AI human body motion capture sensor and automatic control method thereof

Abstract

The invention relates to the technical field of motion capture, and provides an AI human body motion capture sensor and an automatic control method thereof, and the AI human body motion capture sensor comprises a dynamic region identification system, an abnormity degree evaluation system and an intention deduction type triggering system. The automatic control method comprises the following steps: when a semantic block is deformed due to a physical space layout change, automatically correcting a rule action range to ensure that an instruction is always anchored on a dynamically changing semantic region to form an executable instruction set; inputting the behavior state chain output in real time and the corresponding comprehensive anomaly score into a prediction model; the high-score abnormal event temporarily improves the prediction weight of the related behavior mode; the intention deduction type triggering system recalculates the future behavior path and the target area of the target human body, the deduction direction is obviously changed by the abnormal data, and a prediction intention of abnormal perception is generated; and combining the abnormal perception prediction intention with the current environment state, and compiling into an equipment operation pre-sequence. The method completely depends on a data-driven adaptive analysis framework.

Description

Technical Field

[0001] This invention relates to the field of motion capture technology, and in particular to an AI human motion capture sensor and its automated control method. Background Technology

[0002] With the rapid development of the Internet of Things (IoT) and Artificial Intelligence (AI) technologies, human body sensors are increasingly being used in smart homes, security monitoring, health monitoring, and smart buildings. Traditional human body sensors mainly rely on infrared (PIR) or ultrasonic detection technologies, which can only determine the presence of a human body but cannot accurately identify a person's identity, posture, or physiological state. In recent years, AI-based human body sensor technology has gradually emerged, combining computer vision, deep learning, and multimodal perception technologies to achieve higher-precision human body detection, posture estimation, and behavior analysis. However, existing AI human motion capture sensors still suffer from high computational resource consumption, insufficient privacy protection, and poor environmental adaptability, limiting their application in scenarios with high real-time requirements and sensitive privacy.

[0003] Existing technology one, disclosed in publication number CN111444827B, presents a method for intelligent dynamic behavior analysis. Its operation process is as follows: The user places their ID card within the ID card reader of the behavior analysis device. After real-name authentication with their ID card, the user enters the system login interface and clicks the "Start Analysis" button. The user wears a motion capture sensor beforehand and then stands directly in front of the face recognition camera and electronic behavior sensor according to the prompts from the behavior analysis device. While this method can collect multiple sets of user motion image information, facial expression information, and planned behavior feature information, and simultaneously perform multi-dimensional analysis and comparison of the collected information on expressions, actions, and behavioral features, thereby directly enhancing the accuracy of the behavior analysis method, it relies on the user's active cooperation: requiring the user to wear specific sensors and complete identity authentication, it cannot achieve non-contact, real-time monitoring; it has limitations in scenarios: it is only applicable to preset fixed detection areas and cannot dynamically adapt to complex spatial layout changes; and it lacks abnormal behavior prediction: it only performs post-event analysis of actions that have already occurred and cannot extrapolate future behavioral intentions through spatiotemporal trajectories.

[0004] Prior art 2, publication number CN118486087A, discloses an AI-based behavior recognition analysis system, comprising: a data collection module equipped with a set of multimodal sensors for collecting user behavior data; a data preprocessing module, the output of which is electrically connected to the input of which the data collection module is used to format and standardize the collected multimodal data; a feature extraction module, the input of which is connected to the data preprocessing module for extracting behavior-related features from the preprocessed data; and a behavior classification module, the output of which is electrically connected to the behavior classification module. Although the system can efficiently collect data under various environmental conditions by introducing infrared and motion capture sensors, providing richer contextual information and thus enhancing the system's recognition capabilities, it suffers from several shortcomings. Firstly, multimodal data fusion is insufficient: despite employing multimodal sensors, the spatiotemporal alignment and field distortion coupling issues of heterogeneous sensor data remain unresolved. Secondly, static semantic annotation relies on preset environmental partitioning rules, failing to dynamically adjust semantic blocks based on real-time topology changes. Thirdly, anomaly detection is limited to a single dimension: judgment is based solely on behavioral classification thresholds, without incorporating physical field features such as state transition probability and field conduction loss rate to quantify anomaly levels.

[0005] Current technologies, specifically technology two, suffer from limitations in their methods of non-invasive dynamic monitoring and require further improvement in the accuracy of early prediction of abnormal behavior. Therefore, this invention provides an AI-powered human motion capture sensor and its automated control method. Summary of the Invention

[0006] To address the aforementioned technical problems, this invention provides an AI human motion capture sensor, comprising:

[0007] The anomaly assessment system tracks the movement path and dwelling nodes of a human body within a dynamically labeled topology network, forming a coordinate-independent spatiotemporal trajectory sequence. Based on characteristics such as trajectory velocity, direction change rate, and regional dwelling patterns, it automatically segments and labels these segments as coherent behavioral primitives, stringing them together into a time-ordered behavioral state chain. The system calculates a comprehensive anomaly score in real time using dimensions such as state transition probability deviation, duration deviation, and regional taboo triggering. The comprehensive anomaly score, behavioral state chain, and corresponding spatiotemporal trajectory sequence are then encoded and multiplexed, and transmitted to a remote monitoring center via wired or wireless communication interfaces.

[0008] Optional, anomaly assessment system, including:

[0009] The topological trajectory condensation subsystem is used to record the movement path between nodes and mark the stationary nodes by continuously tracking the propagation direction and intensity attenuation ratio of field distortion caused by the human body between topological nodes based on a dynamically labeled topological map; the movement path and the stationary nodes together constitute a spatiotemporal trajectory sequence.

[0010] The behavioral primitive field decomposition subsystem is used to automatically cut the spatiotemporal trajectory sequence into behavioral primitive segments when the rate of change of the propagation delay of the spatiotemporal trajectory sequence exceeds the adaptive threshold, the field strength gradient direction angle is continuously reversed, the propagation intensity attenuation ratio is abruptly triggered to cut the path, or the cumulative amount of field distortion reaches the saturation threshold. Each behavioral primitive segment carries its internal field propagation uniformity characteristics and endpoint field distortion intensity value.

