Intelligent watch based on wearing behavior evolution modeling method and system

Abstract

This invention provides an evolutionary modeling method and system for smartwatches based on wearing behavior, applicable to the field of data acquisition and processing. By introducing a progressive change modeling mechanism for wearing behavior, this invention effectively addresses the problem of existing smartwatches focusing only on instantaneous signal quality anomalies and ignoring long-term wearing behavior evolution. Based on pre-detected wearing areas, it filters implicit feature parameters with physical significance. By judging whether the features have the ability to characterize the wearing behavior state, it avoids misusing noise or regional interference for behavior analysis. It constructs a wearing behavior feature sequence according to time sequence and uses distribution drift detection to analyze the progressive changes in feature mean, fluctuation, and stability, continuously depicting the long-term adjustment process of the user's watch-wearing habits. Furthermore, by using a reference baseline, it distinguishes between normal structural changes and abnormal changes, preventing long-term wearing changes from being misjudged as instantaneous anomalies.

Description

Technical Field

[0001] This invention relates to the field of data acquisition and processing, and in particular to an evolutionary modeling method and system for smartwatches based on wearing behavior. Background Technology

[0002] Existing smartwatches typically use optical, bioelectric, or inertial sensors to continuously monitor users' physiological parameters and behavioral status, and provide health assessments or status reminders based on this.

[0003] However, in long-term actual use, users' wearing behavior of smartwatches does not remain stable, but will gradually change with the use time, environmental conditions and personal habits. For example, slight shifts in the wearing position on the wrist, slow adjustments to the tightness of the fit, or long-term drifts in the wearing posture will not produce significant changes in a short period of time, but will accumulate over a long time scale. Summary of the Invention

[0004] This invention aims to address the current shortcomings of smartwatches, which only focus on anomalies in instantaneous signal quality and lack the ability to continuously model the long-term changes in wearing behavior. It provides a method and system for evolutionary modeling of smartwatches based on wearing behavior.

[0005] The present invention employs the following technical means to solve the technical problem: This invention provides an evolutionary modeling method for smartwatches based on wearing behavior, comprising: Based on the wearing area of ​​the user pre-detected by the smartwatch, the implicit feature parameters of the user are identified, wherein the implicit feature parameters specifically include signal amplitude stability features, periodic signal consistency features, and sensor signal correlation features; Determine whether the implicit feature parameters can reflect the wearing behavior status of the smartwatch; If possible, the wearing behavior feature sequence of the smartwatch is constructed according to the preset time sequence. The changing trend of the wearing behavior feature sequence is analyzed by using the preset distribution drift detection to obtain the gradual change information of the user's wearing behavior during use. The gradual change information specifically includes the slow shift of the feature mean, the continuous change of feature fluctuation, and the reduction and enhancement of feature stability. Determine whether the gradual change information is within the normal fluctuation range corresponding to the preset reference baseline; If so, an evolutionary model is constructed based on the progressive change information to characterize the user's wearing behavior. During the use process, the evolutionary model is periodically updated in a triggered manner. Abnormal physiological monitoring signals of the user are collected. Through the evolutionary model, the abnormal sources of the abnormal physiological monitoring signals are identified. Specifically, the abnormal sources include abnormalities caused by long-term changes in wearing behavior and abnormalities caused by changes in actual physiological state.

[0006] Furthermore, after the step of constructing the wearing behavior feature sequence of the smartwatch according to a preset time sequence, the method further includes: Based on the construction process of the wearing behavior feature sequence, the feature interruption segment of the user's interaction with the smartwatch is identified, wherein the feature interruption segment specifically includes taking off the watch, not wearing it for a long time, and insufficient monitoring conditions; Determine whether the feature interruption segment is marked during the construction process; If so, a behavioral inertia parameter representing the resistance to changes in wearing behavior is introduced into the wearing behavior feature sequence. Based on the behavioral inertia parameter, the degree of tilt of the user's wearing behavior in maintaining the original state within a continuous period is identified. Based on the degree of tilt, the change direction identifier of the wearing behavior feature unit between adjacent units is dynamically collected to generate the evolution direction of the user's wearing behavior in the feature space. Specifically, the evolution direction includes evolution towards a stable state and evolution towards an unstable state.

[0007] Furthermore, the step of obtaining information on the gradual changes in user wearing behavior during use also includes: Based on the changing relationship between adjacent wearing behavior units, candidate change segments are identified when the user's wearing behavior changes, wherein the candidate change segments are specifically time intervals with continuous observation value; Determine whether the candidate change segment has a cumulative effect, wherein the cumulative effect specifically means that it is continued or amplified in subsequent time periods; If not, the candidate change segment is marked as a local perturbation. During the change process, the coupling relationship between the change in wearing behavior characteristics and the stability parameter of wearing behavior is obtained. Based on the coupling relationship, the stage boundary of the change in wearing behavior is constructed. Specifically, the stage boundary includes the pre-change stage, the change transition stage, and the post-change stage.

[0008] Furthermore, the step of collecting the user's abnormal physiological monitoring signals and identifying the abnormal source of the abnormal physiological monitoring signals through the evolutionary model further includes: Based on the abnormality type characterization of the abnormal physiological monitoring signal pre-processed by the smartwatch, the time range of the abnormal physiological monitoring signal is collected, wherein the time range specifically includes the start time, duration and end time; Determine whether the time interval can be aligned with the temporal structure information in the evolutionary model; If so, the evolutionary state information of the smartwatch is identified. Based on the evolutionary state information, the abnormal physiological monitoring signal is structurally matched with the wearing behavior evolution process to obtain the degree of consistency between the abnormal physiological monitoring signal and the wearing behavior evolution process in terms of change dimensions, and a corresponding structural correlation is generated. Specifically, the evolutionary state information includes the current wearing behavior stage, the wearing stability level corresponding to the stage, the evolution direction of the stage, and the change rate of the stage. The change dimensions specifically include the change rhythm, change direction, and change stage.

[0009] Furthermore, the step of determining whether the implicit feature parameters can reflect the wearing behavior state of the smartwatch also includes: Based on the different wearing behavior states of the smartwatch, obtain the difference data of the implicit feature parameters within a preset time period; Determine whether the difference data detects a preset convergence behavior; If so, the characterization degradation information of the implicit feature parameters is identified. Based on the characterization degradation information, the assimilation influence parameters of the changing trend of the implicit feature parameters and the wearing adjustment behavior are collected, showing the same direction but not a responsive change. Based on the assimilation influence parameters, the slow unidirectional drift event of the smartwatch is constructed.

[0010] Furthermore, the step of determining whether the gradual change information is within the normal fluctuation range corresponding to the preset reference baseline also includes: Based on the change path of the progressive change information, the change type of the progressive change information is identified, wherein the change type specifically includes continuous approximation type, step transition type and nonlinear bending type; Determine whether the change type exhibits a preset boundary fitting behavior; If so, a preset alternative reference baseline is introduced for parallel verification. Based on the verification results of the parallel verification, the baseline instability interval of the progressive change information is dynamically generated. Specifically, the alternative reference baseline includes a reference baseline formed in adjacent wearing stages, a statistical baseline of similar user groups, and a short-term adaptive update baseline.

[0011] Furthermore, the step of identifying the user's implicit feature parameters based on the user's pre-detected wearing area by the smartwatch also includes: The smartwatch collects the pre-analyzed regional composition state of the wearing area, wherein the regional composition state specifically includes the contact tightness of the watch strap, the tilt state of the watch body relative to the skin, and the rotation state of the wearing area relative to the limb axis. Determine whether the state of the region is lower than a preset stability level; If so, the wearing area corresponding to the region composition state is marked as a structurally uncertain region, and based on the region composition state, a pool of hidden feature candidates that are related to the current wearing area structural state is dynamically activated.

[0012] This invention also provides an evolutionary modeling system for smartwatches based on wearing behavior, comprising: The identification module is used to identify the user's implicit feature parameters based on the wearing area pre-detected by the smartwatch. The implicit feature parameters specifically include signal amplitude stability features, periodic signal consistency features, and sensor signal correlation features. The judgment module is used to determine whether the implicit feature parameters can reflect the wearing behavior state of the smartwatch; The execution module is used to construct the wearing behavior feature sequence of the smartwatch according to a preset time sequence if possible, and to analyze the changing trend of the wearing behavior feature sequence using a preset distribution drift detection method to obtain the gradual change information of the user's wearing behavior during use. The gradual change information specifically includes the slow shift of the feature mean, the continuous change of feature fluctuation, and the reduction and enhancement of feature stability. The second judgment module is used to determine whether the gradual change information is within the normal fluctuation range corresponding to the preset reference baseline; The second execution module is configured to, if so, construct an evolutionary model to characterize the user's wearing behavior based on the progressive change information, periodically trigger updates to the evolutionary model during the use process, collect abnormal physiological monitoring signals of the user, and identify the abnormal source of the abnormal physiological monitoring signals through the evolutionary model, wherein the abnormal source specifically includes abnormalities caused by long-term changes in wearing behavior and abnormalities caused by changes in actual physiological state.

[0013] Furthermore, it also includes: The second identification module is used to identify the user's characteristic interruption segments on the smartwatch based on the construction process of the wearing behavior feature sequence, wherein the characteristic interruption segments specifically include taking off the watch, not wearing it for a long time, and insufficient monitoring conditions; The third judgment module is used to determine whether the feature interruption segment is marked during the construction process; The third execution module is used to, if so, introduce a behavioral inertia parameter representing the resistance to changes in wearing behavior into the wearing behavior feature sequence, identify the degree of tilt of the user's wearing behavior in maintaining the original state within a continuous period of time based on the behavioral inertia parameter, dynamically collect the change direction identifier of the wearing behavior feature unit between adjacent units based on the degree of tilt, and generate the evolution direction of the user's wearing behavior in the feature space, wherein the evolution direction specifically includes evolution towards a stable state and evolution towards an unstable state.

