and physiological parameters, and map the significant features into behavior labels to obtain behavior label data.

3. The personal health abnormal behavior monitoring system integrating unsupervised learning according to claim 2 is characterized in that: The behavior mapping unit includes: A statistical analysis unit is used to perform statistics on the time segment data contained in each cluster in the clustering result data, and respectively calculate the spatial behavior parameters and physiological parameters of the users in each spatial area in the cluster to obtain a cluster characteristic index set; The significant feature identification unit is used to standardize the spatial behavioral parameters and physiological parameters in the cluster feature index set, and screen out the spatial regions and physiological value intervals with significant features according to the preset behavioral feature threshold to obtain the significant feature set; The label classification unit is used to match the set of significant features with the preset behavior pattern, select labels according to the spatial area and physiological value range, determine the behavior label type that best matches the cluster, and bind the behavior label type to the cluster to obtain behavior label data.

4. The personal health abnormal behavior monitoring system integrating unsupervised learning according to claim 3 is characterized in that: The individual model module includes: The behavior decomposition unit is used to perform sample analysis on each behavior tag in the behavior tag data, extract the behavior occurrence time period, behavior duration, behavior spatial area, behavior occurrence frequency and behavior time concentration corresponding to the tag, and obtain the behavior feature set; The behavior structure unit is used to construct ternary structure data for each behavior tag in the behavior feature unit set. The ternary structure data includes a time period field, a spatial region field, and a behavior feature field. The behavior sequence unit is used to combine all the ternary structure data in chronological order, identify the user's behavior patterns in different time periods and different spatial areas, and obtain the behavior structure sequence; The model building unit is used to define the ternary structure with repeated patterns as periodic behavior according to the behavior structure sequence, and convert it into a regular expression form to generate an individual behavior physical model.

5. The personal health abnormal behavior monitoring system integrating unsupervised learning according to claim 4 is characterized in that: The model building unit includes: The candidate behavior unit is used to repeatedly match the ternary structure data in the behavior structure sequence, and classify the time periods and spatial regions with more than a preset overlap rate in multiple consecutive days and the ternary structures with behavior characteristics exceeding the preset overlap rate as the same candidate behavior segment, thereby obtaining a candidate behavior segment set; a periodicity identification unit, configured to perform statistical calculations on the occurrence frequency and time concentration of each candidate behavior segment based on the candidate behavior segment set, and identify the behavior segment as a periodic behavior when its occurrence frequency is greater than a preset frequency threshold and its time concentration is less than a preset concentration threshold; The rule generation unit is used to reconstruct the pattern of the time period field, spatial area field and behavior feature field in the periodic behavior, extract the average occurrence time, main area coordinates and stable behavior features, and obtain the structured behavior set; The physical modeling unit is used to determine the user's behavior pattern in a specific time period and spatial area based on the structured behavior set, and obtain the individual behavior physical model.

6. The personal health abnormal behavior monitoring system integrating unsupervised learning according to claim 5 is characterized in that: The deviation calculation module includes: A first calculation unit is used to compare the target feature data with the time field, space field, and behavior field of the corresponding behavior tag in the individual behavior physical model, calculate the difference between the behavior feature set, obtain the behavior deviation indicator set, and calculate the deviation degree based on the difference; An anomaly identification unit is used to compare the degree of deviation with a preset deviation threshold, and when the degree of deviation is greater than the preset deviation threshold, determine that the behavior label corresponding to the target feature data is a suspicious behavior label; The behavior recording unit is used to monitor suspicious behavior tags, record the first occurrence time, last occurrence time, consecutive occurrence times and spatial distribution range of suspicious behavior tags, and obtain an observation tag group.

7. The personal health abnormal behavior monitoring system integrating unsupervised learning according to claim 6 is characterized in that: The first computing unit includes: The second calculation unit is configured to obtain a first deviation index item based on the product relationship between the behavior time period, behavior frequency, and behavior concentration of the user's recent behavior, combined with the inhibitory effect of behavior concentration on time-frequency offset; obtain a second deviation index item based on the interaction between the behavior duration and the spatial location of the user's recent behavior, combined with the adjustment function of spatial offset on duration fluctuation; and obtain a third deviation index item based on the joint relationship between the behavior time period, behavior concentration, and behavior duration of the user's recent behavior, combined with the degree of interference of frequency offset on this joint relationship; The first deviation index item, the second deviation index item, and the third deviation index item are integrated to obtain the degree of deviation of the user's recent behavior from the individual behavior physical model.

8. The personal health abnormal behavior monitoring system integrating unsupervised learning according to claim 7 is characterized in that: The anomaly detection module includes: A trend extraction unit is used to extract the continuous occurrence records of suspicious behavior tags in the observation tag group in the time dimension to obtain a trend sample set; A trend analysis unit is used to calculate the average frequency and average deviation of suspicious behavior labels in continuous records based on the trend sample set to obtain a trend feature indicator set; The judgment rule unit is used to judge whether the average frequency of the trend characteristic indicator set exceeds the frequency of behavior of the same type of label in the individual behavior physical model, and at the same time judge whether the average deviation degree exceeds the preset deviation threshold. If any judgment result is yes, it is determined that the behavior label does not meet the absorption condition; The abnormality judgment unit is used to judge whether the suspicious behavior label has a trend reversal within a continuous time window. When there is no trend reversal, the suspicious behavior label is confirmed to be an abnormal behavior.

Images