[0011] The state chain field coupling subsystem is used to trigger semantic block matching by the field distortion intensity of the endpoints of the behavioral primitive segments: the behavioral primitive segments are connected in series through field transmission continuity verification to form a behavioral state chain weighted by block semantics. The state transition intensity is determined by the field distortion transmission loss rate of adjacent behavioral primitive segments.

[0012] The abnormal field distortion assessment subsystem is used for three-dimensional anomaly detection driven by the field transmission characteristics of the behavioral state chain. It generates a spatiotemporally fused anomaly score by integrating three field distortion indicators: state transition probability deviation, duration deviation, and regional taboo triggering. The score weight is adaptively adjusted by the dynamic boundary deformation amplitude of the semantic block.

[0013] Optionally, the state transition probability deviation in the abnormal field distortion assessment subsystem is: the probability density deviation between the actual conduction loss rate and the historical field conduction mode between adjacent behavioral primitive segments; duration deviation is: the statistical deviation between the field conduction uniformity within a single behavioral primitive segment and the historical uniformity distribution of the corresponding semantic block; regional taboo triggering is: when a taboo semantic block suddenly appears at the endpoint of a behavioral primitive, a field distortion conduction backtracking verification based on topological edge weights is triggered.

[0014] Optional, an anomalous field distortion assessment subsystem, including:

[0015] The conduction loss field interference component is used to generate conduction path interference patterns. When the actual loss rate causes distortion of the historical interference pattern on a specific topology path, the field strength coherence attenuation of the specific topology path is calculated as a physical representation of the state transition probability deviation.

[0016] A uniform field resonant component is used to output the field resonant energy dissipation rate as a physical measure of duration deviance when the uniform distribution of the behavioral primitive segment detunes from the cavity characteristic frequency by matching the inherent resonant frequency spectrum of the historical uniform field resonant cavity.

[0017] The taboo backtracking field verification component is used to drive the topological edge weight network to start reverse field propagation triggered by the behavioral primitive endpoints of taboo semantic blocks; backtracking from the endpoints along the ingress propagation path, the propagation trajectory is reconstructed through the edge weight decay gradient; when the reconstructed trajectory conflicts with the field distortion phase of the taboo block dynamic boundary, the length of the illegal propagation path is generated.

[0018] The distortion level fusion component is used to convert the field strength coherence attenuation into topological path interference level, the field resonance energy dissipation rate into resonant cavity dissipation level, and the illegal conduction path length into phase conflict level. The phase conflict level generates a spatiotemporal fusion anomalous field through the field distortion superposition principle, and its field strength peak value is the comprehensive anomalousness score.

[0019] Optional, a distortion level fusion component, comprising:

[0020] The boundary deformation modulator component is used to convert the field strength coherence attenuation into a topological path interference energy level through nonlinear scaling of the deformation amplitude; the field resonance energy dissipation rate is modulated by the deformation amplitude frequency and converted into a resonant cavity dissipation energy level; the illegal conduction path length is converted into a phase conflict energy level by deformation amplitude curvature renormalization.

[0021] The spatiotemporal coupling sub-component of energy levels is used to constrain the topological path interference energy levels in the spatial dimension by the topological edge weights of the original conduction path; the resonant cavity escaping energy levels are bound to the oscillation period in the temporal dimension by the duration of the behavioral primitive segment; the phase conflict energy levels are distributed along the topological nodes of the taboo backtracking trajectory; energy level gradient diffusion is performed; and energy level carriers with separated spatiotemporal properties are formed.

[0022] The distortion field superposition exciton component is used to generate a spatial interference ripple field by superimposing field distortions on the energy level carrier after spatiotemporal coupling. The topological path interference energy level excites the spatial interference ripple field, the resonant cavity escape energy level generates a time-decayed oscillation field, and the phase conflict energy level forms a forbidden gradient vortex field, which generates nonlinear interference in the four-dimensional field coupler.

[0023] The anomalous field strength convergence sub-component is used to propagate the nonlinear interference field within the dynamic boundary of the semantic block. Guided by boundary deformation, it forms an energy flow convergence and is filtered by the historical anomalous field strength baseline. Finally, the field strength peak is generated at the future extension point of the spatiotemporal trajectory, which is the comprehensive anomaly score.

[0024] Optional, anomaly field strength convergence sub-component, including:

[0025] The boundary guidance field propagation module is used to propagate nonlinear interference fields within the dynamic boundary of semantic blocks. The real-time deformation amplitude of the boundary generates a curvature guidance effect, which redirects the energy flow of the interference field along the curvature gradient of the dynamic boundary, forming an energy flow beam that converges towards future spatiotemporal coordinates.

[0026] The historical baseline field filtering module is used to converge the energy flow beam through the historical abnormal field strength baseline stored in the historical field conduction mode database. The normal conduction mode in the historical field conduction mode database forms the field transmission impedance distribution. The component in the energy flow beam that matches the impedance distribution is absorbed and attenuated, and the remaining abnormal energy flow density distribution passes through the baseline.

[0027] The trajectory phase delay compensation module is used to input the future extension points of the spatiotemporal trajectory of the abnormal energy flux density distribution. The topological path length of the future extension points is converted into the field propagation delay. The energy flux density distribution performs phase advance compensation based on the delay to generate a pre-synchronized energy flux hypersurface.

[0028] The field strength peak collapse module is used to cause energy level collapse of the pre-synchronized energy flow hypersurface at the target spatiotemporal coordinate point. It is modulated by the dynamic boundary deformation rate of the semantic block, and the collapse time window is determined by the duration of the behavior primitive segment. Finally, the field strength scalar value of the collapse point is output, which is the comprehensive anomaly score.

[0029] Optionally, data transmission adopts time-division multiplexing, allocating the anomaly score, behavioral state chain, and spatiotemporal trajectory sequence to different time slots for transmission.

[0030] Optionally, it also includes a dynamic region labeling system, which continuously captures multi-dimensional field changes caused by human activity within the covered space through a sensor array; inputs the multi-dimensional field changes into an adaptive spatial modeling engine, and automatically constructs a three-dimensional topological relationship map describing the relative positions and connectivity between spatial points in the covered area through continuous field analysis without a preset grid; and dynamically labels semantic blocks in the topological map based on user-defined semantic rules and the output of the real-time three-dimensional topological relationship map.

[0031] Optionally, it also includes an intent-based triggering system, which integrates behavioral state chains, anomaly scores, and dynamic region semantics to predict the most likely behavioral path and target area of ​​the target human body in the next few seconds through probability graphs.