[0014] Furthermore, the execution module also includes: The identification submodule is used to identify candidate change segments when the user's wearing behavior changes based on the change relationship between adjacent wearing behavior units, wherein the candidate change segments are specifically time intervals with continuous observation value; The judgment submodule is used to determine whether the candidate change segment has a cumulative effect, wherein the cumulative effect is specifically that it is continued or amplified in subsequent time periods; The execution submodule is used to mark the candidate change segment as a local perturbation if no, obtain the coupling relationship between the change of wearing behavior characteristics and the wearing behavior stability parameter during the change process, and construct the stage boundary of the wearing behavior change based on the coupling relationship. The stage boundary specifically includes the pre-change stage, the change transition stage and the post-change stage.

[0015] This invention provides an evolutionary modeling method and system for smartwatches based on wearing behavior, which has the following beneficial effects: This invention effectively addresses the problem of existing smartwatches focusing only on instantaneous signal quality anomalies and ignoring the long-term evolution of wearing behavior by introducing a progressive change modeling mechanism for wearing behavior. Based on the pre-detected wearing area, it filters implicit feature parameters with physical significance. By judging whether the features have the ability to characterize the state of wearing behavior, it avoids misusing noise or regional interference for behavior analysis. It constructs a sequence of wearing behavior features in chronological order and uses distribution drift detection to analyze the progressive changes in the mean, fluctuation, and stability of the features. This can continuously depict the long-term adjustment process of the user's watch-wearing habits. Furthermore, by judging with a reference baseline, it distinguishes between normal structural changes and abnormal changes, preventing long-term wearing changes from being misjudged as instantaneous anomalies. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating an embodiment of the evolutionary modeling method for smartwatches based on wearing behavior according to the present invention. Figure 2 This is a structural block diagram of an embodiment of the smartwatch evolution modeling system based on wearing behavior according to the present invention. Detailed Implementation

[0017] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention. The realization of the purpose, functional features, and advantages of the invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings.

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Reference Appendix Figure 1 The present invention provides an evolutionary modeling method for a smartwatch based on wearing behavior, comprising: S1: Based on the user's pre-detected wearing area by the smartwatch, identify the user's implicit feature parameters, wherein the implicit feature parameters specifically include signal amplitude stability features, periodic signal consistency features, and sensor signal correlation features; S2: Determine whether the implicit feature parameters can reflect the wearing behavior state of the smartwatch; S3: If possible, construct the wearing behavior feature sequence of the smartwatch according to the preset time sequence, and use the preset distribution drift detection to analyze the changing trend of the wearing behavior feature sequence to obtain the gradual change information of the user's wearing behavior during use. The gradual change information specifically includes the slow shift of the feature mean, the continuous change of feature fluctuation, and the reduction and enhancement of feature stability. S4: Determine whether the gradual change information is within the normal fluctuation range corresponding to the preset reference baseline; S5: If so, an evolutionary model for characterizing the user's wearing behavior is constructed based on the progressive change information. The evolutionary model is periodically updated during the use process. Abnormal physiological monitoring signals of the user are collected. The abnormal sources of the abnormal physiological monitoring signals are identified through the evolutionary model. The abnormal sources specifically include abnormalities caused by long-term changes in wearing behavior and abnormalities caused by changes in actual physiological state.

[0020] In this embodiment, the system identifies the user's implicit feature parameters based on the pre-detected wearing area of ​​the smartwatch. These implicit feature parameters specifically include signal amplitude stability characteristics, periodic signal consistency characteristics, and sensor signal correlation characteristics. The system then determines whether these implicit feature parameters can reflect the smartwatch's wearing behavior and executes corresponding steps accordingly. For example, if the system determines that the user's implicit feature parameters cannot reflect the smartwatch's wearing behavior, the system considers the current implicit feature parameters insufficient to characterize the user's wearing behavior. This could be due to factors such as wearing the watch too loosely, rotating it, or partially leaving it unsupported, causing the sensor signal to be primarily affected by random signals. If changes in movement or contact affect the user's behavior rather than actual wearing behavior, the system will pause the modeling of wearing behavior based on the implicit feature parameters, marking the current stage as an uncertain stage. Simultaneously, it will trigger a re-evaluation process of the wearing area and the effectiveness of the features, and adopt a conservative interpretation strategy for abnormal physiological monitoring signals to avoid misinterpreting signal fluctuations caused by wearing instability as real physiological abnormalities. This improves the monitoring reliability of the smartwatch during long-term use. For example, when the system determines that the user's implicit feature parameters can reflect the smartwatch's wearing behavior state, the system will consider the current implicit feature parameters sufficient to characterize the user's wearing behavior state. Based on this, the system constructs a wear behavior feature sequence of the smartwatch according to a pre-set time sequence. Using a pre-defined distribution drift detection method, it analyzes this sequence to obtain information on the gradual changes in user behavior during watch use. These gradual changes specifically include slow shifts in the feature mean, continuous changes in feature fluctuations, and a decrease in feature stability. By constructing the wear behavior feature sequence according to a pre-set time sequence and introducing distribution drift detection for analysis, the system no longer focuses solely on signal changes at a single moment or within a short period, but can continuously capture slow shifts in the feature mean, long-term changes in feature fluctuations, and more. The system uses progressively decreasing or increasing characteristic stability information to model the long-term evolution of user wearing habits during use. This progressive change information provides a structured reference for distinguishing changes in wearing behavior from signal anomalies caused by other factors. This allows the system to make comprehensive judgments based on the long-term trends of wearing behavior during subsequent physiological monitoring or anomaly analysis, avoiding misjudging normal wearing habit evolution as abnormal, thus improving the accuracy and interpretability of anomaly identification results. The system then determines whether these progressive changes are within the normal fluctuation range corresponding to a pre-set reference baseline to execute the appropriate steps.For example, when the system determines that the gradual changes in a user's wearing behavior during use are not within the normal fluctuation range corresponding to a pre-set reference baseline, the system will consider that the user's wearing behavior has undergone a continuous structural change, and the original reference baseline can no longer fully represent the wearing characteristics of the current usage stage. The system will mark the current wearing behavior as an out-of-baseline evolution state, suspend the normality judgment based on the original reference baseline, and trigger the update or expansion process of the wearing behavior evolution model. After the model stabilizes, a new reference baseline will be constructed. During the transition phase, a conservative interpretation strategy will be adopted for abnormal physiological monitoring signals to avoid misjudging the long-term evolution of wearing behavior as an abnormal situation, thereby improving the monitoring accuracy and system adaptability of the smartwatch during long-term use. Conversely, when the system determines that the gradual changes in a user's wearing behavior during use are within the normal fluctuation range corresponding to a pre-set reference baseline, the system will consider that the user's wearing behavior has not undergone a structural change, and the original reference baseline can still represent the wearing characteristics of the current usage stage. The system will construct an evolutionary model to represent the user's wearing behavior based on these gradual changes. During use, the system periodically triggers updates to the evolutionary model, collecting abnormal physiological monitoring signals from the user. The evolutionary model identifies the sources of these anomalies, specifically those caused by long-term changes in wearing behavior and those caused by changes in the user's actual physiological state. Under the premise of a valid reference baseline, the system utilizes gradual changes such as slow shifts in feature mean, continuous changes in feature fluctuations, and increases or decreases in feature stability to construct and periodically update the wearing behavior evolutionary model. This allows the model to gradually absorb subtle changes in the user's wearing habits without disrupting the original behavioral structure, achieving long-term, smooth, and adaptive modeling of wearing behavior. This improves the model's fit to real-world usage scenarios. Furthermore, with a stable and continuously updated wearing behavior evolutionary model, the system can jointly analyze abnormal physiological monitoring signals with the current wearing behavior evolutionary state to distinguish whether the anomaly is caused by long-term changes in wearing behavior or by changes in the user's actual physiological state. This approach effectively reduces the risk of false alarms caused by changes in wearing factors and improves the accuracy and interpretability of anomaly detection results.

[0021] It should be noted that, based on a preset time sequence, a wearing behavior feature sequence of the smartwatch is constructed. A preset distribution drift detection method is used to analyze the changing trend of the wearing behavior feature sequence, obtaining information on the gradual changes in user wearing behavior during use. Specifically: During smartwatch use, the system continuously collects the user's implicit characteristic parameters according to a preset time sequence, including signal amplitude stability characteristics, periodic signal consistency characteristics, and sensor signal correlation characteristics. The system arranges the characteristic values ​​calculated in each time period in chronological order to form a wearing behavior characteristic sequence, thereby reflecting the changing trend of the user's watch-wearing behavior over a continuous period of time. By constructing this sequence, the system can capture the slow changes of characteristics over time, rather than relying solely on instantaneous signal fluctuations, thus obtaining long-term wearing behavior evolution information. Subsequently, the system uses a preset distribution drift detection method to analyze the characteristic sequence to identify gradual shifts in the characteristic mean, continuous changes in characteristic fluctuations, and gradual changes in characteristic stability, such as increases or decreases. This analysis can distinguish between normal wearing habit evolution and abnormal wearing behavior, and provides a continuous and reliable basis for the construction of subsequent evolutionary models and the determination of abnormal physiological signals. Specific examples are as follows: For example, the system collects amplitude stability features hourly, resulting in a 6-hour data sequence [0.95, 0.94, 0.92, 0.90, 0.88, 0.87]. The system constructs a wearing behavior feature sequence based on these feature values ​​in chronological order, reflecting the slow decreasing trend of amplitude stability features during the 6-hour period. Simultaneously, distribution drift detection identifies that the mean of this sequence slowly decreases from 0.95 to 0.87, and the fluctuation range gradually increases from ±0.01 to ±0.02, indicating that the user's wearing method may have undergone slight adjustments during this period, such as wrist movement or changes in strap tightness. Through this gradual change information, the system can continuously update the wearing behavior evolution model, reflecting both long-term wearing habits and helping to distinguish abnormal signals caused by changes in wearing behavior or actual physiological states.

