Old-age service remote monitoring and early warning method based on artificial intelligence

Abstract

The invention relates to the technical field of low-altitude emergency response, and discloses an old-age service remote monitoring and early warning method based on artificial intelligence, and the method comprises the steps: obtaining the spatial positioning data and environmental background data of an old person, recognizing an abnormal interruption point of a moving track based on the spatial positioning data, and when the abnormal interruption point of the moving track is detected, carrying out the early warning of the old person. Activating a local environment disturbance index in the environment background data, generating a behavior abnormal data set, associating a dynamic coupling relationship between the physiological monitoring parameters of the old people and the local environment disturbance index, generating physiological response offset characteristics, and determining the behavior abnormal data set based on the spatial distribution density change of the physiological response offset characteristics. Constructing a spatial risk transfer feature directed graph from the abnormal interruption point of the moving track to the high-risk area, and triggering a dynamic stress response strategy based on the conduction direction of the spatial risk transfer feature directed graph; according to the invention, the problems of early warning lag and incapability of realizing risk blocking in advance caused by data splitting and passive response in a remote monitoring method can be solved.

Description

Technical Field

[0001] This invention relates to the field of technology, and in particular to a method for remote monitoring and early warning of elderly care services based on artificial intelligence. Background Technology

[0002] Current technologies primarily rely on two independent data sources for risk assessment: one is the continuous collection of physiological indicators such as heart rate and blood pressure from wearable devices or environmental sensors, triggering an alarm when the data exceeds a static threshold; the other is monitoring the elderly's activities through video or positioning devices, identifying abnormalities through fall detection algorithms or prolonged periods of stillness. However, both methods lack the simultaneous acquisition and collaborative analysis of multimodal information such as behavioral abnormalities, environmental disturbances, and physiological responses. Existing technologies process these signals fragmentarily, leading to delayed identification of complex risk scenarios and a high false alarm rate, failing to provide effective early warnings before risks escalate.

[0003] Furthermore, existing early warning and response mechanisms are crude and passive. When the system alarms, it typically only sends a simple notification of the anomaly to the caregiver, failing to assess the spatiotemporal transmission trend of the risk, and lacking intelligent intervention capabilities based on the risk transmission path. This results in monitoring responses remaining at the post-event notification level, making it difficult to achieve pre-event risk prevention and dynamic proactive care, and the elderly's true safety status and psychological needs are not met in a closed loop. Therefore, there is an urgent need to develop an intelligent monitoring and early warning method that can deeply integrate multi-dimensional data on behavior, environment, and physiology, dynamically construct a risk spatial transmission model, and trigger precise, tiered proactive intervention strategies. This would address the problems of delayed early warning, high false alarm rates, and passive responses in existing technologies, effectively improving the timeliness, accuracy, and humanistic care level of elderly care monitoring. Summary of the Invention

[0004] This invention provides a remote monitoring and early warning method for elderly care services based on artificial intelligence, the main purpose of which is to address the problems raised in the background section above.

[0005] To achieve the above objectives, this invention provides a remote monitoring and early warning method for elderly care services based on artificial intelligence, comprising: S1. Obtain the elderly person's spatial positioning data and environmental background data; S2. Identify the abnormal interruption points of the elderly person's movement trajectory based on the spatial positioning data; S3. When an abnormal interruption point of the movement trajectory is detected, the local environmental disturbance index in the environmental background data is activated synchronously to generate the abnormal behavior dataset of the elderly. S4. Within the time window of the abnormal behavior dataset, dynamically couple the physiological monitoring parameters of the elderly with the local environmental disturbance indicators in real time to generate the physiological response shift characteristics of the elderly. S5. Based on the spatial distribution density change of the physiological response offset features, construct a directed graph of spatial risk transfer features from the abnormal interruption point of the movement trajectory to the high-risk area; S6. Trigger the elderly person's dynamic stress response strategy based on the transmission direction of the directed graph of the spatial risk transfer characteristics.

[0006] Optionally, obtaining the elderly person's spatial positioning data and environmental background data includes: The spatial positioning data of the elderly person is obtained based on the positioning base station network; The environmental background data of the elderly were acquired using distributed multimodal sensors.

[0007] Optionally, identifying the abnormal interruption points of the elderly person's movement trajectory based on the spatial positioning data includes: The spatial positioning data is converted into a continuous temporal location sequence of the elderly person. Based on the continuous time location sequence, calculate the probability of the elderly person's movement state transition under the preset behavior pattern; When the probability of the movement state transition is lower than the preset state transition threshold and the time-continuous location sequence has not been updated within a preset time period, the location point corresponding to the elderly person is marked as the abnormal interruption point of the elderly person's movement trajectory.

[0008] Optionally, when an abnormal interruption point in the movement trajectory is detected, the local environmental disturbance index in the environmental background data is simultaneously activated to generate the elderly person's abnormal behavior dataset, including: Centered on the point where the movement trajectory was abnormally interrupted, a spherical spatial area of ​​concern for the elderly was delineated; Within a preset time window, extract the regional environmental background data within the spherical space of interest area; The regional environmental background data includes sound spectrum energy distribution, object micro-vibration amplitude, and infrared thermal imaging temperature gradient. Principal component analysis was performed on the environmental background data of the region to calculate the collaborative change index of the spherical spatial region of interest. If the coordinated change index exceeds the preset disturbance threshold, then a subset of the regional environmental background data is marked as the local environmental disturbance index. The temporal and spatial coordinates of the abnormal interruption points of the movement trajectory are spatiotemporally aligned with the local environmental disturbance indicators to generate the abnormal behavior dataset of the elderly.

[0009] Optionally, the step of dynamically coupling the physiological monitoring parameters of the elderly person with the local environmental disturbance index in real time within the time window of the abnormal behavior dataset to generate the physiological response shift characteristics of the elderly person includes: Based on the physiological monitoring device worn by the elderly, heart rate variability frequency domain index, infrared thermal imaging temperature gradient and skin conductance response signal are obtained synchronously with the behavioral abnormality dataset; A recursive analysis is performed on the heart rate variability frequency domain index and the sound spectral energy to generate dynamic coupling strength; The skin electrical response signal and the infrared thermal imaging temperature gradient are interactively analyzed to generate an information flow intensity. The dynamic coupling strength and the information flow strength are integrated into the physiological response shift characteristics of the elderly.

[0010] Optionally, the formula for calculating the dynamic coupling strength is:

[0011] in, The dynamic coupling strength is... To define the start time of the time window for the behavioral anomaly dataset, To define the end time of the time window for the behavioral anomaly dataset, and The adjustment coefficient is obtained by learning from historical data. It is the natural logarithm function. For any consecutive moments within the time window For at any time Heart rate variability is the instantaneous shift of a certain frequency domain power relative to an individual's baseline. To correspond to the standard deviation of the individual baseline for frequency domain power of heart rate variability, Let be the acoustic disturbance effect function. For a moment The amplitude of the energy change in the sound spectrum. For at any time The vector magnitude of the infrared thermal imaging temperature gradient, For at any time The rate of change of skin conductance response signal.