[0032] The present invention provides an automated control method for an AI human motion capture sensor, comprising the following steps:

[0033] Users set regional monitoring conditions and expected actions through an intuitive interface to form an original instruction set; the user-preset rules are matched with the semantic block information output in real time by the dynamic regional identification system; when changes in physical space layout cause semantic block deformation, the scope of the rules is automatically corrected to ensure that the instructions are always anchored on the dynamically changing semantic region, thus forming an executable instruction set.

[0034] The behavioral state chain output in real time by the anomaly assessment system and the corresponding comprehensive anomaly score are input into the prediction model; high-score abnormal events will temporarily increase the prediction weight of related behavioral patterns; based on the anomaly fusion result of the state chain, the intention inference trigger system recalculates the future behavioral path and target area of ​​the target human body, and the abnormal data significantly changes the inference direction, generating the prediction intention of anomaly perception.

[0035] The predicted intent of anomaly perception is combined with the current environmental state and compiled into a device operation pre-sequence. Before the operation pre-sequence takes effect, the consistency between the human state in the latest semantic block output by the dynamic region identification system and the predicted intent is continuously compared. When the verification is successful, the pre-sequence is released. When the deviation exceeds the threshold, the sequence is discarded and re-prediction is triggered, and the final execution instruction is output.

[0036] This invention achieves three-dimensional spatial semantic understanding and intent prediction of human activities through the synergy of multimodal sensing and intelligent algorithms. Specifically, it demonstrates its technological integration value at three levels: First, at the spatial dynamic modeling level, it breaks through the limitations of traditional grid-based spatial segmentation by using non-contact multi-dimensional field perception (energy / pressure / electromagnetism) to establish an adaptive three-dimensional topological relationship graph. This dynamic modeling method without a preset coordinate system can not only reflect the changes in spatial relationships caused by human activities in real time, but also transform the physical space into a computable semantic network through semantic rule injection, providing a structured spatial reference system for subsequent analysis. Second, at the behavioral quantitative assessment level, the behavioral trajectory analysis based on the topological network eliminates the dependence on absolute coordinates. Through feature extraction, continuous motion is discretized into computable behavioral primitives. This anomaly detection mechanism based on state chains achieves irregular quantitative assessment of behavioral patterns through multi-dimensional deviation calculation (probability bias / temporal anomaly / regional taboo). At the predictive decision-making level, the system integrates the triple inputs of spatiotemporal behavior chain, dynamic anomaly degree and regional semantics, and uses a probabilistic graphical model to deduce the Markov process of behavior path. This deduction mechanism predicts the target area and path selection in advance by calculating the conditional probability distribution of state transition, forming a complete reasoning chain from physical signal acquisition to behavioral intention output.

[0037] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings.

[0038] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0039] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0040] Figure 1 This is a block diagram of the AI ​​human motion capture sensor in Embodiment 1 of the present invention;

[0041] Figure 2 This is a schematic diagram of the AI ​​human motion capture sensor in Embodiment 1 of the present invention;

[0042] Figure 3 This is a block diagram of the dynamic area identification system in Embodiment 2 of the present invention;

[0043] Figure 4 This is a block diagram of the anomaly assessment system in Embodiment 3 of the present invention;

[0044] Figure 5 This is a block diagram of the intent-based triggering system in Embodiment 10 of the present invention;

[0045] Figure 6 This is a flowchart of the automated control method for the AI ​​human motion capture sensor in Embodiment 11 of the present invention. Detailed Implementation

[0046] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0047] The terminology used in the embodiments of this application is for the purpose of describing particular embodiments only and is not intended to limit the embodiments of this application. The singular forms “a,” “the,” and “the” used in the embodiments of this application are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any or all possible combinations of one or more associated listed items.

[0048] In the following description, when referring to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application. In the description of this application, it should be understood that the terms "first," "second," "third," etc., are used only to distinguish similar objects and are not necessarily used to describe a specific order or sequence, nor should they be construed as indicating or implying relative importance. Those skilled in the art can understand the specific meaning of the above terms in this application according to the specific circumstances.

[0049] Example 1: As Figure 1 As shown, this embodiment of the invention provides an AI human motion capture sensor, comprising:

[0050] The dynamic area identification system uses an intelligent sensor array composed of various heterogeneous sensors (pyroelectric, MEMS barometer, electromagnetic induction coil, seismic sensor) to continuously capture multi-dimensional field changes (energy disturbances, micro-pressure fluctuations, electromagnetic field distortions, etc.) caused by human activities within the covered space. These multi-dimensional field changes are then input into an adaptive spatial modeling engine, which automatically constructs a 3D topological map describing the relative positions and connectivity between spatial points in the covered area through continuous field analysis without a preset grid. Based on user-defined semantic rules (such as "rest area" and "restricted area") and the output of the real-time 3D topological map, semantic blocks in the topological map are dynamically labeled.

[0051] An anomaly assessment system tracks the movement path and dwelling nodes of a human body within a dynamically labeled topology network, forming a coordinate-independent spatiotemporal trajectory sequence. Based on characteristics such as trajectory velocity, direction abrupt change rate, and regional dwelling patterns, it automatically segments and labels these segments as coherent behavioral primitives, chaining them into a time-ordered behavioral state chain. A comprehensive anomaly score is calculated in real-time using dimensions such as state transition probability deviation, duration deviation, and regional taboo triggering. The comprehensive anomaly score, behavioral state chain, and corresponding spatiotemporal trajectory sequence are encoded and multiplexed, then transmitted to a remote monitoring center via wired or wireless communication interfaces. Data transmission employs time-division multiplexing, allocating the anomaly score, behavioral state chain, and spatiotemporal trajectory sequence to different time slots for transmission. The intent-driven triggering system is defined as the remote monitoring center.

[0052] The intent-based triggering system is used to comprehensively analyze behavioral state chains, anomaly scores, and dynamic region semantics to predict the most likely behavioral path and target area of ​​a human body in the next few seconds using a probability graph.