[0022] It should be added that, based on the aforementioned gradual change information, an evolutionary model is constructed to characterize the user's wearing behavior. During the usage process, the evolutionary model is periodically updated in a triggered manner, and abnormal physiological monitoring signals of the user are collected. Through the evolutionary model, the abnormal sources of these physiological monitoring signals are identified, specifically as follows: The system constructs an evolutionary model to characterize user wearing behavior based on the gradual changes in user wearing behavior. This model can depict the evolutionary patterns of user wearing behavior characteristics over different time periods, including the slow shift of the characteristic mean, the continuous changes in characteristic fluctuations, and the enhancement or reduction of characteristic stability. During watch use, the system uses a periodic trigger update mechanism to regularly update the evolutionary model, enabling the model to continuously reflect the long-term evolution of user wearing habits rather than remaining in an initial state. Simultaneously, the system continuously collects abnormal physiological monitoring signals from users and uses the evolutionary model to correlate these signals with the evolutionary trend of wearing behavior, thereby identifying the source of abnormal signals. Specifically, the system can distinguish whether abnormal signals are caused by long-term changes in user wearing behavior or by changes in actual physiological state, reducing false alarms caused by unstable wearing and improving the accuracy and interpretability of anomaly detection. Specific examples are as follows: For example, the system collects the user's amplitude stability characteristics, periodic signal consistency characteristics, and sensor signal correlation characteristics hourly over a week, and constructs an evolutionary model based on the aforementioned gradual change information. Assuming the amplitude stability characteristics show a slow downward trend during the week, gradually decreasing from 0.95 to 0.87, while the fluctuation range slightly expands, the system periodically (e.g., daily or every two days) updates the evolutionary model, incorporating these changes to reflect the long-term adjustment process of the user wearing the watch. Subsequently, the system collects an abnormal heart rate signal and, through evolutionary model analysis, finds that this signal is highly correlated with the downward trend of the amplitude stability characteristics, indicating that the abnormality may be partly caused by long-term changes in wearing behavior, rather than entirely by abnormal physiological state. Conversely, if a heart rate signal is outside the model's prediction range and the wearing behavior characteristics are stable, it can be determined that the abnormality mainly originates from the user's actual physiological state. This evolutionary model-based analysis method can continuously reflect changes in wearing behavior and improve the accuracy and reliability of abnormality detection.

[0023] In this embodiment, after step S3 of constructing the wearing behavior feature sequence of the smartwatch according to a preset time sequence, the method further includes: S301: Based on the construction process of the wearing behavior feature sequence, identify the feature interruption segment of the user's interaction with the smartwatch, wherein the feature interruption segment specifically includes taking off the watch, not wearing it for a long time, and insufficient monitoring conditions; S302: Determine whether the feature interruption segment is marked during the construction process; S303: If so, then a behavioral inertia parameter representing the resistance to changes in wearing behavior is introduced into the wearing behavior feature sequence. Based on the behavioral inertia parameter, the degree of tilt of the user's wearing behavior in maintaining the original state within a continuous period is identified. Based on the degree of tilt, the change direction identifier of the wearing behavior feature unit between adjacent units is dynamically collected to generate the evolution direction of the user's wearing behavior in the feature space. Specifically, the evolution direction includes evolution towards a stable state and evolution towards an unstable state.

[0024] In this embodiment, based on the construction process of the wearing behavior feature sequence, the system identifies the characteristic interruption segments of the user's interaction with the smartwatch. These interruption segments specifically include removing the watch, prolonged periods without wearing it, and insufficient monitoring conditions. The system then determines whether these interruption segments were marked during the construction process to execute corresponding steps. For example, if the system determines that the user's characteristic interruption segments with the smartwatch were not marked during the construction process, the system considers the current wearing behavior feature sequence to have a potential risk of discontinuity. The system will then perform a backtracking analysis of the feature sequence to identify and supplement the marked features. The system identifies and isolates interrupted segments from the progressive change analysis, structurally corrects the wearing behavior feature sequence, and delays or adjusts the update of the wearing behavior evolution model before the correction is complete. This prevents unmarked interrupted segments from interfering with the evolution modeling, ensuring the accuracy and continuity of the wearing behavior modeling. For example, if the system determines that the user has marked the interrupted segments of the smartwatch's features during the construction process, it considers the current wearing behavior feature sequence to have no potential risk. The system then introduces a behavioral inertia parameter representing the resistance to changes in wearing behavior into the wearing behavior feature sequence. Based on these behavioral inertia parameters, the system identifies the degree of tilt in maintaining the user's wearing behavior over a continuous period. According to different degrees of tilt, it dynamically collects the direction of change of the wearing behavior feature units between adjacent units, generating the evolution direction of the user's wearing behavior in the feature space. This evolution direction specifically includes evolution towards a stable state and evolution towards an unstable state. By dynamically collecting the direction of change of the wearing behavior feature units between adjacent units, the system can identify the evolution direction of the wearing behavior in the feature space, thereby distinguishing whether the wearing state is trending towards stability or gradually moving towards instability. This achieves a directional description of the evolution trend of the wearing behavior, rather than making static judgments based solely on amplitude changes. Furthermore, based on the evolution direction of the wearing behavior in the feature space, the system can make a comprehensive judgment in subsequent physiological monitoring anomaly analysis, combining whether the wearing behavior is evolving towards a stable state or an unstable state. This avoids misjudging small fluctuations in the stable evolution stage as anomalies and increases attention to abnormal signals in the unstable evolution stage. This method effectively improves the robustness and interpretability of anomaly source identification, enabling smartwatches to more accurately reflect real physiological state changes in long-term use scenarios, improving overall monitoring effectiveness and user experience.

[0025] It should be noted that a behavioral inertia parameter representing the resistance to changes in wearing behavior is introduced into the wearing behavior feature sequence. Based on the behavioral inertia parameter, the degree of tilt in which the user's wearing behavior maintains its original state over a continuous period is identified. Based on the degree of tilt, the direction of change of the wearing behavior feature unit between adjacent units is dynamically collected to generate the evolution direction of the user's wearing behavior in the feature space, specifically: Assuming the wear behavior feature sequence has been marked with interrupted segments and its continuity is confirmed, the system further introduces a behavioral inertia parameter to the wear behavior feature sequence, representing the resistance to changes in wear behavior. This behavioral inertia parameter is not a single feature value, but is determined by the amplitude, consistency, and persistence of changes in adjacent wear behavior feature units within a multi-dimensional feature space. It quantifies the user's tendency to maintain the original wear state when transitioning from one time unit to the next. Based on this behavioral inertia parameter, the system determines the degree of tilt in maintaining the original state of the user's wear behavior over a continuous period. The degree of tilt... This describes whether the wearing behavior tends to maintain the existing state or is more prone to state shift. Furthermore, based on the tilt degree, the system performs directional analysis on the change relationship between adjacent wearing behavior feature units, dynamically collecting the corresponding change direction identifiers. This ensures that each feature unit transition includes not only information on "whether it changes" but also directional information on "to what state it changes to." By continuously accumulating the change direction identifiers in the feature space, the system can generate the evolution direction of the user's wearing behavior, which is used to characterize the overall evolution trend of the wearing behavior in the time dimension. This elevates the description of wearing behavior from a static feature set to a dynamic process with directional constraints and evolutionary semantics. Specific examples are as follows: For example, the system collects the user's signal amplitude stability characteristics, periodic signal consistency characteristics, and sensor signal correlation characteristics over multiple consecutive time periods, and constructs a corresponding wearing behavior feature sequence. Assuming that each feature exhibits only minor amplitude changes over three consecutive time periods, and the direction of change is consistent, the system will obtain a high behavioral inertia parameter when calculating the relationship between adjacent feature units. This indicates that the user's wearing behavior has a strong ability to maintain its state during this phase. At this point, the system determines that the wearing behavior has a high degree of inclination to maintain its original state over consecutive time periods and assigns a "moving towards a stable state" direction indicator to adjacent feature units. Subsequently, if the system... If, within a given time period, the amplitude stability characteristic shows an increased decrease, the periodic consistency fluctuations are enhanced, and the direction of change becomes more dispersed, then the corresponding behavioral inertia parameter gradually decreases. Based on this, the system determines that the wearing behavior's tendency to maintain the original state is weakening, and dynamically collects the change direction indicators of "evolution towards an unstable state." By continuously recording the above change direction indicators in the feature space, the system can form the evolution trajectory of the user's wearing behavior from stable to unstable, or identify the evolution direction of its return from unstable to stable after wearing adjustments. This provides a more refined and interpretable directional basis for subsequent updates to the wearing behavior evolution model and the identification of abnormal physiological signal sources.

[0026] In this embodiment, step S3, which involves obtaining information on the gradual changes in user wearing behavior during use, further includes: S31: Based on the change relationship between adjacent wearing behavior units, identify candidate change segments when the user's wearing behavior changes, wherein the candidate change segments are specifically time intervals with continuous observation value; S32: Determine whether the candidate change segment has a cumulative effect, wherein the cumulative effect is specifically that it is continued or amplified in subsequent time periods; S33: If not, the candidate change segment is marked as a local perturbation. During the change process, the coupling relationship between the change of wearing behavior characteristics and the stability parameter of wearing behavior is obtained. Based on the coupling relationship, the stage boundary of the change of wearing behavior is constructed. The stage boundary specifically includes the pre-change stage, the change transition stage, and the post-change stage.