[0012] Optionally, the construction of a directed graph of spatial risk transfer features from the abnormal interruption point of the movement trajectory to the high-risk area based on the spatial distribution density change of the physiological response offset features includes: Multiple consecutive physiological response offset feature vectors are mapped to a three-dimensional spatial grid; Density estimation is performed on the feature vector distribution of the three-dimensional spatial grid to form the feature density field of the three-dimensional spatial grid. The feature density field is decomposed into vector field to separate the divergence-free components of the feature density field. The abnormal interruption point of the movement trajectory is determined as the source point, and the path is deduced along the streamline direction of the non-dispersion component to obtain the path deduction result of the elderly. Based on a pre-set database of hazardous sources, the locations of hazardous sources in the path deduction results are marked as high-risk areas; The complete inference path structure from the source point to the high-risk area is encoded as a directed graph of spatial risk transfer characteristics.

[0013] Optionally, the preset hazard source database includes: coordinates of sharp furniture corners, boundaries of slippery floor areas, locations of medicine storage cabinets, and areas with concentrated power outlets.

[0014] Optionally, the dynamic stress response strategy triggered by the transmission direction of the directed graph based on the spatial risk transfer characteristics includes: Analyze the directed graph of spatial risk transfer characteristics to determine the transmission direction from the source point to the high-risk area; If the transmission direction points to the high-risk area, then the elderly person's dynamic stress response strategy is activated according to the risk type of the high-risk area.

[0015] Optionally, the method for remote monitoring and early warning of elderly care services based on artificial intelligence as described in claim 9 is characterized in that the dynamic stress response strategy includes: Using a smart speaker, targeted auditory intervention is performed on the elderly person using preset soothing voice messages; Along the transmission path, a guide light strip is dynamically generated by smart lamps to visually guide the elderly person away from the current trajectory; If the elderly person does not provide the expected feedback, the physical protection device in the high-risk area will be automatically activated, and a composite alarm message will be sent to the preset emergency contact.

[0016] Compared with the prior art, the present invention has the following beneficial effects: This method introduces synchronous activation and dynamic coupling logic through spatiotemporal collaborative analysis and deep fusion of multi-source heterogeneous data. When an abnormal interruption of the movement trajectory is detected, the system does not make an isolated judgment, but immediately and synchronously activates the environmental disturbance indicators of multiple physical fields such as sound, vibration, and heat around the corresponding spatiotemporal point. These environmental anomalies are then nonlinearly and dynamically coupled with physiological parameters such as the elderly person's heart rate variability and skin conductance within a unified time window. This multi-dimensional cross-validation mechanism, which uses behavioral abnormalities as clues, environmental disturbances as corroborating evidence, and physiological reactions as core verification, enables the system to accurately distinguish between genuine health and safety risks and ordinary temporary stillness. This significantly reduces the false alarm rate at the root and can capture early and subtle physiological deviations caused by environmental stress before significant and irreversible changes occur in physiological indicators, thus achieving early warning of risks.

[0017] After accurately identifying risks, this method further utilizes the spatial distribution density of continuous physiological response characteristics to extract the divergence-free components of risk transmission through vector field decomposition technology. This allows for the proactive deduction of the spatial transmission path of risks from abnormal locations to potentially high-risk areas, encoded as a directed graph. This triggers dynamic stress response strategies matched to the risk transmission path and level: first, gentle audio-visual guidance intervention via smart devices attempts to help the elderly autonomously escape the risk trajectory; if the expected feedback is not obtained, specific physical protective devices in the high-risk area are activated, and a precise alarm containing the risk type, location, and transmission path is simultaneously sent to emergency contacts. This forms a complete intelligent closed loop of accurate identification, path prediction, tiered intervention, and effect feedback, greatly improving the targeting and success rate of rescue efforts and demonstrating the proactive protection and humanistic care of technology for the safety of the elderly. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating an artificial intelligence-based remote monitoring and early warning method for elderly care services, provided as an embodiment of the present invention.

[0019] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0020] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0021] This application provides an artificial intelligence-based remote monitoring and early warning method for elderly care services. The executing entity of this artificial intelligence-based remote monitoring and early warning method for elderly care services includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application: a server, a terminal, etc. In other words, the artificial intelligence-based remote monitoring and early warning method for elderly care services can be executed by software or hardware installed on a terminal device or a server device, and the software can be a blockchain platform. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster. The server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0022] Reference Figure 1The diagram shown is a flowchart illustrating a remote monitoring and early warning method for elderly care services based on artificial intelligence, according to an embodiment of the present invention. In this embodiment, the remote monitoring and early warning method for elderly care services based on artificial intelligence includes: S1. Obtain the elderly person's spatial positioning data and environmental background data; In this embodiment, obtaining the elderly person's spatial positioning data and environmental background data includes: The spatial positioning data of the elderly person is obtained based on the positioning base station network; The environmental background data of the elderly were acquired using distributed multimodal sensors.

[0023] Specifically, spatial positioning data refers to digital information captured in real time or near real time through specific technical means, which can uniquely identify and reflect the precise three-dimensional location and time-varying trajectory of an elderly person in their living or activity space.

[0024] Specifically, environmental background data refers to a collection of multi-dimensional, multi-physical field information that is continuously collected by a series of non-invasive sensing devices within the same activity space of the elderly, reflecting the physical state and events of that space.

[0025] Furthermore, a comprehensive spatial coverage network of positioning base stations is deployed within the elderly living environments requiring monitoring. These base stations, serving as reference points with known fixed coordinates, are installed on the ceilings or walls of rooms, forming an invisible positioning field. The elderly person needs to wear a lightweight positioning tag integrated with a specific radio frequency or ultra-wideband signal transmission module.

[0026] When an elderly person moves around in their environment, the tag they wear continuously emits a wireless signal carrying a unique identifier. Multiple base stations in the network simultaneously receive this signal and accurately measure the propagation time or angle of arrival from the tag to each base station. The central processing unit aggregates the measurement data from all base stations and uses multi-point positioning technology to calculate the tag's precise coordinates in the current three-dimensional space in real time. This process is continuous, generating a high-frequency, high-precision spatial positioning data stream that accurately depicts the elderly person's real-time location and movement trajectory.

[0027] Furthermore, to achieve comprehensive and seamless environmental perception, a series of different types of environmental sensor nodes are deployed in key activity areas for the elderly, such as the living room, bedroom, corridor, and bathroom, to form a sensor network based on the functional and risk characteristics of each area.

[0028] These nodes may include: directional microphone arrays deployed in the corners of the room to collect ambient sounds and analyze their spectral energy distribution to identify abnormal sounds such as breaking glass or heavy objects falling; micro-vibration sensors attached to the bottom of key furniture to monitor the amplitude of minute vibrations caused by objects being bumped, moved, or leaned against; and wide-angle infrared thermal imagers installed on the ceiling to scan the entire room non-contactly and generate infrared thermal imaging temperature gradient maps that reflect the temperature distribution of human bodies and object surfaces to detect the presence of people, areas with abnormal body temperatures, or abnormal heat sources.

[0029] All sensor nodes synchronously acquire raw signals according to a unified clock. After local preliminary signal conditioning and analog-to-digital conversion, the processed feature data is transmitted to the central processing unit via the network for aggregation, alignment and fusion, ultimately forming a multimodal environmental background data stream that is strictly synchronized with the environmental spatial coordinates and timestamps.