[0053] The working principle and beneficial effects of the above technical solution are as follows: The dynamic area identification system of this embodiment is used to continuously capture multi-dimensional field changes (energy disturbances, micro-pressure fluctuations, electromagnetic field distortions, etc.) caused by human activities within the covered space through a sensor array; inputting the multi-dimensional field changes into an adaptive spatial modeling engine, and automatically constructing a three-dimensional topological relationship map of the relative positions and connectivity between spatial points in the covered area through continuous field analysis without a preset grid; dynamically labeling semantic blocks in the topological map based on user-defined semantic rules (such as "rest area", "restricted area") and the output of the real-time three-dimensional topological relationship map; the anomaly evaluation system is used for Based on a dynamically labeled topology graph, the system tracks the movement path and dwelling nodes of a human body within the topological network, forming a coordinate-independent spatiotemporal trajectory sequence. Based on characteristics such as trajectory velocity, direction abrupt change rate, and regional dwelling patterns, the system automatically segments and labels these segments as coherent behavioral primitives, stringing them together into a time-ordered behavioral state chain. A comprehensive anomaly score is calculated in real-time using dimensions such as state transition probability deviation, duration deviation, and regional taboo triggering. An intent-based triggering system is used to integrate the behavioral state chain, comprehensive anomaly score, and dynamic regional semantics, using a probabilistic graph to predict the most likely behavioral path and target region of the target human body within the next few seconds (see appendix for details). Figure 2 The above-mentioned solution, through the collaboration of multimodal sensing and intelligent algorithms, achieves three-dimensional spatial semantic understanding and intention prediction of human activities. Specifically, it demonstrates its technological integration value at three levels: First, at the spatial dynamic modeling level, it breaks through the limitations of traditional gridded spatial segmentation through non-contact multi-dimensional field perception (energy / pressure / electromagnetism), establishing an adaptive three-dimensional topological relationship graph. This dynamic modeling method without a preset coordinate system can not only reflect changes in spatial relationships caused by human activities in real time, but also transform physical space into a computable semantic network through semantic rule injection, providing a structured spatial reference system for subsequent analysis. Second, at the behavioral quantitative assessment level, behavior trajectory analysis based on topological networks eliminates absolute coordinate dependence. Through feature extraction, continuous motion is discretized into computable behavioral primitives. This anomaly detection mechanism based on state chains achieves irregular quantitative assessment of behavioral patterns through multi-dimensional deviation calculation (probability bias / temporal anomaly / regional taboo). At the predictive decision-making level, the system integrates the triple inputs of spatiotemporal behavior chain, dynamic anomaly degree and regional semantics, and uses a probabilistic graphical model to deduce the Markov process of behavior path. This deduction mechanism predicts the target area and path selection in advance by calculating the conditional probability distribution of state transition, forming a complete reasoning chain from physical signal acquisition to behavioral intention output.

[0054] In summary, this embodiment realizes physical field changes, 3D topology construction, semantic space annotation, behavioral primitive parsing, anomaly quantification, and intent probability inference. The entire process does not require a pre-set behavioral rule base and relies entirely on a data-driven adaptive analysis framework.

[0055] Example 2: As Figure 3 As shown, based on Embodiment 1, the dynamic region identification system provided in this embodiment of the invention includes:

[0056] The multidimensional field dynamic coupling subsystem is used to realize that the multidimensional field changes caused by human activities, such as energy disturbances, micro-pressure fluctuations and electromagnetic field distortions, are captured by the sensor array as spatiotemporally continuous heterogeneous manifolds, which are directly input into the adaptive spatial modeling engine to trigger the non-discrete field analysis process.

[0057] The continuous field topological condensation subsystem is used to perform field strength self-organization in a spatial framework without a pre-set grid by leveraging the inherent gradient characteristics and boundary effects of heterogeneous manifolds. The evolution process of field strength self-organization automatically condenses spatial positional correlations, generating a three-dimensional connected skeleton that describes the relative distances and path reachability between points. The three-dimensional connected skeleton is a dynamic topological relationship graph covering the spatial region, whose nodes are naturally formed by field strength extrema and whose edges are dynamically connected by field strength attenuation paths.

[0058] The semantic rule field mapping subsystem is used to compile user-preset semantic rules into topological constraints; changes in the node connectivity density and edge weights of the real-time 3D topological graph trigger the constraint matching engine, marking connected subgraphs that meet specific motion pattern thresholds as prototypes of semantic blocks.

[0059] The block boundary adaptive subsystem is used to regulate the spatial extensibility of semantic block prototypes by the field distortion conductance between adjacent nodes in the topology graph. When human movement causes a sudden change in conductance, the constraint matching engine automatically shrinks or expands the block boundary through topology edge reconstruction, and finally outputs a dynamic labeled topology graph of semantic blocks whose boundaries deform in real time with field distortion.

[0060] The working principle and beneficial effects of the above technical solution are as follows: The multi-dimensional field dynamic coupling subsystem of this embodiment is used to realize that the multi-dimensional field changes caused by human activities, such as energy disturbances, micro-pressure fluctuations, and electromagnetic field distortions, are captured by the sensor array as spatiotemporally continuous heterogeneous manifolds within the covered space. These are directly input into the adaptive spatial modeling engine to trigger a non-discrete field analysis process. The continuous field topology condensation subsystem is used to perform field strength self-organization in a spatial framework without a preset grid by utilizing the gradient characteristics and boundary effects inherent in the heterogeneous manifold. The evolution process of field strength self-organization automatically condenses spatial position correlations, generating a three-dimensional connected skeleton describing the relative distance and path reachability between points. The three-dimensional connected skeleton represents the dynamic topological relationship of the covered spatial region. The graph, whose nodes are naturally formed by extreme points of field strength and whose edges are dynamically connected by field strength attenuation paths, is described in two parts. A semantic rule field mapping subsystem is used to compile user-preset semantic rules into topological constraints. Changes in the node connectivity density and edge weights of the real-time 3D topological graph trigger a constraint matching engine, marking connected subgraphs that meet specific motion mode thresholds as prototypes of semantic blocks. A block boundary adaptive subsystem is used to control the spatial extensibility of the semantic block prototypes based on the field distortion conductance between adjacent nodes in the topological graph. When human movement causes a sudden change in conductance, the constraint matching engine automatically shrinks or expands the block boundaries through topological edge reconstruction, ultimately outputting a dynamically labeled topological graph of semantic blocks whose boundaries deform in real time with field distortion. This dynamic region identification system constructs a spatial semantic parsing system with autonomous evolution capabilities through real-time capture and calculation of multi-dimensional physical field changes. The multi-dimensional field dynamic coupling subsystem transforms the complex physical disturbances (energy / pressure / electromagnetic field) caused by human activity into a spatiotemporally continuous heterogeneous manifold data stream, providing the system with the original field strength gradient distribution; this non-contact sensing method achieves a holographic mathematical description of three-dimensional space. The continuous field topological condensation subsystem generates a dynamic topological graph without a preset reference frame, leveraging the self-organizing properties of manifold data. The three-dimensional connected skeleton of this graph has two core features: the node distribution reflects the natural emergence of field strength extrema, and the edge connections embody the dynamic transmission characteristics of physical field attenuation paths. The semantic rule field mapping subsystem transforms user-defined semantic rules into topological constraints, achieving the transformation from physical field changes to semantic concepts by real-time matching of motion pattern features in connected subgraphs; this mapping mechanism maintains the mathematical isomorphism between the physical and semantic layers. The block boundary adaptive subsystem enables real-time deformation of the geometric boundaries of semantic blocks through dynamic calculation of field distortion conductance; when human movement alters the local field strength distribution, the system adjusts the block boundaries through topological edge reconstruction, ensuring synchronous updates of semantic annotations and physical space changes.