[0027] In this embodiment, the system identifies candidate change segments in user wearing behavior based on the changing relationships between adjacent wearing behavior units. These candidate change segments are specifically time intervals with sustained observation value. The system then determines whether these candidate change segments have a cumulative effect, specifically whether they are continued or amplified in subsequent time periods, and executes corresponding steps accordingly. For example, when the system determines that a candidate change segment in user wearing behavior has a cumulative effect, it considers the candidate change segment to be continued or amplified in subsequent time periods, belonging to a valid change segment with sustained evolution value. The system incorporates this change segment into the wearing behavior evolution modeling process, assigns it a higher weight, and expands the corresponding time observation window to continuously track the change trend. Simultaneously, it updates the behavioral inertia parameters and the evolution direction of wearing behavior in the feature space based on this change segment, enabling the evolution model to reflect the impact of this change segment on the long-term evolution of user wearing behavior, thereby improving the accuracy of wearing behavior analysis and anomaly source identification. Conversely, when the system determines that a candidate change segment in user wearing behavior does not have a cumulative effect, it considers the candidate change segment to be an invalid change segment without sustained evolution value. The system marks candidate change segments as local perturbations and acquires the coupling relationship between changes in wearing behavior characteristics and wearing behavior stability parameters during the change process. Based on different coupling relationships, it constructs stage boundaries for changes in wearing behavior, specifically including the pre-change stage, the transition stage, and the post-change stage. Based on different coupling relationships, the system can identify the inherent stage structure in the process of changes in wearing behavior and construct stage boundaries such as the pre-change stage, the transition stage, and the post-change stage. This transforms what was originally considered an invalid local perturbation into an interpretable and localizable structured change process, thereby improving the precision of wearing behavior analysis. At the same time, by constructing clear stage boundaries, the system can differentiate interpretations based on the specific stage of an anomaly in subsequent wearing behavior evolution analysis or abnormal physiological signal determination. For example, it can distinguish between abnormal signals occurring in the transition stage and those occurring in the pre-change or post-change stages, avoiding misjudging short-term anomalies during local perturbations as long-term wearing behavior evolution or real physiological anomalies. This approach enhances the interpretability and robustness of the system in the process of changes in wearing behavior and improves the overall monitoring effect of smartwatches in complex usage scenarios.

[0028] It should be noted that the candidate change segments are marked as local perturbations. During the change process, the coupling relationship between the changes in wearing behavior characteristics and the wearing behavior stability parameters is obtained. Based on the coupling relationship, the stage boundary of the wearing behavior change is constructed, specifically as follows: When the system determines that a candidate change segment does not have a cumulative effect, it marks it as a local perturbation to indicate that the change does not have the ability to drive the long-term evolution of wearing behavior. Based on this, the system does not directly ignore the local perturbation, but continuously acquires the coupling relationship between changes in wearing behavior characteristics and wearing behavior stability parameters during the change process. This coupling relationship describes the mutual influence between the magnitude and rate of change of wearing behavior characteristics and the overall stability of wearing behavior during the occurrence of a local perturbation. By analyzing the change patterns of the coupling relationship, the system identifies different states of wearing behavior before, during, and after a local perturbation, thereby constructing the stage boundaries of wearing behavior changes. The constructed stage boundaries divide the change process into a pre-change stage, a transition stage, and a post-change stage, structuring the local perturbation into a process with clear start and end points and internal stages, rather than treating it as simple noise, providing clear temporal boundaries and semantic annotations for subsequent wearing behavior analysis. Specific examples are as follows: For example, if the system detects a short-term drop in the signal amplitude stability feature within a certain time period, and identifies this time period as a candidate change segment, and determines that the change does not continue or amplify in subsequent time periods, the system marks it as a local disturbance. During this disturbance, the system simultaneously monitors the wearing behavior stability parameter, finding that while the amplitude stability feature decreases, the stability parameter briefly decreases and then quickly recovers to its original level. Based on this coupling relationship between feature change and stability parameter, the system divides the stable state before the disturbance into the pre-change stage, the time interval between the synchronous decrease in feature change and stability into the change transition stage, and the time interval between feature recovery and stability recovery into the post-change stage. By constructing these stage boundaries, the system can clearly define the scope of influence of the local disturbance and its ending position, thereby avoiding mistaking short-term wearing adjustments for continuous evolution behavior and providing a basis for the staged interpretation of subsequent abnormal signals.

[0029] In this embodiment, step S5, which involves collecting abnormal physiological monitoring signals from the user and identifying the abnormal source of the abnormal physiological monitoring signals using the evolutionary model, further includes: S51: Based on the abnormality type characterization of the abnormal physiological monitoring signal pre-processed by the smartwatch, the time range of the abnormal physiological monitoring signal is collected, wherein the time range specifically includes the start time, duration and end time; S52: Determine whether the time interval can be aligned with the time structure information in the evolution model; S53: If so, identify the evolutionary state information of the smartwatch, and perform structural matching between the abnormal physiological monitoring signal and the wearing behavior evolution process based on the evolutionary state information. Obtain the degree of consistency between the abnormal physiological monitoring signal and the wearing behavior evolution process in the change dimension, and generate the corresponding structural correlation. The evolutionary state information specifically includes the current wearing behavior stage, the wearing stability level corresponding to the stage, the evolution direction of the stage, and the change rate of the stage. The change dimension specifically includes the change rhythm, change direction, and change stage.

[0030] In this embodiment, the system, based on the pre-defined anomaly type characterization of abnormal physiological monitoring signals by the smartwatch, collects the time intervals of these abnormal physiological monitoring signals. The time intervals specifically include the start time, duration, and end time. The system then determines whether these time intervals can align with the temporal structure information in the evolutionary model to execute corresponding steps. For example, when the system determines that the time interval of the abnormal physiological monitoring signal cannot align with the temporal structure information in the evolutionary model, the system considers the abnormal signal to lack a temporal semantic correspondence with the current wearing behavior evolution process. The system then marks the abnormal physiological monitoring signal as a time-incompatible anomaly and removes it from the system. In the evolutionary correlation analysis of wearing behavior, the abnormality is separated and instead judged independently based on pre-defined abnormality types. This avoids inaccurate attribution of abnormality sources and further triggers a temporal structure verification mechanism to improve the adaptability of subsequent evolutionary models to the temporal characteristics of abnormal signals. For example, when the system determines that the temporal range of an abnormal physiological monitoring signal can align with the temporal structure information in the evolutionary model, the system considers the abnormal signal to correspond to the current wearing behavior evolution process in temporal semantics. The system will then identify the smartwatch's evolutionary state information, which specifically includes the current wearing behavior stage, the corresponding wearing stability level, the evolutionary direction of the stage, and the... The system analyzes the rate of change in each stage of the wearing behavior. Based on this evolutionary state information, it performs structural matching between abnormal physiological monitoring signals and the wearing behavior evolution process to obtain the degree of consistency between the abnormal physiological monitoring signals and the wearing behavior evolution process in terms of change dimensions, specifically including change rhythm, change direction, and change stage, generating corresponding structural correlations. The system identifies the current wearing behavior stage, corresponding wearing stability level, evolutionary direction, and change rate, embedding abnormal physiological monitoring signals into the wearing behavior evolution framework for structural matching. Simultaneously, it evaluates the consistency between abnormal signals and the wearing behavior evolution process across multiple change dimensions, including change rhythm, change direction, and change stage. The system can characterize the relationship between anomalies and wearing behavior at the structural level, rather than just at the level of numerical correlation or temporal overlap. This improves the precision and expressive power of anomaly analysis. Based on the generated structural correlation, the system can more accurately distinguish whether abnormal physiological monitoring signals are highly correlated with the evolution of wearing behavior or, although aligned in time, have a low correlation in the dimension of structural change. This provides a quantitative basis for determining the source of anomalies. This approach effectively reduces the risk of misjudgment caused by changes in wearing behavior, enhances the interpretability of anomaly analysis results, and enables smartwatches to output more reliable anomaly judgment results with evolutionary semantic support in complex usage scenarios and long-term monitoring.

[0031] It should be noted that, in identifying the evolutionary state information of the smartwatch, and based on this evolutionary state information, structural matching is performed between the abnormal physiological monitoring signals and the wearing behavior evolution process to obtain the degree of consistency between the abnormal physiological monitoring signals and the wearing behavior evolution process in terms of change dimensions, thereby generating corresponding structural correlations. Specifically: The system first identifies the evolutionary state information of the smartwatch, including the current stage of the wearing behavior, the corresponding level of wearing stability, the direction of evolution, and the rate of change. This evolutionary state information not only describes the static characteristics of the user's wearing behavior but also depicts the dynamic evolution trend of the wearing behavior over a continuous time period. Subsequently, based on this evolutionary state information, the system embeds abnormal physiological monitoring signals into the wearing behavior evolution framework for structural matching. The core of structural matching lies in analyzing the consistency between abnormal signals and the evolutionary model across multiple dimensions of change, including the rhythm, direction, and stage of change, thereby quantifying the synchronicity and correlation between abnormal signals and the wearing behavior evolution process. Through this multi-dimensional matching, the system can establish a structured correlation between abnormal signals and the wearing behavior evolution process, rather than relying solely on simple temporal overlap or the similarity of a single feature. This generates a structural correlation index that reflects the degree of correlation between abnormalities and the wearing behavior evolution, providing a quantifiable basis for determining the source of abnormalities. Specific examples are as follows: For example, the system collects amplitude stability characteristics, periodic signal consistency characteristics, and sensor signal correlation characteristics of the user's watch over a continuous time period and constructs a wearing behavior evolution model. During a certain period, the user experiences an abnormal heart rate signal, the time interval of which aligns with a certain wearing behavior stage in the evolution model. The system identifies the evolutionary state information of this stage: this stage is a transitional phase of wearing behavior, with a moderate level of stability, a gradual trend towards stability, and a slowly decreasing rate of change. Subsequently, the system maps the abnormal heart rate signal to this evolutionary stage and performs comparative analysis on the dimensions of change rhythm, change direction, and the evolution rate corresponding to the stage. It finds that the change rhythm of the abnormal signal highly matches the rhythm of the evolutionary stage, the change direction is consistent with the evolution direction, and the time of the abnormal signal occurrence is precisely in the middle of the stage transition. Based on these matching results, the system generates a high-structure correlation index, indicating that the abnormal heart rate signal may be closely related to the wearing behavior evolution process, thus providing a structured and interpretable reference for subsequent anomaly source determination.

[0032] It should be added that the anomaly type characterization includes at least one or more of the following: whether the anomaly is a sudden anomaly or a persistent anomaly, whether the anomaly exhibits periodic distortion or overall amplitude drift, and whether the anomaly is accompanied by a decrease in signal integrity.

[0033] In this embodiment, step S2, which determines whether the implicit feature parameters can reflect the wearing behavior state of the smartwatch, further includes: S21: Based on the different wearing behavior states of the smartwatch, obtain the difference data of the implicit feature parameters within a preset time period; S22: Determine whether the difference data detects a preset convergence behavior; S23: If so, identify the characterization degradation information of the implicit feature parameter, collect the assimilation influence parameter that shows the same direction but non-responsive change in the changing trend of the implicit feature parameter and the wearing adjustment behavior based on the characterization degradation information, and construct the slow unidirectional drift event of the smartwatch based on the assimilation influence parameter.