[0030] In summary, acquiring the elderly person's spatial positioning data and environmental background data is the core task of the perception layer of the entire intelligent monitoring and early warning system. Spatial positioning data accurately delineates the dynamic coordinates of the monitored individual in the physical world, serving as the absolute reference frame for behavioral analysis. When subsequent steps detect abnormal behavior in the elderly person, the system can not only know where the person is, but also immediately retrieve and understand what environmental changes are occurring there. This provides crucial cross-validation information for determining whether the abnormality is due to a sudden health event, an accidental fall, or normal temporary rest, greatly improving the accuracy, reliability, and depth of contextual awareness in subsequent risk identification and assessment.

[0031] S2. Identify the abnormal interruption points of the elderly person's movement trajectory based on the spatial positioning data; In this embodiment, identifying the abnormal interruption points of the elderly person's movement trajectory based on the spatial positioning data includes: The spatial positioning data is converted into a continuous temporal location sequence of the elderly person. Based on the continuous time location sequence, calculate the probability of the elderly person's movement state transition under the preset behavior pattern; When the probability of the movement state transition is lower than the preset state transition threshold and the time-continuous location sequence has not been updated within a preset time period, the location point corresponding to the elderly person is marked as the abnormal interruption point of the elderly person's movement trajectory.

[0032] Specifically, spatial positioning data refers to a data stream that reflects the elderly person's real-time three-dimensional coordinates.

[0033] Specifically, abnormal interruption points in the movement trajectory refer to spatial locations in the elderly person's movement trajectory that are determined by the system analysis to be abnormal, unexpected stagnation or cessation of activity.

[0034] Specifically, a time-continuous location sequence refers to a data chain that arranges and connects original, discrete spatial positioning data in a strict chronological order to form a complete and coherent trajectory that reflects the continuous change of an elderly person's location from one moment to the next.

[0035] Specifically, a preset behavioral pattern refers to a set of regular behaviors that the system summarizes based on the elderly person's historical activity data over a long period of time, and that conform to their personal living habits.

[0036] Specifically, the mobility transition probability is an indicator used to quantitatively assess how much an elderly person's current behavior conforms to their individual normal pattern.

[0037] Specifically, the preset state transition threshold is a pre-set boundary value used to determine whether the probability of a state transition is as low as an abnormal level.

[0038] Specifically, the preset duration is a pre-defined time length value used to determine whether the unupdated state of the position sequence has lasted long enough to constitute an interruption.

[0039] Furthermore, the system receives raw spatial positioning data packets with timestamps from the positioning base station network. First, these data packets are strictly sorted according to their timestamps to ensure that the order from earliest to latest is correct.

[0040] Next, based on the preset data fusion frequency, interpolation or smoothing is performed to account for any minor timestamp deviations or data packet loss, generating a uniform and continuous sequence of location points on the timeline. Each location point contains a precise timestamp and a set of three-dimensional spatial coordinates.

[0041] Ultimately, this sequence exists in the system as a virtual trajectory line connected by countless points in space and time, accurately recording the spatial location of the elderly person at every moment over a period of time.

[0042] Furthermore, the system first invokes the personalized preset behavior pattern library established for the elderly person. Then, it matches and compares the real-time segments of the currently analyzed continuous time location sequence with the preset behavior patterns. The matching process is not a simple judgment of location overlap, but rather an analysis of whether the movement direction, speed, acceleration, and functional area reflected in the current trajectory segment conform to the typical statistical characteristics of a known behavior pattern at the same time and location.

[0043] Based on the degree of conformity of these multi-dimensional characteristics, the system calculates a probability value between 0 and 1 through an internal evaluation mechanism; this probability value represents the likelihood that the elderly person will continue to move in the current manner or transition to the next expected state, according to their historical habits.

[0044] Furthermore, the system continues to perform the aforementioned probability calculations and location update monitoring.

[0045] Simultaneously, two key conditions are checked: First, whether the calculated real-time motion state transition probability remains consistently below a preset state transition threshold; second, whether the spatial coordinates of the latest data point in the continuous time-series position sequence have not changed by more than a preset minimum displacement over a period exceeding a preset duration. When both conditions are met, the judgment logic is triggered. Once the judgment is successful, the system officially marks the last recorded stationary spatial coordinate point in the continuous time-series position sequence as a point of abnormal motion trajectory interruption. This point is assigned an event tag and associated with the time of occurrence, the duration of continuous stationary state, and an extremely low transition probability value, for subsequent in-depth analysis.

[0046] In summary, the system accurately and automatically filters out truly alarming abnormal pauses from continuous normal activity. The analysis is based on transforming raw location data into a continuous temporal location sequence. Personalized and intelligent anomaly detection is achieved by introducing preset behavioral patterns and movement state transition probabilities based on individual history, avoiding the incompatibility of fixed rules with individual differences. False alarms are significantly reduced by setting preset state transition thresholds and preset durations as dual judgment conditions. The final output of abnormal movement trajectory interruption points is not a simple coordinate, but a high-value risk event signal with anomaly intensity and temporal attributes, providing precise spatiotemporal guidance and decision-making basis, ensuring the efficient and accurate operation of the entire monitoring system.

[0047] S3. When an abnormal interruption point of the movement trajectory is detected, the local environmental disturbance index in the environmental background data is activated synchronously to generate the abnormal behavior dataset of the elderly. In this embodiment, when an abnormal interruption point in the movement trajectory is detected, the local environmental disturbance index in the environmental background data is simultaneously activated to generate the elderly person's abnormal behavior dataset, including: Centered on the point where the movement trajectory was abnormally interrupted, a spherical spatial area of ​​concern for the elderly was delineated; Within a preset time window, extract the regional environmental background data within the spherical space of interest area; The regional environmental background data includes sound spectrum energy distribution, object micro-vibration amplitude, and infrared thermal imaging temperature gradient. Principal component analysis was performed on the environmental background data of the region to calculate the collaborative change index of the spherical spatial region of interest. If the coordinated change index exceeds the preset disturbance threshold, then a subset of the regional environmental background data is marked as the local environmental disturbance index. The temporal and spatial coordinates of the abnormal interruption points of the movement trajectory are spatiotemporally aligned with the local environmental disturbance indicators to generate the abnormal behavior dataset of the elderly.

[0048] Specifically, the local environmental disturbance index refers to the set of data features extracted from the vast environmental background data after detecting an abnormal interruption point in the movement trajectory. These features are significant and involve coordinated changes in multiple physical fields that occur in a specific space and time near the abnormal point.

[0049] Specifically, the behavioral anomaly dataset is a structured collection of data. Its core is the result of accurately binding and associating the abnormal interruption points of the movement trajectory representing the elderly person's own behavioral abnormalities with local environmental disturbance indicators representing abnormal events occurring simultaneously in the environment.

[0050] Specifically, the spherical space of interest is a spherical three-dimensional space defined by the three-dimensional coordinates of the point where the movement trajectory is abnormally interrupted, with a preset length as the radius.

[0051] Specifically, the preset time window refers to a time interval before and after the moment when the abnormal interruption point of the movement trajectory occurs.

[0052] Specifically, regional environmental background data refers to a subset of environmental background data collected by sensors deployed in and around the defined spherical spatial area of ​​interest during a preset time window.

[0053] Specifically, sound spectrum energy distribution refers to the characteristic description extracted from regional environmental background data about how the energy of each frequency component of environmental sound changes over time within a preset time window.

[0054] Specifically, the amplitude of object micro-vibration refers to the sequence of intensity changes of minute mechanical vibrations on the surface of an object caused by force within a preset time window, extracted from regional environmental background data and recorded by micro-vibration sensors deployed on key furniture or facilities.