[0061] In summary, this embodiment establishes a dynamic annotation system that does not require pre-divided grids and can autonomously adapt to changes in spatial state caused by human activities. It is particularly suitable for intelligent spatial scenarios that require real-time reflection of the interaction between humans and the environment.

[0062] Example 3: As Figure 4 As shown, based on Example 1, the anomaly assessment system provided in this embodiment of the invention includes:

[0063] The topology trajectory aggregation subsystem is used to continuously track the transmission direction and intensity attenuation ratio of field distortion caused by the human body between topology nodes based on a dynamically labeled topology map. When the transmission intensity attenuation ratio is continuously higher than a preset motion threshold, it is recorded as a movement path between nodes, and the path weight is dynamically updated by the transmission delay. When the transmission intensity accumulates in a single node and the attenuation ratio is lower than the dwell threshold, it is marked as a dwell node, and its duration is determined by the field distortion stability. The movement path and the dwell node together constitute a spatiotemporal trajectory sequence, where the node correlation is locked by both the transmission direction and the intensity attenuation ratio.

[0064] The behavioral primitive field decomposition subsystem is used to reflect the fluctuation of trajectory velocity by the rate of change of propagation delay in the spatiotemporal trajectory sequence. The direction of change of field strength gradient between topological nodes determines the motion turning characteristics of the direction change rate. The sudden change of the propagation intensity attenuation ratio between adjacent nodes of the moving path triggers path cutting. The accumulation of field distortion at the stationary node triggers the termination of stationing when it reaches the saturation threshold. When the rate of change of propagation delay of the spatiotemporal trajectory sequence exceeds the adaptive threshold, the direction angle of the field strength gradient is continuously reversed, the sudden change of the propagation intensity attenuation ratio triggers path cutting, or the accumulation of field distortion reaches the saturation threshold, the spatiotemporal trajectory sequence is automatically cut into behavioral primitive segments. Each behavioral primitive segment carries its internal field propagation uniformity characteristics and the field distortion intensity value at the endpoint.

[0065] Among these, "resident nodes" and "resident patterns" are two concepts at different levels. The former forms the basis and detection criteria for the latter. A resident node is an instantaneous, microscopic spatiotemporal unit; it represents the point at which, by analyzing field distortion signals, a person stops moving at a specific location or node in the topological network. The criterion is that the transmission intensity accumulates at a single node and the attenuation ratio is lower than the resident threshold, and the duration is determined by the stability of the field distortion. A resident pattern is a continuous, macroscopic behavioral characteristic, characterized by a series of temporally continuous and spatially similar clusters of resident nodes, supplemented by higher-level features such as duration, frequency, and periodicity. For example: Long-term static resident pattern: In the rest area semantic block, the appearance of a cluster of resident nodes with a long duration may be interpreted as sleeping or sitting still. High-frequency short-term resident pattern: In the kitchen area, the appearance of multiple short-lived, alternating clusters of resident nodes may be interpreted as cooking or preparing food. Periodic movement-resident pattern: On a corridor path, the appearance of a regular sequence of alternating movement paths and short-lived resident nodes may be interpreted as loitering. The dwelling pattern is a statistical learning and behavioral abstraction of dwelling nodes in the spatiotemporal dimension; by tracking and recording the occurrence patterns of dwelling nodes in historical data, a normal dwelling pattern archive is established for different semantic blocks; when a new real-time dwelling node sequence deviates significantly from these historical patterns, i.e., the duration deviation is high, anomaly evaluation is triggered.

[0066] The triggering logic for automatic segmentation of the behavioral primitive field decomposition subsystem is as follows:

[0067] The rate of change of transmission delay is the instantaneous rate of change of the time required for a signal to travel from one sensor node to an adjacent node; it directly reflects the fluctuation of human movement speed; the value is stable during uniform motion; the value increases during acceleration or deceleration; when the rate of change of transmission delay exceeds an adaptive threshold, it dynamically adjusts based on historical average speed, indicating that the movement speed has changed drastically, such as a sudden start or stop; at this moment, the preceding and following trajectories are divided into two different behavioral primitive segments, separating the slow walking segment from the fast running segment;

[0068] The field strength gradient direction angle continuously reverses. The field strength gradient direction indicates the direction of the strongest signal source, i.e., the direction of human movement. The continuous reversal of the direction angle means that the angle has undergone multiple violent fluctuations of positive and negative values ​​in a short period of time. This corresponds to behaviors such as turning in place, hesitation, or lingering in a small area. When multiple reversals of the direction angle are detected in a very short period of time, it is determined that this is an independent behavioral element with turning characteristics, and it is cut out from the stable straight trajectory.

[0069] A sudden change in the transmission intensity attenuation ratio indicates a sudden change in the relative geometric relationship between the human body and the sensor node. When a human body moves rapidly from one topological node to another non-adjacent node, such as flashing from one room doorway to another, skipping corridor nodes in between, the signal transmission path will suddenly change, causing a discontinuous jump in the attenuation ratio. This jump triggers path cutting, marking the beginning and end of a flashing or long-distance movement behavior primitive.