[0034] In this embodiment, the system acquires differential data of implicit feature parameters within a pre-set time period based on different user wearing behaviors of the smartwatch. The system then determines whether these differential data detects a pre-defined convergence pattern and executes corresponding steps accordingly. For example, if the system determines that the differential data of implicit feature parameters of the smartwatch within a pre-set time period does not detect a pre-defined convergence pattern, the system considers that the user's wearing behavior has not shown a convergent or stable trend within the observation period, indicating a risk of wearing behavior fluctuation. The system marks the observation segment as a fluctuation segment, extends the observation window, or adjusts the sampling granularity to capture possible long-term convergence trends. Simultaneously, within this fluctuation segment, the system re-evaluates the behavioral inertia parameters and evolution direction, and weights this fluctuation segment in the wearing behavior evolution model and abnormal signal analysis, thereby ensuring the accuracy of refined analysis and long-term evolution modeling of user wearing behavior trends. For example, if the system determines that the differential data of implicit feature parameters of the smartwatch within a pre-set time period detects a pre-defined convergence pattern, the system considers that the user's wearing behavior has not shown a convergent or stable trend within the observation period, indicating a risk of wearing behavior fluctuation. When user wearing behavior shows a convergence or stabilization trend during the observation period, the system identifies the representational degradation information of these implicit feature parameters. Based on the representational degradation information, it collects the changing trends of these implicit feature parameters and the assimilation influence parameters when wearing adjustment behavior shows changes in the same direction but not responsive. Based on different assimilation influence parameters, it constructs a slow unidirectional drift event of the smartwatch. By introducing assimilation influence parameters, the system can capture the potential contribution of user wearing behavior to feature stability and evolution direction in the analysis, achieving a refined characterization of gradual changes. At the same time, based on different assimilation influence parameters, the system integrates the user's long-term small unidirectional changes to construct a slow unidirectional drift event of the smartwatch. This event can characterize the gradual shift trend of user wearing behavior in a stable state, thereby providing long-term behavioral evolution information for the evolutionary model. This enables the system to distinguish between slow drift caused by wearing behavior and abnormal signals caused by changes in real physiological state, improving the robustness and interpretability of long-term anomaly identification, and providing a quantitative basis for the periodic update of the evolutionary model and the prediction of wearing behavior trends.

[0035] It should be noted that the identification of the latent feature parameters' representation degradation information, based on the representation degradation information, collects the assimilation influence parameters of the latent feature parameters' change trend and the wearing adjustment behavior showing a unidirectional but non-responsive change. Based on the assimilation influence parameters, a slow unidirectional drift event of the smartwatch is constructed, specifically: The system first continuously observes the latent feature parameters, identifying degenerative information that still appears in stable or converging states, such as a slow decrease in signal amplitude stability, a gradual weakening of periodic consistency, or a continuous decrease in the correlation of multi-sensor signals. Then, the system jointly analyzes the changing trends of these feature parameters with the user's wearing adjustment behavior during use. When the two show a consistent direction of change, but the wearing adjustment does not immediately trigger a significant response in the feature parameters, the system classifies it as a "co-directional but non-responsive change." Based on this, the system extracts assimilation influence parameters to quantify the potential cumulative effect of wearing adjustment behavior on the long-term evolution of latent feature parameters. Finally, based on these assimilation influence parameters, the system integrates the originally scattered and weak co-directional changes to construct a unidirectional drift event that reflects the long-term, slow evolution of wearing behavior, thus providing continuous and interpretable long-term change input for the evolutionary model. Specific examples are as follows: For example, a user gradually develops a habit of slightly shifting the wearing position of their smartwatch outwards during daily use, but each adjustment is small and does not trigger significant signal quality anomalies in the short term. The system observes that while the heart rate signal's periodicity remains consistent, the signal amplitude stability characteristic shows a continuous, unidirectional, slight decrease over several days, while the sensor signal's related characteristics also slowly weaken. The system further identifies that these trends are consistent with the user's wearing adjustment behavior in direction, but there is no situation where a single adjustment immediately triggers a sudden change in characteristics. Therefore, the corresponding assimilation effect parameter is extracted. Based on this parameter, the system constructs this series of small but continuous changes into a slow, unidirectional drift event to characterize the gradual shift in the user's wearing behavior under stable usage conditions. This allows for effective differentiation between long-term drift caused by wearing behavior and actual physiological state changes in subsequent abnormal physiological signal analysis.

[0036] In this embodiment, step S4, which determines whether the gradual change information is within the normal fluctuation range corresponding to the preset reference baseline, further includes: S41: Based on the change path of the progressive change information, identify the change type of the progressive change information, wherein the change type specifically includes continuous approximation type, step transition type and nonlinear bending type; S42: Determine whether the change type exhibits a preset boundary fitting behavior; S43: If so, then a preset alternative reference baseline is introduced for parallel verification. Based on the verification results of the parallel verification, the baseline instability interval of the progressive change information is dynamically generated. The alternative reference baseline specifically includes the reference baseline formed in adjacent wearing stages, the statistical baseline of similar user groups, and the short-term adaptive update baseline.

[0037] In this embodiment, the system identifies the change types of progressive change information based on the change path of the progressive change information. These change types specifically include continuous approximation, step transition, and nonlinear bending. The system then determines whether these change types exhibit pre-defined boundary-fitting behavior and executes corresponding steps accordingly. For example, if the system determines that the change type of the progressive change information does not exhibit pre-defined boundary-fitting behavior, the system considers that although the currently identified progressive change has certain continuity or stage characteristics in its change path, its evolution trajectory does not gradually approach the preset normal boundary, stable boundary, or risk boundary, nor does it form a restricted or fitting trend near the boundary. The system will mark the corresponding progressive change information as "non-boundary-dominated change" and adopt an extended observation strategy, encrypting the time segment corresponding to this change type or extending the observation window to capture whether it subsequently shows a trend of fitting towards a certain boundary. Simultaneously, temporarily... The weight of this change in the anomaly source determination is reduced to avoid misjudgment due to an undefined change path, and relevant change paths are included in candidate evolution branches to provide a basis for subsequent evolution model updates or boundary reassessment. For example, when the system determines that the change type of progressive change information exhibits a pre-defined boundary-fitting behavior, the system will consider that the currently identified progressive change evolution trajectory is gradually approaching the preset normal boundary, stable boundary, or risk boundary, forming a restricted or fitting trend near the boundary. The system will introduce a pre-defined alternative reference baseline for parallel verification. The alternative reference baseline specifically includes the reference baseline formed by adjacent wearing stages, the statistical baseline of similar user groups, and the short-term adaptive update baseline. Based on the verification results of the parallel verification, the baseline instability interval of progressive change information is dynamically generated. By identifying whether the progressive change type exhibits boundary-fitting behavior, the system can determine whether the change trajectory has entered an evolutionary state constrained by the boundary. Building upon this foundation, alternative reference baselines from multiple sources are introduced for parallel verification to avoid distortion caused by aging, drift, or individual differences in a single reference baseline. Furthermore, by introducing parallel verification of multiple baselines and generating baseline instability intervals during the boundary fitting stage, the system can more precisely distinguish between anomalies caused by long-term changes in wearing behavior and those caused by changes in actual physiological state. When abnormal physiological monitoring signals fall into the baseline instability interval, the system can reduce the certainty of judging a single anomaly and instead make a comprehensive judgment based on the evolution direction and rate of change, thereby reducing false alarms and missed alarms and providing a more robust and reliable decision-making basis for subsequent risk warnings or wearing prompts.

[0038] It should be noted that a pre-defined alternative reference baseline is introduced for parallel verification. Based on the verification results of the parallel verification, the baseline instability interval of the progressively changing information is dynamically generated, specifically as follows: When the system identifies that the gradual change information exhibits boundary-fitting behavior, to avoid the failure or misjudgment of a single reference baseline near the boundary, the system introduces alternative reference baselines from a preset source for parallel verification. This parallel verification process is not a simple comparison, but rather projects the gradual change information onto reference baselines formed in adjacent wearing stages, statistical baselines of similar user groups, and short-term adaptive update baselines, respectively, and calculates their deviation degree, fit strength, and change consistency. The system integrates the verification results of multiple baselines, identifies the consistent and divergent intervals in the judgment results between different baselines, and dynamically marks the time range in which the applicability of the baselines shows obvious divergence or fluctuation as the baseline unstable interval. This unstable interval is used to characterize the boundary-sensitive state of the current gradual change information, providing a buffer space for subsequent evolution model adjustments and anomaly judgments. Specific examples are as follows: For example, during a certain usage phase, the amplitude stability of the user's wearing behavior gradually approaches the preset stability boundary. The system uses a reference baseline formed in the most recent wearing phase, a statistical baseline of a user group of the same age and similar wearing habits, and a short-term baseline that has been adaptively updated in the last 24 hours for parallel verification. The results show that the baseline of adjacent wearing phases determines that the change is still within the normal range, while the group statistical baseline shows that it has approached the risk threshold, and the short-term adaptive baseline shows fluctuations in the determination result. Based on this, the system identifies the time interval of the divergence in the determination of multiple baselines and dynamically generates this interval as the baseline instability interval. Within this interval, the system will not immediately trigger an abnormal alarm, but will continuously observe the subsequent change trend and evolution direction, thereby avoiding misjudgment caused by baseline switching or boundary aging, and providing a basis for the smooth update of the wearing behavior evolution model.

[0039] In this embodiment, step S1, which identifies the user's implicit feature parameters based on the smartwatch's pre-detected wearing area, further includes: S11: Collect the pre-analyzed regional composition state of the wearing area by the smartwatch, wherein the regional composition state specifically includes the contact tightness state of the watch strap, the tilt state of the watch body relative to the skin, and the rotation state of the wearing area relative to the limb axis. S12: Determine whether the state of the region is lower than a preset stability level; S13: If so, mark the wearing area corresponding to the region composition state as a structurally uncertain region, and dynamically activate the hidden feature candidate pool that is related to the current wearing area structural state based on the region composition state.