[0055] Specifically, the infrared thermal imaging temperature gradient refers to the rate and direction of temperature change in a spherical area of ​​interest captured by an infrared thermal imager within a preset time window, extracted from regional environmental background data.

[0056] Specifically, the Co-change Index is a quantitative indicator calculated by performing a multivariate comprehensive analysis of regional environmental background data.

[0057] Specifically, the preset disturbance threshold is a pre-set standard value used to determine whether the cooperative change index is high enough to confirm that a significant cooperative disturbance has occurred in the environment.

[0058] Furthermore, upon receiving information about the location of the abnormal interruption in the movement trajectory, the system immediately defines a spherical region in the 3D spatial model, centered on the coordinates of that location, based on preset radius parameters. The system overlays and matches this spherical region with a pre-constructed 3D digital map of the indoor space, automatically identifying all environmental sensors within this spherical area, such as specifically numbered microphones, vibration sensors, and thermal imager pixel areas. Simultaneously, the system records the geometric boundaries of this spherical region, preparing for subsequent spatial data filtering. This delineation process ensures that subsequent environmental analysis is strictly limited to the spatial area closest to and most relevant to the location of the elderly person's abnormal behavior.

[0059] Furthermore, the system expands forward and backward from the point where the movement trajectory abnormally interrupts, using this point as a reference. For example, the window is set from T1 seconds before the interruption to T2 seconds after the interruption. Next, based on the sensor list identified in the previous step and located within the spherical area of ​​interest, the system precisely extracts, by timestamp, all raw signals or pre-processed feature data generated by each relevant sensor within the preset time window from the continuously recorded full-volume environmental background data stream. For example, it extracts audio waveforms or spectral image segments from a specified microphone array within this time period; it extracts the acceleration amplitude sequence from a specific vibration sensor within this time period; and it extracts the temperature value matrix sequence of the corresponding spatial pixels from the thermal imager within this time period. All these sensor- and time-aligned data segments together constitute the regional environmental background data.

[0060] Furthermore, the system extracts regional environmental background data and performs standardized preprocessing to eliminate differences in the dimensions and magnitudes of different sensors. Then, the system examines these multivariate time series data as a whole, using principal component analysis, a data processing technique, to identify a few comprehensive directions that best represent the common trends of change from multiple dimensions such as sound, vibration, and thermal imaging.

[0061] After the analysis is completed, the eigenvalues ​​corresponding to the first principal component or the variance proportions they explain are extracted and quantified into a scalar value, which is the co-variance index. The level of this index directly reflects the overall intensity of consistent and drastic changes in the physical field signals of multiple environments within a preset time window and a spherical region of interest.

[0062] Furthermore, the calculated cooperative change index is compared with a preset disturbance threshold in the database for that environment. If the index value is greater than the threshold, it is determined that significant multi-field cooperative disturbances have indeed occurred within that spatiotemporal range. Once the determination is made, a labeling operation is performed: the environmental background data of the region corresponding to the high index is labeled with a specific disturbance, and separated from ordinary environmental data as a usable, meaningful data packet. This labeled data packet is the local environmental disturbance index. If the index does not exceed the threshold, it is considered that there is no significant cooperative anomaly in the environment, and the relevant data will not be labeled with this index.

[0063] Furthermore, a new data set structure is created. First, it writes all the attributes of the abnormal interruption point of the movement trajectory, including the timestamp of its occurrence, three-dimensional spatial coordinates, duration, and associated low state transition probability value, as the first core part of this set.

[0064] Next, the system will write the local environmental disturbance indicator data packet, which was marked when the determination was successful in this step, as the second core part. During the writing process, the system will ensure that the time range information attached to the local environmental disturbance indicator data packet is precisely aligned with the timestamp of the abnormal interruption point of the movement trajectory; Simultaneously, its spatial source information is also associated with the location coordinates. Through this precise spatiotemporal alignment and binding, a unified behavioral anomaly dataset containing evidence of behavioral anomalies and environmental synchronization anomalies is formally generated. This dataset is a complete event record, awaiting further in-depth comprehensive analysis in subsequent steps.

[0065] In summary, by defining a spherical area of ​​interest and a preset time window, the system achieves intelligent focusing on massive amounts of environmental data, significantly improving processing efficiency. By extracting multi-physics data such as sound spectrum energy distribution, object micro-vibration amplitude, and infrared thermal imaging temperature gradient, a three-dimensional, non-line-of-sight perception capability for the same event is constructed. By calculating the cooperative change index and comparing it with a preset disturbance threshold, the system achieves intelligent quantitative judgment of the intensity of environmental anomalies, avoiding interference caused by false alarms from a single sensor.

[0066] S4. Within the time window of the abnormal behavior dataset, dynamically couple the physiological monitoring parameters of the elderly with the local environmental disturbance indicators in real time to generate the physiological response shift characteristics of the elderly. In this embodiment, the step of dynamically coupling the physiological monitoring parameters of the elderly person with the local environmental disturbance index in real time within the time window of the abnormal behavior dataset to generate the physiological response shift characteristics of the elderly person includes: Based on the physiological monitoring device worn by the elderly, heart rate variability frequency domain index, infrared thermal imaging temperature gradient and skin conductance response signal are obtained synchronously with the behavioral abnormality dataset; A recursive analysis is performed on the heart rate variability frequency domain index and the sound spectral energy to generate dynamic coupling strength; The skin electrical response signal and the infrared thermal imaging temperature gradient are interactively analyzed to generate an information flow intensity. The dynamic coupling strength and the information flow strength are integrated into the physiological response shift characteristics of the elderly.

[0067] The formula for calculating the dynamic coupling strength is:

[0068] in, The dynamic coupling strength is... To define the start time of the time window for the behavioral anomaly dataset, To define the end time of the time window for the behavioral anomaly dataset, and The adjustment coefficient is obtained by learning from historical data. It is the natural logarithm function. For any consecutive moments within the time window For at any time Heart rate variability is the instantaneous shift of a certain frequency domain power relative to an individual's baseline. To correspond to the standard deviation of the individual baseline for frequency domain power of heart rate variability, Let be the acoustic disturbance effect function. For a moment The amplitude of the energy change in the sound spectrum. For at any time The vector magnitude of the infrared thermal imaging temperature gradient, For at any time The rate of change of skin conductance response signal.

[0069] Specifically, the time window of a behavioral anomaly dataset refers to the specific, defined start and end time range associated with the behavioral anomaly dataset.

[0070] Specifically, physiological monitoring parameters refer to biological signals that reflect the core vital signs and autonomic nervous system status of the elderly, which are continuously collected and transmitted through specialized medical or health monitoring devices worn by the elderly.

[0071] Specifically, local environmental disturbance indicators refer to a set of labeled, spatiotemporally aligned, multiphysics-based environmental anomaly data features that are aligned with behavioral interruptions.

[0072] Specifically, dynamic coupling refers to the fact that the physiological monitoring parameters of the elderly and the indicators of local environmental disturbances do not change in isolation within a time window, but rather there is a synergistic evolution relationship in which they influence each other, follow each other, and are related to each other.

[0073] Specifically, physiological response offset features are a comprehensive, quantitative feature vector or set of indicators.

[0074] Specifically, the heart rate variability frequency domain index is a characteristic value calculated from the heartbeat interval sequence within a time window obtained from the electrocardiogram monitoring device worn by the elderly through frequency domain analysis.