[0070] When the accumulated field distortion reaches the saturation threshold, the signal strength at the dwell node will accumulate and tend to a stable value. When the dwell time is determined to end, the accumulated field distortion starts to decrease from the stable value, that is, the attenuation ratio is higher than the motion threshold again, which means that the human body begins to leave the current dwell point. The moment when this dwell time ends marks the end of a dwell behavior primitive and the beginning of the next movement behavior primitive.

[0071] The state chain field coupling subsystem is used to trigger semantic block matching based on the field distortion intensity of the endpoints of behavioral primitives: if the endpoint is located in a high-activity semantic block (passage area), the connection is strengthened; if it is located in a low-activity block (rest area), the connection is weakened. Behavioral primitives are connected in series through field transmission continuity verification to form a behavioral state chain weighted by block semantics. The state transition strength is determined by the field distortion transmission loss rate of adjacent behavioral primitives.

[0072] An abnormal field distortion assessment subsystem is used for three-dimensional anomaly detection driven by the field transmission characteristics of the behavioral state chain: it generates a spatiotemporally fused anomaly score by comprehensively considering three field distortion indicators: state transition probability deviation, duration deviation, and regional taboo triggering. The score weight is adaptively adjusted by the dynamic boundary deformation amplitude of the semantic block.

[0073] State transition probability deviation: the deviation between the actual conduction loss rate and the probability density of the historical field conduction mode between adjacent behavioral primitive segments;

[0074] Duration deviation: The statistical deviation between the uniformity of field propagation within a single behavioral primitive segment and the historical uniformity distribution of the corresponding semantic block;

[0075] Tabu Trigger: When a taboo semantic block suddenly appears at the endpoint of a behavioral primitive, a backtracking verification based on the field distortion propagation of topological edge weights is triggered.

[0076] The working principle and beneficial effects of the above technical solution are as follows: The topology trajectory condensation subsystem of this embodiment is used to continuously track the transmission direction and intensity attenuation ratio of field distortion caused by the human body between topology nodes based on a dynamically labeled topology map. When the transmission intensity attenuation ratio is continuously higher than a preset motion threshold, it is recorded as a movement path between nodes, and the path weight is dynamically updated by the transmission delay. When the transmission intensity accumulates in a single node and the attenuation ratio is lower than the dwell threshold, it is marked as a dwell node, and its duration is determined by the field distortion stability. The movement path and the dwell node together constitute a spatiotemporal trajectory sequence, wherein the node correlation is determined by the transmission direction and intensity attenuation ratio. The intensity attenuation ratio is dual-locked; the behavioral primitive field decomposition subsystem uses the propagation delay change rate of the spatiotemporal trajectory sequence to reflect the fluctuation of trajectory velocity, and the direction of field intensity gradient change between topological nodes determines the motion turning characteristics of the direction change rate; the abrupt change in the propagation intensity attenuation ratio between adjacent nodes of the moving path triggers path cutting, and the accumulated field distortion of the stationary node triggers stationing termination when it reaches the saturation threshold; when the propagation delay change rate of the spatiotemporal trajectory sequence exceeds the adaptive threshold, the field intensity gradient direction angle continuously reverses, the abrupt change in the propagation intensity attenuation ratio triggers path cutting, or the accumulated field distortion reaches the saturation threshold, the spatiotemporal trajectory sequence is automatically cut. Each behavioral primitive segment carries its internal field conduction uniformity characteristics and endpoint field distortion intensity values. The state chain field coupling subsystem uses the endpoint field distortion intensity of the behavioral primitive segment to trigger semantic block matching: if the endpoint is located in a high-activity semantic block (passage area), the connection is strengthened; if it is located in a low-activity block (rest area), the connection is weakened. Behavioral primitive segments are linked together through field conduction continuity verification to form a behavioral state chain weighted by block semantics. The state transition strength is determined by the field distortion conduction loss rate of adjacent behavioral primitive segments. The abnormal field distortion evaluation subsystem uses the field conduction characteristics of the behavioral state chain to drive three... Anomaly Detection: A spatiotemporally fused anomaly score is generated by integrating three field distortion indicators: state transition probability deviation, duration deviation, and regional taboo triggering. The score weights are adaptively adjusted by the dynamic boundary deformation amplitude of the semantic block. State transition probability deviation: The probability density deviation between the actual transmission loss rate and the historical field transmission pattern between adjacent behavioral primitive segments. Duration deviation: The statistical deviation between the field transmission uniformity within a single behavioral primitive segment and the historical uniformity distribution of the corresponding semantic block. Regional taboo triggering: When a taboo semantic block suddenly appears at the endpoint of a behavioral primitive, a backtracking verification of field distortion transmission based on topological edge weights is triggered. The above scheme achieves accurate quantitative representation of dynamic motion trajectories through a topological trajectory condensation subsystem, decomposing continuous motion into a spatiotemporal sequence of movement paths and stationary nodes. Path weights are dynamically updated through transmission delay, and node correlation is locked by both transmission direction and intensity attenuation ratio, forming a trajectory model with physical field transmission characteristics.The behavioral primitive field decomposition subsystem automatically segments spatiotemporal trajectories based on field propagation parameters such as the rate of change of propagation delay, the direction angle of the field strength gradient, and the abrupt change in attenuation ratio, extracting behavioral primitive segments with internal field propagation uniformity characteristics. Cutting is triggered by four types of physical thresholds—adaptive threshold, direction angle reversal threshold, path cutting threshold, and saturation threshold—ensuring the consistency of physical field characteristics of the behavioral primitives. The state chain field coupling subsystem establishes a mapping relationship between behavioral primitives and semantic blocks, achieving block matching through endpoint field distortion intensity, and constructing a semantically weighted behavioral state chain based on field propagation continuity verification. The state transition intensity is determined by the field distortion propagation loss rate of adjacent primitive segments, forming a semantically constrained behavioral chain. The anomalous field distortion assessment subsystem achieves multi-scale anomaly detection through three-dimensional indicators: temporal anomaly detection: state transition probability deviation quantifies the deviation of the transmission loss rate from historical patterns; persistent behavioral anomaly: duration deviation measures the statistical difference between field transmission uniformity and historical distribution; spatial violation detection: regional taboo triggering mechanism verifies the field distortion transmission path of taboo blocks; through adaptive weight adjustment of semantic block boundary deformation, the final output is a spatiotemporal fusion anomaly score.