[0040] In this embodiment, the system collects data from the smartwatch regarding the pre-analyzed regional composition state of the wearing area. This regional composition state specifically includes the contact tightness of the watchband, the tilt angle of the watch body relative to the skin, and the rotation state of the wearing area relative to the limb axis. The system then determines whether these regional composition states are below a pre-set stability level to execute corresponding steps. For example, if the system determines that the pre-analyzed regional composition state of the smartwatch is not below the pre-set stability level, it considers the contact tightness between the watchband and skin sufficient, the tilt angle of the watch body relative to the skin not significantly off, and the rotation state of the wearing area relative to the limb axis within a reasonable range. The system incorporates the current regional composition state as valid input into the wearing behavior feature sequence and evolution model to continuously track the gradual changes in wearing behavior. Simultaneously, the regional composition parameters corresponding to this stable state are used to update or strengthen the current reference baseline to improve the accuracy of subsequent stability level judgments. Furthermore, for the processing of physiological monitoring signals, the system maintains a conventional collection and analysis strategy, without additionally widening or tightening the anomaly judgment threshold, thereby ensuring monitoring continuity while avoiding excessive intervention that could affect the user's normal wearing experience. For example, when the system determines that the smartwatch... If the pre-analyzed region structure of the watch's wearing area falls below a pre-set stability level, the system considers the smartwatch to be physically unstable, and its positioning on the wrist is no longer fixed. The system marks these regions as structurally uncertain areas. Based on these regional structures, it dynamically activates a pool of latent feature candidates that are correlated with the current structural state of the wearing area. By dynamically activating this pool, the system can focus on feature types sensitive to changes in wearing conditions within structurally uncertain areas, such as signal-related features sensitive to tilt changes or amplitude stability features sensitive to contact changes. Marking the wearing area as a structurally uncertain area, combined with the dynamic activation of the latent feature candidate pool, helps assign different explanatory weights to data generated within this area in subsequent wearing behavior evolution modeling and abnormal physiological signal analysis. Furthermore, the system can distinguish between feature fluctuations caused by physical wearing instability and actual physiological state changes, reducing the misidentification of wearing problems as physiological abnormalities. This improves the stability of the evolutionary model and the reliability of anomaly source determination, providing more robust technical support for long-term smartwatch monitoring and user experience.

[0041] It should be added that the latent feature candidate pool is used to limit the range of latent feature parameters that are physically interpretable under the current wearing area conditions.

[0042] Reference Appendix Figure 2An evolutionary modeling system for a smartwatch based on wearing behavior, as described in one embodiment of the present invention, includes: The identification module 10 is used to identify the user's implicit feature parameters based on the wearing area pre-detected by the smartwatch. The implicit feature parameters specifically include signal amplitude stability features, periodic signal consistency features, and sensor signal correlation features. The judgment module 20 is used to determine whether the implicit feature parameters can reflect the wearing behavior state of the smartwatch; The execution module 30 is used to construct the wearing behavior feature sequence of the smartwatch according to a preset time sequence if possible, and to analyze the changing trend of the wearing behavior feature sequence using a preset distribution drift detection method to obtain the gradual change information of the user's wearing behavior during use. The gradual change information specifically includes the slow shift of the feature mean, the continuous change of feature fluctuation, and the reduction and enhancement of feature stability. The second judgment module 40 is used to determine whether the gradual change information is within the normal fluctuation range corresponding to the preset reference baseline; The second execution module 50 is configured to, if so, construct an evolutionary model to characterize the user's wearing behavior based on the progressive change information, periodically trigger updates to the evolutionary model during the use process, collect abnormal physiological monitoring signals of the user, and identify the abnormal source of the abnormal physiological monitoring signals through the evolutionary model, wherein the abnormal source specifically includes abnormalities caused by long-term changes in wearing behavior and abnormalities caused by changes in actual physiological state.

[0043] In this embodiment, the identification module 10 identifies the user's implicit feature parameters based on the pre-detected wearing area of ​​the smartwatch. These implicit feature parameters specifically include signal amplitude stability characteristics, periodic signal consistency characteristics, and sensor signal correlation characteristics. Then, the judgment module 20 determines whether these implicit feature parameters can reflect the smartwatch's wearing behavior state, and executes corresponding steps accordingly. For example, if the system determines that the user's implicit feature parameters cannot reflect the smartwatch's wearing behavior state, the system considers the current implicit feature parameters insufficient as a basis for characterizing the user's wearing behavior state. For example, wearing the watch too loosely, rotating it, or partially leaving it suspended could lead to insufficient sensor signal... Primarily influenced by random movement or changes in contact, rather than actual wearing behavior, the system will pause wearing behavior modeling based on the implicit feature parameters, marking the current stage as an uncertain wearing behavior stage. Simultaneously, it will trigger a re-evaluation process of the wearing area and the effectiveness of the features, and adopt a conservative interpretation strategy for abnormal physiological monitoring signals to avoid misjudging signal fluctuations caused by wearing instability as real physiological abnormalities, thereby improving the monitoring reliability of the smartwatch during long-term use. For example, when the system determines that the user's implicit feature parameters can reflect the wearing behavior state of the smartwatch, the execution module 30 will consider the current implicit feature parameters sufficient to characterize the user's wearing behavior. The system uses a pre-set time sequence to construct a sequence of smartwatch wearing behavior characteristics. It then employs a pre-defined distribution drift detection method to analyze this sequence, obtaining information on the gradual changes in the user's wearing behavior during watch use. These gradual changes include slow shifts in the feature mean, continuous changes in feature fluctuations, and a decrease or enhancement in feature stability. The second judgment module 40 then determines whether these gradual changes are within the normal fluctuation range corresponding to a pre-set reference baseline, and executes the corresponding steps accordingly. For example, if the system determines that the gradual changes in the user's wearing behavior during use are not within the normal fluctuation range, the system will proceed accordingly. Within the normal fluctuation range corresponding to the pre-set reference baseline, the system will consider that the user's wearing behavior has undergone continuous structural changes, and the original reference baseline can no longer fully represent the wearing characteristics of the current use stage. The system will mark the current wearing behavior as an evolutionary state beyond the baseline, suspend the normality judgment based on the original reference baseline, and trigger the update or expansion process of the wearing behavior evolution model. After the model stabilizes, a new reference baseline will be constructed. In the transition phase, a conservative interpretation strategy will be adopted for abnormal physiological monitoring signals to avoid misjudging the long-term evolution of wearing behavior as an abnormal situation, thereby improving the monitoring accuracy and system adaptability of the smartwatch during long-term use.For example, when the system determines that the gradual changes in the user's wearing behavior during use are within the normal fluctuation range corresponding to a pre-set reference baseline, the second execution module 50 will consider that the user's wearing behavior has not undergone structural changes, and the original reference baseline can still characterize the wearing characteristics of the current usage stage. Based on these gradual changes, the system will construct an evolutionary model to characterize the user's wearing behavior. During watch use, the evolutionary model will be periodically updated, and abnormal physiological monitoring signals of the user will be collected. Through the evolutionary model, the abnormal sources of these physiological monitoring signals will be identified. These abnormal sources specifically include abnormalities caused by long-term changes in wearing behavior and abnormalities caused by changes in actual physiological state.

[0044] In this embodiment, it also includes: The second identification module is used to identify the user's characteristic interruption segments on the smartwatch based on the construction process of the wearing behavior feature sequence, wherein the characteristic interruption segments specifically include taking off the watch, not wearing it for a long time, and insufficient monitoring conditions; The third judgment module is used to determine whether the feature interruption segment is marked during the construction process; The third execution module is used to, if so, introduce a behavioral inertia parameter representing the resistance to changes in wearing behavior into the wearing behavior feature sequence, identify the degree of tilt of the user's wearing behavior in maintaining the original state within a continuous period of time based on the behavioral inertia parameter, dynamically collect the change direction identifier of the wearing behavior feature unit between adjacent units based on the degree of tilt, and generate the evolution direction of the user's wearing behavior in the feature space, wherein the evolution direction specifically includes evolution towards a stable state and evolution towards an unstable state.

[0045] In this embodiment, the system identifies characteristic interruptions in the user's interaction with the smartwatch based on the construction process of the wearing behavior feature sequence. These interruptions specifically include removing the watch, prolonged periods without wearing it, and insufficient monitoring conditions. The system then determines whether these interruptions were marked during the construction process and executes corresponding steps accordingly. For example, if the system determines that the user did not mark the characteristic interruptions during the construction process, it considers the current wearing behavior feature sequence to have a potential discontinuity risk. The system performs a backtracking analysis of the feature sequence, identifies and supplements the marking of the corresponding interruptions, and isolates these interruptions from the progressive change analysis. This structural correction of the wearing behavior feature sequence, and before the correction is complete, the system delays or corrects the wearing behavior. The wearing behavior evolution model is updated to avoid interference from unmarked interruptions, thus ensuring the accuracy and continuity of wearing behavior modeling. For example, when the system determines that the user has marked the feature interruption segment of the smartwatch during the construction process, the system will consider that there is no potential risk in the current wearing behavior feature sequence. The system will introduce behavioral inertia parameters that characterize the resistance to changes in wearing behavior into the wearing behavior feature sequence. Based on these behavioral inertia parameters, the system will identify the degree of tilt of the user's wearing behavior in maintaining the original state over a continuous period of time. Based on different degrees of tilt, the system will dynamically collect the change direction identifiers of the wearing behavior feature units between adjacent units to generate the evolution direction of the user's wearing behavior in the feature space. The evolution direction specifically includes evolution towards a stable state and evolution towards an unstable state.