[0075] Specifically, infrared thermal imaging temperature gradient refers to the spatial variation characteristics of body surface temperature extracted from a thermal imaging sequence taken within a time window of an elderly person's body using a non-contact infrared thermal imager.

[0076] Specifically, the skin conductance signal is an electrical signal that reflects the activity level of the skin's sweat glands, collected by a skin conductance sensor worn by the elderly.

[0077] Specifically, sound spectrum energy is a component of local environmental disturbance indicators. It refers to the data on the change of energy distribution of environmental sound at various frequencies near the abnormal location over time within an abnormal time window.

[0078] Specifically, recursive analysis is a data processing method used to analyze the dynamic relationship between two time series.

[0079] Specifically, dynamic coupling strength is used to comprehensively characterize the degree of correlation and synergistic change between the two time series—heart rate variability frequency domain index and sound spectral energy—in terms of dynamic behavior within the analysis time window.

[0080] Specifically, interactive analysis is a data processing method used to measure the direction and magnitude of information transfer or causal relationship between two signals.

[0081] Specifically, information flow intensity is a quantitative result obtained through interaction analysis.

[0082] Furthermore, the system first reads the precise start and end timestamps of the event occurrence time window recorded in the behavioral abnormality dataset. Then, the system sends a command to the gateway managing the physiological monitoring devices, requesting the retrieval of the raw physiological data streams collected by different devices within this specific time window for the elderly individual. This includes: acquiring raw electrocardiogram waveform data from the electrocardiogram monitoring device; acquiring infrared image sequences of the elderly person's body surface area from the infrared thermal imager; and acquiring raw skin conductivity data from the skin conductivity sensor.

[0083] Upon acquiring the raw data, the system immediately performs preprocessing and feature extraction: R-wave detection is performed on the electrocardiogram (ECG), adjacent heartbeat interval sequences are calculated, and spectral analysis is conducted on these sequences to obtain the frequency domain index of heart rate variability; face or hand region recognition and tracking are performed on the infrared image sequences, the temperature distribution within the region is calculated, and the infrared thermal imaging temperature gradient features changing over time are extracted; the raw skin conductivity signal is filtered and denoised, and its skin conductance response signal changing over time is calculated. All extracted feature signals are strictly aligned to the time axis of the behavioral anomaly dataset.

[0084] Furthermore, the system first standardizes the aligned heart rate variability frequency domain index and sound spectral energy. Then, the system performs phase space reconstruction on these two time series respectively, mapping the one-dimensional time series to trajectories in a high-dimensional space.

[0085] Then, the system calculates the recursive properties of their respective evolution trajectories in these two reconstructed phase spaces, paying particular attention to the probability and structure of recursive point pairs between the two trajectories appearing in the state space. By quantifying this recursive similarity across system trajectories, the system calculates a comprehensive metric, which is the dynamic coupling strength.

[0086] Furthermore, the aligned time series of the skin conductance response (SCR) signal and the time series of the infrared thermography temperature gradient are input into the analysis module. First, the statistical dependence between these two sequences is assessed. Then, using a specific information-theoretic metric, the directional information transfer from the infrared thermography temperature gradient sequence to the SCR signal sequence is calculated. This calculation excludes the influence of the sequence's own historical values, focusing on assessing the extent to which changes in the ambient temperature gradient at the previous time step reduce the uncertainty of the SCR signal at the next time step. The average value of this calculated directional information transfer is quantified as the information flow intensity.

[0087] Furthermore, the two scalar values ​​of dynamic coupling strength and information flow strength calculated in the previous two steps are treated as two components of a two-dimensional vector. These are then integrated into a unified, multidimensional physiological response offset feature. This feature may be a vector containing these two intensity values ​​and their ratio, or it may be transformed into points in a higher-dimensional or more abstract feature space through a nonlinear mapping function. This final physiological response offset feature, as a whole, encapsulates the comprehensive strength and pattern information of the coupling between the elderly's cardiac autonomic nervous system response and cutaneous sympathetic nervous system response and environmental disturbances during abnormal events.

[0088] In summary, the above steps strongly correlate external behavior and environmental anomalies with internal, objective physiological evidence, thus providing a physiological basis for determining the true impact of abnormal events on the health of the elderly. By acquiring frequency domain indices of heart rate variability, infrared thermography temperature gradients, and skin conductance signals that are completely synchronized with abnormal events, the system advances the analysis to the level of vital signs. Through recursive analysis to generate dynamic coupling strength, the system quantifies the immediate impact of environmental noise disturbance on cardiac autonomic nerve function; through interactive analysis to generate information flow intensity, the system quantifies the induced efficacy of drastic changes in environmental temperature on sympathetic nerve excitation.

[0089] Specifically, dynamic coupling strength It is a scalar value calculated using the above formula, used to comprehensively quantify the time window in which abnormal behavioral events occur. Internally, the intensity of the overall and cumulative stress response of the elderly's physiological system to two simultaneous key environmental disturbances.

[0090] Specifically, the start time of the time window and the end time Together, they define the precise time range for calculating coupling strength.

[0091] Specifically, the adjustment coefficient and , are two weighting parameters learned through analysis of the elderly person's long-term historical data. They are used to adjust the contribution ratio of the center-acoustic coupling term and the skin conductance-temperature coupling term to the final total coupling strength, respectively.

[0092] Specifically, the natural logarithm function `<value>` is a mathematical function used to perform a non-linear transformation on the value within its parentheses. It converts the offset ratio of heart rate variability into a logarithmic value whose growth gradually slows.

[0093] Specifically, continuous time , is the integral variable, representing the time window Each point in time within the continuous flow.

[0094] Specifically, instantaneous offset It refers to a specific moment. The difference between the power value of an elderly person's heart rate variability in a specific frequency band and the average power value of that frequency band when he is at rest.

[0095] Specifically, standard deviation , is a measure of the statistical dispersion of the frequency band power of the elderly person's heart rate variability around their personal baseline.

[0096] Specifically, the acoustic disturbance influence function It is a measure of the amplitude of abrupt changes in the energy of the sound spectrum. The input is a function. Its output value is between 0 and 1, used to characterize the time interval [0, 1]. The potential stress response of the detected sound mutations.

[0097] Specifically, the amplitude of the change in sound spectrum energy , refers to the time The energy value of ambient sound near an anomaly location within a specific frequency band of interest, relative to the sudden increase in the background noise level of that area.

[0098] Specifically, the vector magnitude of the infrared thermal imaging temperature gradient , refers to the time The intensity value of the direction of the most drastic temperature change in the space near the anomalous location, measured by an infrared thermal imager. It is obtained by calculating the spatial derivative of the temperature field, and its modulus reflects the degree of inhomogeneity or drastic change in the local thermal environment.

[0099] Specifically, the rate of change of skin conductance response signal , refers to the time The rate of change of skin conductance signals in the elderly over time is usually obtained by calculating the derivative or difference of the signal over a very short period of time.

[0100] In summary, this calculation formula is the essence of quantifying physiological-environment interactions in this method. It unifies two different types of physiological-environment interactions under a single mathematical framework for continuous-time quantification and fusion. By performing logarithmic normalization of heart rate deviation based on individual baselines and introducing an influence function for sound disturbances, the formula fully considers individual differences and stimulus specificity, making the calculation results more physiologically meaningful and individually adaptive. Through time integration and averaging, the formula reflects the persistence and cumulative effect of stress responses, more closely resembling real physiological processes. The final calculated... As a robust and comprehensive quantitative indicator, the risk value can reliably distinguish between real risk events with strong physiological coupling and false alarms with almost no physiological coupling, greatly enhancing the objectivity, accuracy and reliability of system risk assessment and providing key scientific data support for subsequent steps.