[0077] In summary, this embodiment constructs a behavior analysis framework based on a physical field transmission model, organically coupling motion trajectories, behavioral primitives, and semantic blocks through field distortion transmission characteristics, achieving end-to-end computation from raw motion data to semantic anomaly assessment. The subsystems form a progressive processing chain: trajectory aggregation, behavior deconstruction, semantic coupling, and anomaly assessment, ultimately outputting anomaly detection results with physical field transmission interpretability.

[0078] Example 4: Based on Example 3, the abnormal field distortion assessment subsystem provided in this embodiment of the invention includes:

[0079] The conduction loss field interference component is used to spatiotemporally align the actual conduction loss rate between adjacent behavioral primitive segments with the historical field conduction mode database to generate a conduction path interference map. When the actual loss rate causes distortion of the historical interference map on a specific topological path, the field strength coherence attenuation of the specific topological path is calculated as a physical representation of the state transition probability deviation.

[0080] The uniform field resonance component is used to input the uniformity of field conduction within the behavioral primitive segment to the historical uniform field resonant cavity corresponding to the semantic block. By matching the inherent resonant frequency spectrum of the historical uniform field resonant cavity, when the uniformity distribution of the behavioral primitive segment is detuned to the characteristic frequency of the cavity, the output field resonance energy dissipation rate is used as a physical measure of the duration deviation.

[0081] The taboo backtracking field verification component is used to drive the topological edge weight network to start reverse field propagation triggered by the behavioral primitive endpoints of taboo semantic blocks; backtracking from the endpoints along the ingress propagation path, the propagation trajectory is reconstructed through the edge weight decay gradient; when the reconstructed trajectory conflicts with the field distortion phase of the taboo block dynamic boundary, the length of the illegal propagation path is generated.

[0082] The distortion level fusion component is used to modulate the physical dimensions of field strength coherence attenuation, field resonance energy dissipation rate, and illegal conduction path length using the semantic block boundary deformation amplitude. It converts field strength coherence attenuation into topological path interference level, field resonance energy dissipation rate into resonant cavity dissipation level, and illegal conduction path length into phase conflict level. The three phase conflict levels generate a spatiotemporal fusion anomalous field through the field distortion superposition principle, and the peak field strength is the comprehensive anomalousness score.

[0083] The working principle and beneficial effects of the above technical solution are as follows: The conduction loss field interference component in this embodiment is used to spatiotemporally align the actual conduction loss rate between adjacent behavioral primitive segments with the historical field conduction mode database to generate a conduction path interference map; when the actual loss rate causes distortion of the historical interference map on a specific topological path, the field strength coherence attenuation of the specific topological path is calculated as a physical representation of the state transition probability deviation; the homogeneous field resonance component is used to input the field conduction homogeneity within the behavioral primitive segment into the historical homogeneous field resonant cavity corresponding to the semantic block; by matching the inherent resonant frequency spectrum of the historical homogeneous field resonant cavity, when the homogeneous distribution of the behavioral primitive segment detunes from the cavity characteristic frequency, the field resonance energy dissipation rate is output as a physical measure of the duration deviation; taboo backtracking field verification The verification component is used to initiate reverse field propagation by the endpoint-driven topological edge weight network triggered by the taboo semantic block. It traces back along the ingress propagation path from the endpoint, reconstructing the propagation trajectory through the edge weight decay gradient. When the reconstructed trajectory conflicts with the dynamic boundary of the taboo block due to field distortion phase, an illegal propagation path length is generated. The distortion level fusion component modulates the physical dimensions of field strength coherence attenuation, field resonance energy dissipation rate, and illegal propagation path length using the deformation amplitude of the semantic block boundary. It converts field strength coherence attenuation into topological path interference levels, field resonance energy dissipation rate into resonant cavity dissipation levels, and illegal propagation path length into phase conflict levels. These three phase conflict levels generate a spatiotemporally fused anomalous field through the field distortion superposition principle, and the peak field strength is the comprehensive anomalousness score. The anomalous field distortion assessment subsystem of the above scheme achieves physical quantification and fusion analysis of anomalous phenomena during the propagation of behavioral primitives through the collaborative work of multiple components. The conduction loss field interference component establishes a spatiotemporal mapping relationship between historical field modes and real-time conduction, transforming loss anomalies on the topological path into calculable field strength coherence attenuation; revealing the physical essence of the state transition probability deviation of the conduction path. The homogeneous field resonance component, through a resonant cavity matching mechanism, quantifies conduction detuning phenomena within behavioral primitive segments into energy dissipation rate parameters; these parameters objectively reflect the duration deviation between the actual conduction process and the historical steady-state mode. The taboo backtracking field verification component constructs an inverse field conduction model, accurately capturing the phase conflict characteristics of illegal conduction paths and semantic boundaries through weighted attenuation gradient reconstruction technology, outputting physically meaningful path length values. The distortion level fusion component, through unified dimensional modulation, achieves field-theoretical normalization of three types of heterogeneous physical parameters—interference levels, dissipation levels, and conflict levels; based on the comprehensive anomaly score generated by the field distortion superposition principle, a unified evaluation system for multi-dimensional anomaly characteristics is established.

[0084] In summary, this embodiment transforms the three types of anomalous phenomena—topological interference, resonance detuning, and boundary conflict—in the process of behavioral primitive transmission into quantifiable and comparable energy level parameters, ultimately outputting a comprehensive anomaly assessment result with a clear physical interpretation. The data flow between each component strictly follows the physical laws of field transmission, forming a closed-loop anomaly detection-quantification-fusion chain.

[0085] Example 5: Based on Example 4, the distortion level fusion component provided in this embodiment of the invention includes:

[0086] The boundary deformation modulatory subcomponent is used to apply the dynamic boundary deformation amplitude of semantic blocks to three inputs: field strength coherence attenuation, field resonance energy dissipation rate, and illegal conduction path length.

[0087] The field strength coherence attenuation is transformed into a topological path interference energy level through nonlinear scaling of the deformation amplitude; the field resonance energy dissipation rate is transformed into a resonant cavity dissipation energy level by frequency modulation of the deformation amplitude; the length of the illegal conduction path is transformed into a phase conflict energy level by deformation amplitude curvature renormalization.