[0046] In this embodiment, the execution module further includes: The identification submodule is used to identify candidate change segments when the user's wearing behavior changes based on the change relationship between adjacent wearing behavior units, wherein the candidate change segments are specifically time intervals with continuous observation value; The judgment submodule is used to determine whether the candidate change segment has a cumulative effect, wherein the cumulative effect is specifically that it is continued or amplified in subsequent time periods; The execution submodule is used to mark the candidate change segment as a local perturbation if no, obtain the coupling relationship between the change of wearing behavior characteristics and the wearing behavior stability parameter during the change process, and construct the stage boundary of the wearing behavior change based on the coupling relationship. The stage boundary specifically includes the pre-change stage, the change transition stage and the post-change stage.

[0047] In this embodiment, the system identifies candidate change segments in user wearing behavior based on the changing relationships between adjacent wearing behavior units. These candidate change segments are specifically time intervals with sustained observation value. The system then determines whether these candidate change segments have a cumulative effect, specifically whether they are continued or amplified in subsequent time periods, and executes corresponding steps accordingly. For example, when the system determines that a candidate change segment in user wearing behavior has a cumulative effect, it considers that the candidate change segment is continued or amplified in subsequent time periods, belonging to a valid change segment with sustained evolutionary value. The system incorporates this change segment into the wearing behavior evolution modeling process, assigns it a higher weight, and expands the corresponding time observation window to continuously track the changing trend. Simultaneously, based on the change segment, the inertia parameters of the behavior and the evolution direction of the wearing behavior in the feature space are updated, so that the evolution model can reflect the impact of the change segment on the long-term evolution of the user's wearing behavior, thereby improving the accuracy of wearing behavior analysis and anomaly source identification. For example, when the system determines that the candidate change segment when the user's wearing behavior changes does not have a cumulative effect, the system will consider the candidate change segment to be an invalid change segment without continuous evolution value. The system will mark the candidate change segment as a local perturbation, obtain the coupling relationship between the change of wearing behavior characteristics and the wearing behavior stability parameters during the change process, and construct the stage boundary of the wearing behavior change according to different coupling relationships. The stage boundary specifically includes the pre-change stage, the change transition stage, and the post-change stage.

[0048] In this embodiment, the second execution module further includes: The acquisition submodule is used to acquire the time range of the abnormal physiological monitoring signal based on the abnormal type characterization pre-performed by the smartwatch on the abnormal physiological monitoring signal, wherein the time range specifically includes the start time, duration and end time; The second judgment submodule is used to determine whether the time interval can be aligned with the time structure information in the evolution model; The second execution submodule is used to identify the evolutionary state information of the smartwatch if the condition is met, and to perform structural matching between the abnormal physiological monitoring signal and the wearing behavior evolution process based on the evolutionary state information, thereby obtaining the degree of consistency between the abnormal physiological monitoring signal and the wearing behavior evolution process in terms of change dimensions and generating corresponding structural correlations. Specifically, the evolutionary state information includes the current wearing behavior stage, the wearing stability level corresponding to the stage, the evolution direction of the stage, and the change rate of the stage. The change dimensions specifically include the change rhythm, change direction, and change stage.

[0049] In this embodiment, the system, based on the pre-defined anomaly type characterization of abnormal physiological monitoring signals by the smartwatch, collects the time intervals of these abnormal physiological monitoring signals. These time intervals specifically include the start time, duration, and end time. The system then determines whether these time intervals can align with the temporal structure information in the evolutionary model to execute corresponding steps. For example, when the system determines that the time interval of the abnormal physiological monitoring signal cannot align with the temporal structure information in the evolutionary model, the system considers the abnormal signal to lack a temporal semantic correspondence with the current wearing behavior evolution process. The system marks the abnormal physiological monitoring signal as a time-incompatible anomaly, removes it from the wearing behavior evolution correlation analysis, and independently interprets it based on the pre-defined anomaly type to avoid inaccurate attribution of the anomaly source and further... A temporal structure verification mechanism is triggered to improve the adaptability of the subsequent evolutionary model to the temporal characteristics of abnormal signals. For example, when the system determines that the temporal range of the abnormal physiological monitoring signal can be aligned with the temporal structure information in the evolutionary model, the system will consider that the abnormal signal corresponds to the current wearing behavior evolution process in terms of temporal semantics. The system will identify the evolutionary state information of the smartwatch, which specifically includes the current wearing behavior stage, the wearing stability level corresponding to the stage, the evolution direction of the stage, and the rate of change of the stage. Based on this evolutionary state information, the abnormal physiological monitoring signal is structurally matched with the wearing behavior evolution process to obtain the degree of consistency between the abnormal physiological monitoring signal and the wearing behavior evolution process in terms of change dimensions. The change dimensions specifically include the change rhythm, change direction, and change stage, generating the corresponding structural correlation.

[0050] In this embodiment, the determination module further includes: The acquisition submodule is used to acquire the difference data of the implicit feature parameters within a preset time period based on the different wearing behavior states of the smartwatch; The third judgment submodule is used to determine whether the difference data detects a preset convergence behavior; The third execution submodule is used to identify the characterization degradation information of the implicit feature parameters if the condition is met, collect the assimilation influence parameters of the changing trend of the implicit feature parameters and the wearing adjustment behavior in the same direction but not in a responsive manner based on the characterization degradation information, and construct the slow unidirectional drift event of the smartwatch based on the assimilation influence parameters.

[0051] In this embodiment, the system acquires differential data of implicit feature parameters within a pre-set time period based on different user wearing behaviors of the smartwatch. The system then determines whether these differential data detects a pre-defined convergence pattern and executes corresponding steps accordingly. For example, if the system determines that the differential data of the implicit feature parameters of the smartwatch within the pre-set time period does not detect a pre-defined convergence pattern, the system considers that the user's wearing behavior has not shown a convergent or stable trend during the observation period, indicating a risk of wearing behavior fluctuation. The system will mark the observation segment as a fluctuation segment, extend the observation window, or adjust the sampling granularity to capture possible long-term convergence trends. Simultaneously, within this fluctuation segment, the system will re-evaluate the behavioral inertia parameters and their evolution. The system determines the direction and weights the fluctuation range in the wearing behavior evolution model and abnormal signal analysis to ensure the accuracy of refined analysis and long-term evolution modeling of user wearing behavior changes. For example, when the system detects that the difference data of the implicit feature parameters of the smartwatch within a preset time period shows a preset convergence, the system will consider that the user wearing behavior shows a convergence or stable trend during the observation period. The system will identify the representation degradation information of these implicit feature parameters, and collect the assimilation influence parameters when the changing trend of these implicit feature parameters and the wearing adjustment behavior show the same direction but non-responsive changes. Based on different assimilation influence parameters, the system constructs the slow unidirectional drift event of the smartwatch.

[0052] In this embodiment, the second determination module further includes: The second identification submodule is used to identify the change type of the gradual change information based on the change path of the gradual change information, wherein the change type specifically includes continuous approximation type, step transition type and nonlinear bending type; The fourth judgment submodule is used to determine whether the change type exhibits a preset boundary fitting behavior; The fourth execution submodule is used to introduce a preset alternative reference baseline for parallel verification if the condition is met. Based on the verification results of the parallel verification, the baseline instability interval of the progressive change information is dynamically generated. Specifically, the alternative reference baseline includes a reference baseline formed in adjacent wearing stages, a statistical baseline of similar user groups, and a short-term adaptive update baseline.

[0053] In this embodiment, the system identifies the change types of progressive change information based on the change path of the progressive change information. These change types specifically include continuous approximation, step transition, and nonlinear bending. The system then determines whether these change types exhibit pre-defined boundary-fitting behavior to execute corresponding steps. For example, if the system determines that the change type of the progressive change information does not exhibit pre-defined boundary-fitting behavior, the system considers that although the currently identified progressive change has certain continuity or stage characteristics in its change path, its evolution trajectory does not gradually approach the preset normal boundary, stable boundary, or risk boundary, nor does it form a restricted or fitting trend near the boundary. The system will mark the corresponding progressive change information as "non-boundary-dominated change" and adopt an extended observation strategy, such as encrypting the sampling of the time segment corresponding to this change type or extending the observation window. The system detects whether the gradual change tends to converge towards a certain boundary, while temporarily reducing the weight of this change in the anomaly source determination to avoid misjudgment due to an undefined change path. It also incorporates relevant change paths into candidate evolutionary branches to provide a basis for subsequent evolutionary model updates or boundary reassessment. For example, when the system determines that the gradual change information exhibits a pre-defined boundary-converging behavior, it considers the currently identified gradual change evolution trajectory to gradually approach a pre-defined normal boundary, stable boundary, or risk boundary, forming a restricted or converging trend near the boundary. The system then introduces a pre-defined alternative reference baseline for parallel verification. This alternative reference baseline specifically includes a reference baseline formed during adjacent wearing stages, a statistical baseline of similar user groups, and a short-term adaptive update baseline. Based on the verification results of the parallel verification, the system dynamically generates the baseline instability interval for the gradual change information.

[0054] In this embodiment, the identification module further includes: The second acquisition submodule is used to acquire the pre-analyzed regional composition state of the wearing area by the smartwatch, wherein the regional composition state specifically includes the contact tightness state of the watch strap, the tilt state of the watch body relative to the skin, and the rotation state of the wearing area relative to the limb axis. The fifth judgment submodule is used to determine whether the state of the region is lower than a preset stability level; The fifth execution submodule is used to mark the wearing area corresponding to the regional composition state as a structurally uncertain region if the condition is met, and to dynamically activate the hidden feature candidate pool that is related to the current structural state of the wearing area based on the regional composition state.