[0101] S5. Based on the spatial distribution density change of the physiological response offset features, construct a directed graph of spatial risk transfer features from the abnormal interruption point of the movement trajectory to the high-risk area; In this embodiment, constructing a directed graph of spatial risk transfer features from the abnormal interruption point of the movement trajectory to the high-risk area based on the spatial distribution density change of the physiological response offset features includes: Multiple consecutive physiological response offset feature vectors are mapped to a three-dimensional spatial grid; Density estimation is performed on the feature vector distribution of the three-dimensional spatial grid to form the feature density field of the three-dimensional spatial grid. The feature density field is decomposed into vector field to separate the divergence-free components of the feature density field. The abnormal interruption point of the movement trajectory is determined as the source point, and the path is deduced along the streamline direction of the non-dispersion component to obtain the path deduction result of the elderly. Based on a pre-set database of hazardous sources, the locations of hazardous sources in the path deduction results are marked as high-risk areas; The complete inference path structure from the source point to the high-risk area is encoded as a directed graph of spatial risk transfer characteristics.

[0102] The preset hazard source database includes: coordinates of sharp furniture corners, boundaries of wet and slippery floor areas, locations of medicine storage cabinets, and areas with concentrated power outlets.

[0103] Specifically, the spatial distribution density change of physiological response offset features refers to the differences and trends in the degree of aggregation or sparsity of discrete physiological response offset feature vectors in different spatial regions after they are placed into a three-dimensional environmental model according to the spatial location of the elderly person corresponding to the time of their generation.

[0104] Specifically, the abnormal interruption point of the movement trajectory refers to the spatial location point where the elderly person's movement stopped abnormally; it is the starting spatial coordinate of the entire abnormal event. Specifically, a high-risk area refers to a specific spatial location or region that, after analysis in this step, poses an immediate physical threat to the elderly.

[0105] Specifically, the directed graph of spatial risk transfer characteristics is a risk transmission model represented by a graphical structure. It starts at the point where the movement trajectory is abnormally interrupted and ends at the high-risk area. The connecting path in between represents the potential spatial direction and path of risk transmission from the starting point to the ending point.

[0106] Specifically, the physiological response offset feature vector is a quantitative feature data that characterizes the comprehensive physiological stress state of the elderly.

[0107] Specifically, a three-dimensional spatial grid is a regular three-dimensional cell array formed by discretizing the physical space of the elderly person's actual living environment in the digital world.

[0108] Specifically, the characteristic density field is a continuous scalar field distribution formed on a three-dimensional spatial grid after density estimation operations.

[0109] Specifically, the divergence-free component is a mathematical component obtained by performing a vector field decomposition operation on the characteristic density field.

[0110] Specifically, the source point refers to the spatial location that is identified as the starting point for risk transmission analysis, i.e., the point where the movement trajectory is abnormally interrupted.

[0111] Specifically, streamline direction refers to the tangent direction of the continuous curve formed by the instantaneous velocity direction of an imaginary tiny particle moving in a field described by divergence components.

[0112] Specifically, the path deduction result refers to one or more continuous spatial trajectories obtained by gradually tracing and extending along the streamline direction without divergence components, starting from the aforementioned source point.

[0113] Specifically, the pre-set hazard source database is a pre-established electronic database that stores the precise coordinates of all known potential physical hazards or areas in the elderly person's living environment.

[0114] Furthermore, the system first acquires a series of physiological response offset feature vectors generated within the time window of the abnormal event, each vector bearing a precise timestamp of its generation. Then, based on these timestamps, the system retrieves the three-dimensional spatial coordinates of the elderly person's location from a continuous temporal location sequence.

[0115] Next, the system places these feature vectors into the established 3D spatial grid: determining which grid cell each feature vector's coordinates fall into, and storing that vector as a data record for that grid cell. If multiple feature vectors fall into a grid cell, they will all be recorded under that cell. This process organizes discrete feature data with spatial attributes into a structured spatial framework.

[0116] Furthermore, after completing the eigenvector mapping, the system traverses the entire 3D spatial mesh. For each mesh cell, the system considers not only the number of existing eigenvectors within that cell but also the distribution of eigenvectors in its neighboring cells. Using a spatial statistical method, the system calculates a density value for each mesh cell. This value depends not only on the number of eigenvectors within the cell but also on the number and distance of eigenvectors in nearby cells, thus forming a smooth and continuous density distribution.

[0117] Ultimately, each grid cell is assigned a density value, and the distribution of all these density values ​​across the entire 3D grid constitutes a continuous characteristic density field. This field visually displays where the hotspots of physiological stress responses are located in 3D space.

[0118] Furthermore, the system treats the feature density field as a scalar field and calculates its gradient. The gradient itself is a vector field pointing in the direction of the fastest increase in density. The system then performs a mathematical decomposition on this gradient vector field. This decomposition aims to break down the original gradient field into several subfields with different physical meanings. One key subfield is the divergence-free component.

[0119] The system extracts the portion of the gradient field that satisfies the condition of zero divergence through specific mathematical operations. The streamlines of this separated divergence-free vector field can better describe the organized, cyclical, or conductive trends in density changes, while filtering out the parts that are randomly diffused or uniformly converged.

[0120] Furthermore, the system first locates the precise grid cell where the movement trajectory is abnormally interrupted in three-dimensional space, and sets this point as the starting point of path tracing, i.e. the source point.

[0121] Then, the system queries the direction of the divergence-free vector field at that point. Starting from the source point, the system moves forward a small step along the field direction at that point, reaching a new position. At the new position, the system queries the field direction again and continues to move along the new direction. This process is like letting a point drift along the streamlines of the vector field. The system continues this iterative tracking until the preset tracking boundary conditions are reached.

[0122] Ultimately, this spatial trajectory, formed by connecting consecutive small steps, is the result of path deduction. It depicts the possible spatial extension trend of physiological stress hotspots starting from the anomalous points.

[0123] Furthermore, the path derivation results obtained in the previous step are compared with the preset hazard source database for spatial geometric relationships. The system checks whether the derivation path passes through, approaches, or terminates within the coordinates of any hazard source or area boundary recorded in the database. Once it is found that the path intersects with or is sufficiently close to a known hazard source, the system immediately marks the specific location of that hazard source as a high-risk area for this event. If the path does not intersect with any known hazard source, the system may mark the path endpoint or determine that there is no directly high-risk area.

[0124] Furthermore, the system creates a graphical data structure. It sets the source point as the starting node of the graph and the location of the high-risk area as the ending node.

[0125] Then, the key spatial points or turning points traversed by the path deduction result are set as intermediate nodes of the graph. Next, these nodes are connected sequentially with directed edges according to the actual order of the path, pointing from the starting node to the ending node, forming a directed path.

[0126] Ultimately, this complete graphical structure, encompassing a set of nodes, a set of edges, and spatial coordinate attributes, constitutes the directed graph of spatial risk transfer characteristics. This graph intuitively encapsulates the core spatial logic of where risk originates, along which path it travels, and where it ends.

[0127] In summary, by constructing a characteristic density field and analyzing its divergence-free components, the above steps enable the system to discern organized risk transmission trends hidden behind spatial distributions from seemingly discrete physiological response data. Through path extrapolation, the system achieves a forward-looking simulation of the direction of risk movement. By comparing with a hazard source database and marking high-risk areas, the system translates abstract transmission trends into concrete, threatening targets.

[0128] The resulting directed graph of spatial risk transfer characteristics is a powerful tool for situational understanding and decision support. It enables the early warning system to have predictive capabilities, answering the question of what the next specific danger is most likely to be if the elderly person's current discomfort or risk state continues, and from which path the danger will approach. This provides an irreplaceable scientific basis for triggering precise response strategies with spatial predictability and proactive blocking capabilities in subsequent steps, greatly improving the initiative and intelligence level of monitoring.

[0129] S6. Trigger the elderly person's dynamic stress response strategy based on the transmission direction of the directed graph of the spatial risk transfer characteristics.

[0130] In this embodiment, the dynamic stress response strategy triggered by the transmission direction of the directed graph based on the spatial risk transfer characteristics includes: Analyze the directed graph of spatial risk transfer characteristics to determine the transmission direction from the source point to the high-risk area; If the transmission direction points to the high-risk area, then the elderly person's dynamic stress response strategy is activated according to the risk type of the high-risk area.

[0131] In this embodiment, the dynamic stress response strategy includes: Using a smart speaker, targeted auditory intervention is performed on the elderly person using preset soothing voice messages; Along the transmission path, a guide light strip is dynamically generated by smart lamps to visually guide the elderly person away from the current trajectory; If the elderly person does not provide the expected feedback, the physical protection device in the high-risk area will be automatically activated, and a composite alarm message will be sent to the preset emergency contact.

[0132] Specifically, the spatial risk transfer characteristic directed graph refers to a data model that encapsulates the risk transmission logic in a graphical structure.

[0133] Specifically, the transmission direction refers to the spatial direction from the source point to the high-risk area as clearly indicated by the directed edge in the directed graph of spatial risk transfer characteristics.

[0134] Specifically, the source point refers to the identified point where the movement trajectory is abnormally interrupted, which serves as the starting point for risk transmission analysis. It represents the initial location where the elderly person may currently be in a state of risk or discomfort.

[0135] Specifically, high-risk areas are spatial locations that pose a specific and immediate threat to the elderly, as identified through path deduction and comparison with a hazard source database, such as slippery surfaces and sharp table corners.

[0136] Specifically, dynamic stress response strategy is a set of multi-level proactive intervention and protection measures that are automatically triggered and executed based on real-time analysis of the risk transmission situation.

[0137] Specifically, the preset soothing voice is pre-recorded or synthesized and stored in the system, specifically designed to provide emotional comfort and cognitive guidance to elderly people when they may feel stressed, confused, or uncomfortable.

[0138] Specifically, directional auditory intervention refers to the system using spatial acoustics technology to control the playback of a smart speaker, making the preset soothing voice sound as if it is coming from the elderly person's location or a safe direction they are expected to go, thereby achieving spatial auditory guidance and directional communication for the elderly person, enhancing the accuracy and immersion of the intervention.

[0139] Specifically, the transmission path is the spatial trajectory depicted in the directed graph of spatial risk transfer characteristics, from the source point to the high-risk area.

[0140] Specifically, physical protective devices refer to physical safety equipment that is pre-installed near high-risk areas and can be activated remotely by the system.

[0141] Specifically, composite alarm information refers to a detailed notification sent by the system to emergency contacts that not only includes simple abnormal alarms, but also integrates key information about the entire event.

[0142] Furthermore, firstly, the system identifies the starting and ending nodes in the graph. Next, the system traverses the directed edge sequence connecting these two nodes, analyzing the composite direction of the direction vectors of these edges in three-dimensional space. The system calculates the overall vector direction from the source point coordinates to the high-risk area coordinates, and combines this with the order of the path nodes in the directed graph to make a clear topological judgment: whether there exists a directed path starting from the source point and ultimately pointing to the high-risk area. If it exists, the transmission direction is determined to point to the high-risk area, indicating a clear trend of risk development towards that specific hazard source. If it does not exist, the transmission direction is determined to be invalid or unable to point to the specific high-risk area.

[0143] Furthermore, once the system determines that the transmission direction is valid, it immediately queries a predefined response strategy template and parameters that match the risk type registered in the hazard source database for the high-risk area. For example, for slippery risks, the strategy might prioritize voice reminders about slippery ground and increased path lighting; for sharp risks, the strategy might focus more on activating physical protective devices. Based on the strategy template, the system generates a specific, executable sequence of instructions and immediately initiates the execution process.

[0144] Furthermore, based on the elderly person's real-time location and the indoor acoustic model, the system calculates a list of smart speakers that need to be activated to make the sound appear to come from a specific direction, along with their respective volume and phase parameters. Then, the system selects a pre-set reassuring voice message from its voice library that best matches the current risk type and the elderly person's personalized settings. Next, the system sends a coordinated playback command to the selected group of smart speakers, ensuring that the elderly person can clearly hear the appropriately reassuring voice message from the target direction, thus completing a precise directional auditory intervention.

[0145] Furthermore, the system acquires the spatial coordinate sequence of the conduction path. Then, it controls all smart lights located along and near this path. The system instructs these lights to sequentially illuminate at higher brightness according to the path's order, and uniformly adjust them to a specific, eye-catching color temperature, thus forming a continuous, prominent guide light strip on the ground or wall. Further, the system can control the light strip to create a visual effect of flowing from the elderly person's current position towards a safe area to the side of the path, clearly indicating a desired direction of movement, attempting to guide the elderly person instinctively to move in the opposite direction or to the side of the light strip's flow, thereby deviating from their current trajectory and moving away from the risky path.

[0146] Furthermore, while performing the aforementioned audio-visual interventions, the system activates a feedback monitoring timer and continuously monitors the elderly person's location movement and physiological signals. If, within a preset reasonable timeframe, the system detects that the elderly person's location has not significantly changed to avoid deviating from the risk path, or that their physiological stress signals continue to worsen, it is determined that the expected feedback has not been generated. At this point, the system immediately performs an escalation operation: First, it sends an activation command to the physical protective devices deployed in the high-risk area, causing them to quickly deploy or activate and physically isolate the source of danger. Almost simultaneously, the system automatically generates a composite alarm message, which packages all key data of the event, and immediately sends it to all preset emergency contacts via SMS, app push notifications, or automatic telephone dialing.

[0147] In summary, the above steps determine the transmission direction by analyzing the directed graph, ensuring the necessity and relevance of the response. The activated dynamic stress response strategy embodies an intelligent, tiered processing logic: first, gentle directional auditory intervention is used to attempt to awaken and guide the elderly to avoid danger independently; if this fails, a more explicit escape route is provided through visual guidance light strips; if still unsuccessful, the ultimate protective measure is decisively taken—activating physical protective devices to block physical risks and sending a composite alarm message to call for human rescue.

[0148] This invention respects and utilizes the elderly's autonomy to the greatest extent possible, only resorting to mandatory protection and external intervention when necessary, thus achieving a balance between technological intervention and humanistic care. Ultimately, this method not only issues an alarm but also substantially participates in the risk mitigation process, greatly improving the effectiveness, timeliness, and reliability of emergency protection for elderly people living alone, truly putting intelligent monitoring to the ultimate goal of protection.