[0055] In this embodiment, the system collects the pre-analyzed regional composition state of the wearing area from the smartwatch. This regional composition state specifically includes the contact tightness of the watchband, the tilt angle of the watch body relative to the skin, and the rotation state of the wearing area relative to the limb axis. The system then determines whether these regional composition states are below a pre-set stability level to execute corresponding steps. For example, if the system determines that the pre-analyzed regional composition state of the wearing area from the smartwatch is not below the pre-set stability level, the system considers the contact tightness between the watchband and skin sufficient, the tilt angle of the watch body relative to the skin not significantly off, and the rotation state of the wearing area relative to the limb axis within a reasonable range. The system then incorporates the current regional composition state as valid input into the wearing behavior feature sequence and evolution model to continuously track the gradual changes in wearing behavior. The system uses the region composition parameters corresponding to the stable state to update or strengthen the current reference baseline, thereby improving the accuracy of subsequent stability level judgments. Furthermore, for the processing of physiological monitoring signals, the system maintains the conventional acquisition and analysis strategy, without additionally relaxing or tightening the anomaly judgment threshold. This ensures monitoring continuity while avoiding excessive intervention that could affect the user's normal wearing experience. For example, when the system determines that the pre-analyzed region composition state of the smartwatch is lower than the pre-set stability level, the system considers the smartwatch to have exhibited unstable physical wearing characteristics, and the watch's positioning on the wrist is no longer fixed. The system marks the wearing areas corresponding to these region composition states as structurally uncertain regions. Based on these region composition states, it dynamically activates a pool of latent feature candidates that are correlated with the current structural state of the wearing area.

[0056] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An evolutionary modeling method for smartwatches based on wearing behavior, characterized in that, Includes the following steps: Based on the wearing area of ​​the user pre-detected by the smartwatch, the implicit feature parameters of the user are identified, wherein the implicit feature parameters specifically include signal amplitude stability features, periodic signal consistency features, and sensor signal correlation features; Determine whether the implicit feature parameters can reflect the wearing behavior status of the smartwatch; If possible, the wearing behavior feature sequence of the smartwatch is constructed according to the preset time sequence. The changing trend of the wearing behavior feature sequence is analyzed by using the preset distribution drift detection to obtain the gradual change information of the user's wearing behavior during use. The gradual change information specifically includes the slow shift of the feature mean, the continuous change of feature fluctuation, and the reduction and enhancement of feature stability. Determine whether the gradual change information is within the normal fluctuation range corresponding to the preset reference baseline; If so, an evolutionary model is constructed based on the progressive change information to characterize the user's wearing behavior. During the use process, the evolutionary model is periodically updated in a triggered manner. Abnormal physiological monitoring signals of the user are collected. Through the evolutionary model, the abnormal sources of the abnormal physiological monitoring signals are identified. Specifically, the abnormal sources include abnormalities caused by long-term changes in wearing behavior and abnormalities caused by changes in actual physiological state.

2. The evolutionary modeling method for smartwatches based on wearing behavior according to claim 1, characterized in that, After the step of constructing the wearing behavior feature sequence of the smartwatch according to a preset time sequence, the method further includes: Based on the construction process of the wearing behavior feature sequence, the feature interruption segment of the user's interaction with the smartwatch is identified, wherein the feature interruption segment specifically includes taking off the watch, not wearing it for a long time, and insufficient monitoring conditions; Determine whether the feature interruption segment is marked during the construction process; If so, a behavioral inertia parameter representing the resistance to changes in wearing behavior is introduced into the wearing behavior feature sequence. Based on the behavioral inertia parameter, the degree of tilt of the user's wearing behavior in maintaining the original state within a continuous period is identified. Based on the degree of tilt, the change direction identifier of the wearing behavior feature unit between adjacent units is dynamically collected to generate the evolution direction of the user's wearing behavior in the feature space. Specifically, the evolution direction includes evolution towards a stable state and evolution towards an unstable state.

3. The evolutionary modeling method for smartwatches based on wearing behavior according to claim 1, characterized in that, The step of obtaining information on the gradual changes in user wearing behavior during use also includes: Based on the changing relationship between adjacent wearing behavior units, candidate change segments are identified when the user's wearing behavior changes, wherein the candidate change segments are specifically time intervals with continuous observation value; Determine whether the candidate change segment has a cumulative effect, wherein the cumulative effect specifically means that it is continued or amplified in subsequent time periods; If not, the candidate change segment is marked as a local perturbation. During the change process, the coupling relationship between the change in wearing behavior characteristics and the stability parameter of wearing behavior is obtained. Based on the coupling relationship, the stage boundary of the change in wearing behavior is constructed. Specifically, the stage boundary includes the pre-change stage, the change transition stage, and the post-change stage.

4. The evolutionary modeling method for smartwatches based on wearing behavior according to claim 1, characterized in that, The step of collecting the user's abnormal physiological monitoring signals and identifying the source of the abnormality in the abnormal physiological monitoring signals through the evolutionary model further includes: Based on the abnormality type characterization of the abnormal physiological monitoring signal pre-processed by the smartwatch, the time range of the abnormal physiological monitoring signal is collected, wherein the time range specifically includes the start time, duration and end time; Determine whether the time interval can be aligned with the temporal structure information in the evolutionary model; If so, the evolutionary state information of the smartwatch is identified. Based on the evolutionary state information, the abnormal physiological monitoring signal is structurally matched with the wearing behavior evolution process to obtain the degree of consistency between the abnormal physiological monitoring signal and the wearing behavior evolution process in terms of change dimensions, and a corresponding structural correlation is generated. Specifically, the evolutionary state information includes the current wearing behavior stage, the wearing stability level corresponding to the stage, the evolution direction of the stage, and the change rate of the stage. The change dimensions specifically include the change rhythm, change direction, and change stage.

5. The evolutionary modeling method for smartwatches based on wearing behavior according to claim 1, characterized in that, The step of determining whether the implicit feature parameters can reflect the wearing behavior state of the smartwatch further includes: Based on the different wearing behavior states of the smartwatch, obtain the difference data of the implicit feature parameters within a preset time period; Determine whether the difference data detects a preset convergence behavior; If so, the characterization degradation information of the implicit feature parameters is identified. Based on the characterization degradation information, the assimilation influence parameters of the changing trend of the implicit feature parameters and the wearing adjustment behavior are collected, showing the same direction but not a responsive change. Based on the assimilation influence parameters, the slow unidirectional drift event of the smartwatch is constructed.

6. The evolutionary modeling method for smartwatches based on wearing behavior according to claim 1, characterized in that, The step of determining whether the gradual change information is within the normal fluctuation range corresponding to the preset reference baseline further includes: Based on the change path of the progressive change information, the change type of the progressive change information is identified, wherein the change type specifically includes continuous approximation type, step transition type and nonlinear bending type; Determine whether the change type exhibits a preset boundary fitting behavior; If so, a preset alternative reference baseline is introduced for parallel verification. Based on the verification results of the parallel verification, the baseline instability interval of the progressive change information is dynamically generated. Specifically, the alternative reference baseline includes a reference baseline formed in adjacent wearing stages, a statistical baseline of similar user groups, and a short-term adaptive update baseline.

7. The evolutionary modeling method for smartwatches based on wearing behavior according to claim 1, characterized in that, The step of identifying the user's implicit feature parameters based on the user's pre-detected wearing area by the smartwatch further includes: The smartwatch collects the pre-analyzed regional composition state of the wearing area, wherein the regional composition state specifically includes the contact tightness of the watch strap, the tilt state of the watch body relative to the skin, and the rotation state of the wearing area relative to the limb axis. Determine whether the state of the region is lower than a preset stability level; If so, the wearing area corresponding to the region composition state is marked as a structurally uncertain region, and based on the region composition state, a pool of hidden feature candidates that are related to the current wearing area structural state is dynamically activated.

8. An evolutionary modeling system for smartwatches based on wearing behavior, characterized in that, include: The identification module is used to identify the user's implicit feature parameters based on the wearing area pre-detected by the smartwatch. The implicit feature parameters specifically include signal amplitude stability features, periodic signal consistency features, and sensor signal correlation features. The judgment module is used to determine whether the implicit feature parameters can reflect the wearing behavior state of the smartwatch; The execution module is used to construct the wearing behavior feature sequence of the smartwatch according to a preset time sequence if possible, and to analyze the changing trend of the wearing behavior feature sequence using a preset distribution drift detection method to obtain the gradual change information of the user's wearing behavior during use. The gradual change information specifically includes the slow shift of the feature mean, the continuous change of feature fluctuation, and the reduction and enhancement of feature stability. The second judgment module is used to determine whether the gradual change information is within the normal fluctuation range corresponding to the preset reference baseline; The second execution module is configured to, if so, construct an evolutionary model to characterize the user's wearing behavior based on the progressive change information, periodically trigger updates to the evolutionary model during the use process, collect abnormal physiological monitoring signals of the user, and identify the abnormal source of the abnormal physiological monitoring signals through the evolutionary model, wherein the abnormal source specifically includes abnormalities caused by long-term changes in wearing behavior and abnormalities caused by changes in actual physiological state.

9. The smartwatch evolutionary modeling system based on wearing behavior according to claim 8, characterized in that, Also includes: The second identification module is used to identify the user's characteristic interruption segments on the smartwatch based on the construction process of the wearing behavior feature sequence, wherein the characteristic interruption segments specifically include taking off the watch, not wearing it for a long time, and insufficient monitoring conditions; The third judgment module is used to determine whether the feature interruption segment is marked during the construction process; The third execution module is used to, if so, introduce a behavioral inertia parameter representing the resistance to changes in wearing behavior into the wearing behavior feature sequence, identify the degree of tilt of the user's wearing behavior in maintaining the original state within a continuous period of time based on the behavioral inertia parameter, dynamically collect the change direction identifier of the wearing behavior feature unit between adjacent units based on the degree of tilt, and generate the evolution direction of the user's wearing behavior in the feature space, wherein the evolution direction specifically includes evolution towards a stable state and evolution towards an unstable state.

10. The smartwatch evolutionary modeling system based on wearing behavior according to claim 8, characterized in that, The execution module further includes: The identification submodule is used to identify candidate change segments when the user's wearing behavior changes based on the change relationship between adjacent wearing behavior units, wherein the candidate change segments are specifically time intervals with continuous observation value; The judgment submodule is used to determine whether the candidate change segment has a cumulative effect, wherein the cumulative effect is specifically that it is continued or amplified in subsequent time periods; The execution submodule is used to mark the candidate change segment as a local perturbation if no, obtain the coupling relationship between the change of wearing behavior characteristics and the wearing behavior stability parameter during the change process, and construct the stage boundary of the wearing behavior change based on the coupling relationship. The stage boundary specifically includes the pre-change stage, the change transition stage and the post-change stage.

Images