[0149] Therefore, the embodiments should be regarded as exemplary and non-limiting in all respects. The scope of the invention is not limited to the foregoing description, and all variations within the meaning and scope of equivalents falling within the protection scope are intended to be included in the invention.

[0150] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence (AI) refers to the theories, methods, technologies, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0151] Furthermore, the inclusion of a single word does not exclude other units or steps, and the singular does not exclude the plural. Terms such as "first," "second," etc., are used to indicate names and do not indicate any particular order.

[0152] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A remote monitoring and early warning method for elderly care services based on artificial intelligence, characterized in that, The method includes: S1. Obtain the elderly person's spatial positioning data and environmental background data; S2. Identify the abnormal interruption points of the elderly person's movement trajectory based on the spatial positioning data; S3. When an abnormal interruption point of the movement trajectory is detected, the local environmental disturbance index in the environmental background data is activated synchronously to generate the abnormal behavior dataset of the elderly. S4. Within the time window of the abnormal behavior dataset, dynamically couple the physiological monitoring parameters of the elderly with the local environmental disturbance indicators in real time to generate the physiological response shift characteristics of the elderly. S5. Based on the spatial distribution density change of the physiological response offset features, construct a directed graph of spatial risk transfer features from the abnormal interruption point of the movement trajectory to the high-risk area; S6. Trigger the elderly person's dynamic stress response strategy based on the transmission direction of the directed graph of the spatial risk transfer characteristics.

2. The method for remote monitoring and early warning of elderly care services based on artificial intelligence as described in claim 1, characterized in that, The acquisition of the elderly person's spatial positioning data and environmental background data includes: The spatial positioning data of the elderly person is obtained based on the positioning base station network; The environmental background data of the elderly were acquired using distributed multimodal sensors.

3. The method for remote monitoring and early warning of elderly care services based on artificial intelligence as described in claim 1, characterized in that, The step of identifying abnormal interruption points in the elderly person's movement trajectory based on the spatial positioning data includes: The spatial positioning data is converted into a continuous temporal location sequence of the elderly person. Based on the continuous time location sequence, calculate the probability of the elderly person's movement state transition under the preset behavior pattern; When the probability of the movement state transition is lower than the preset state transition threshold and the time-continuous location sequence has not been updated within a preset time period, the location point corresponding to the elderly person is marked as the abnormal interruption point of the elderly person's movement trajectory.

4. The method for remote monitoring and early warning of elderly care services based on artificial intelligence as described in claim 1, characterized in that, When an abnormal interruption point in the movement trajectory is detected, the local environmental disturbance index in the environmental background data is simultaneously activated to generate the elderly person's abnormal behavior dataset, including: Centered on the point where the movement trajectory was abnormally interrupted, a spherical spatial area of ​​concern for the elderly was delineated; Within a preset time window, extract the regional environmental background data within the spherical space of interest area; The regional environmental background data includes sound spectrum energy distribution, object micro-vibration amplitude, and infrared thermal imaging temperature gradient. Principal component analysis was performed on the environmental background data of the region to calculate the collaborative change index of the spherical spatial region of interest. If the coordinated change index exceeds the preset disturbance threshold, then a subset of the regional environmental background data is marked as the local environmental disturbance index. The temporal and spatial coordinates of the abnormal interruption points of the movement trajectory are spatiotemporally aligned with the local environmental disturbance indicators to generate the abnormal behavior dataset of the elderly.

5. The method for remote monitoring and early warning of elderly care services based on artificial intelligence as described in claim 4, characterized in that, Within the time window of the abnormal behavior dataset, the dynamic coupling relationship between the elderly person's physiological monitoring parameters and the local environmental disturbance indicators is dynamically correlated in real time to generate the elderly person's physiological response shift characteristics, including: Based on the physiological monitoring device worn by the elderly, heart rate variability frequency domain index, infrared thermal imaging temperature gradient and skin conductance response signal are obtained synchronously with the behavioral abnormality dataset; A recursive analysis is performed on the heart rate variability frequency domain index and the sound spectral energy to generate dynamic coupling strength; The skin electrical response signal and the infrared thermal imaging temperature gradient are interactively analyzed to generate an information flow intensity. The dynamic coupling strength and the information flow strength are integrated into the physiological response shift characteristics of the elderly.

6. The method for remote monitoring and early warning of elderly care services based on artificial intelligence as described in claim 5, characterized in that, The formula for calculating the dynamic coupling strength is:

7. Among them, The dynamic coupling strength is... To define the start time of the time window for the behavioral anomaly dataset, To define the end time of the time window for the behavioral anomaly dataset, and The adjustment coefficient is obtained by learning from historical data. It is the natural logarithm function. For any consecutive moments within the time window For at any time Heart rate variability is the instantaneous shift of a certain frequency domain power relative to an individual's baseline. To correspond to the standard deviation of the individual baseline for frequency domain power of heart rate variability, Let be the acoustic disturbance effect function. For a moment The amplitude of the energy change in the sound spectrum. For at any time The vector magnitude of the infrared thermal imaging temperature gradient, For at any time The rate of change of skin conductance response signal.

8. The method for remote monitoring and early warning of elderly care services based on artificial intelligence as described in claim 1, characterized in that, The spatial distribution density change based on the physiological response offset features is used to construct a directed graph of spatial risk transfer features from the abnormal interruption point of the movement trajectory to the high-risk area, including: Multiple consecutive physiological response offset feature vectors are mapped to a three-dimensional spatial grid; Density estimation is performed on the feature vector distribution of the three-dimensional spatial grid to form the feature density field of the three-dimensional spatial grid. The feature density field is decomposed into vector field to separate the divergence-free components of the feature density field. The abnormal interruption point of the movement trajectory is determined as the source point, and the path is deduced along the streamline direction of the non-dispersion component to obtain the path deduction result of the elderly. Based on a pre-set database of hazardous sources, the locations of hazardous sources in the path deduction results are marked as high-risk areas; The complete inference path structure from the source point to the high-risk area is encoded as a directed graph of spatial risk transfer characteristics.

9. The method for remote monitoring and early warning of elderly care services based on artificial intelligence as described in claim 7, characterized in that, The preset hazard source database includes: coordinates of sharp furniture corners, boundaries of wet and slippery floor areas, locations of medicine storage cabinets, and areas with concentrated power outlets.

10. The method for remote monitoring and early warning of elderly care services based on artificial intelligence as described in claim 7, characterized in that, The dynamic stress response strategy for the elderly triggered by the transmission direction of the directed graph based on the spatial risk transfer characteristics includes: Analyze the directed graph of spatial risk transfer characteristics to determine the transmission direction from the source point to the high-risk area; If the transmission direction points to the high-risk area, then the elderly person's dynamic stress response strategy is activated according to the risk type of the high-risk area.

11. The method for remote monitoring and early warning of elderly care services based on artificial intelligence as described in claim 9, characterized in that, The dynamic stress response strategy includes: Using a smart speaker, targeted auditory intervention is performed on the elderly person using preset soothing voice messages; Along the transmission path, a guide light strip is dynamically generated by smart lamps to visually guide the elderly person away from the current trajectory; If the elderly person does not provide the expected feedback, the physical protection device in the high-risk area will be automatically activated, and a composite alarm message will be sent to the preset emergency contact.

Images