[0088] The spatiotemporal coupling sub-component of energy levels is used to constrain the topological path interference energy levels in the spatial dimension by the topological edge weights of the original conduction path; the resonant cavity escaping energy levels are bound to the oscillation period in the temporal dimension by the duration of the behavioral primitive segment; the phase conflict energy levels are distributed along the topological nodes of the taboo backtracking trajectory; energy level gradient diffusion is performed; and energy level carriers with separated spatiotemporal properties are formed.

[0089] The distortion field superposition exciton component is used to generate a spatial interference ripple field by superimposing field distortions on the energy level carrier after spatiotemporal coupling. The topological path interference energy level excites the spatial interference ripple field, the resonant cavity escape energy level generates a time-decayed oscillation field, and the phase conflict energy level forms a forbidden gradient vortex field, which generates nonlinear interference in the four-dimensional field coupler.

[0090] The anomalous field strength convergence sub-component is used to propagate the nonlinear interference field within the dynamic boundary of the semantic block. Guided by boundary deformation, it forms an energy flow convergence and is filtered by the historical anomalous field strength baseline. Finally, the field strength peak is generated at the future extension point of the spatiotemporal trajectory, which is the comprehensive anomaly score.

[0091] The working principle and beneficial effects of the above technical solution are as follows: In this embodiment, the boundary deformation modulator sub-component is used to apply the dynamic boundary deformation amplitude of the semantic block to three inputs: field strength coherence attenuation, field resonance energy dissipation rate, and illegal conduction path length. The field strength coherence attenuation is converted into a topological path interference level through nonlinear scaling of the deformation amplitude. The field resonance energy dissipation rate is converted into a resonant cavity dissipation level by frequency modulation of the deformation amplitude. The illegal conduction path length is converted into a phase conflict level by deformation amplitude curvature renormalization. The energy level spatiotemporal coupling sub-component is used to constrain the topological path interference level in the spatial dimension with the topological edge weight of the original conduction path. The resonant cavity dissipation level in the time dimension is constrained by the duration of the behavioral primitive segment. The process involves binding the oscillation period; distributing phase-conflicting energy levels along the topological nodes of the taboo backtracking trajectory; performing energy level gradient diffusion; forming a spatiotemporally separated energy level carrier; using a distortion field superposition excitation sub-component for the spatiotemporally coupled energy level carrier, the topological path interference energy level excites a spatial interference ripple field through field distortion superposition, the resonant cavity escape energy level generates a time-decaying oscillation field, and the phase-conflicting energy level forms a taboo gradient vortex field, generating nonlinear interference in a four-dimensional field coupler; and using an anomalous field strength convergence sub-component for the propagation of the nonlinear interference field within the dynamic boundary of the semantic block, guided by boundary deformation to form energy flow convergence, filtered by the historical anomalous field strength baseline; and finally generating a field strength peak at the future extension point of the spatiotemporal trajectory, which is the comprehensive anomalousness score. The boundary deformation dimension modulation sub-component of the above scheme transforms physical quantities into three types of characteristic energy levels through the triple action of semantic block boundary deformation: modulation of field strength coherence attenuation, regulation of field resonance energy dissipation rate, and renormalization of illegal conduction path length. These transformations are: topological path interference energy level or spatial dimension, resonant cavity dissipation energy level or temporal dimension, and phase conflict energy level or taboo dimension. The energy level spatiotemporal coupling sub-component imposes dimensional constraints on the three types of energy levels: the topological path interference energy level is limited by the topological edge weight of the original conduction path; the resonant cavity dissipation energy level oscillates synchronously with the duration of the behavioral primitive segment; and the phase conflict energy level is distributed along the taboo backtracking trajectory nodes, forming a discretized energy level carrier with clear spatiotemporal attributes. The distortion field superposition excitation sub-component realizes the field-domain transformation of the three-dimensional energy level carrier: spatial interference ripple field or topological path interference energy level excitation, time-decayed oscillation field or resonant cavity dissipation energy level generation, and taboo gradient vortex field or phase conflict energy level formation; nonlinear interference in the four-dimensional field coupler generates a composite field effect. The anomalous field strength convergence sub-component guides the energy flow through dynamic boundaries and combines it with historical field strength baseline filtering to make the nonlinear interference field form a field strength peak at the future extension point of the spatiotemporal trajectory; this peak is the comprehensive anomalous quantification index output by the system.

[0092] In summary, this embodiment uses dynamic boundary deformation to drive the transformation of multi-dimensional physical quantities, establishes an energy level coupling mechanism under spatiotemporal constraints, and ultimately realizes the quantitative assessment of abnormal states in complex systems; the system output value reflects the comprehensive degree of anomaly of the monitored object in the spatiotemporal continuum.

[0093] Example 6: Based on Example 5, the abnormal field strength convergence sub-component provided in this embodiment of the invention includes:

[0094] The boundary guidance field propagation module is used to propagate nonlinear interference fields within the dynamic boundary of semantic blocks. The real-time deformation amplitude of the boundary generates a curvature guidance effect, which redirects the energy flow of the interference field along the curvature gradient of the dynamic boundary, forming an energy flow beam that converges towards future spatiotemporal coordinates.

[0095] The historical baseline field filtering module is used to converge the energy flow beam through the historical abnormal field strength baseline stored in the historical field conduction mode database. The normal conduction mode in the historical field conduction mode database forms the field transmission impedance distribution. The component in the energy flow beam that matches the impedance distribution is absorbed and attenuated, and the remaining abnormal energy flow density distribution passes through the baseline.

[0096] The trajectory phase delay compensation module is used to input the future extension points of the spatiotemporal trajectory of the abnormal energy flux density distribution. The topological path length of the future extension points is converted into the field propagation delay. The energy flux density distribution performs phase advance compensation based on the delay to generate a pre-synchronized energy flux hypersurface.

[0097] The field strength peak collapse module is used to cause energy level collapse of the pre-synchronized energy flow hypersurface at the target spatiotemporal coordinate point. It is modulated by the dynamic boundary deformation rate of the semantic block, and the collapse time window is determined by the duration of the behavior primitive segment. Finally, the field strength scalar value of the collapse point is output, which is the comprehensive anomaly score.

[0098] The working principle and beneficial effects of the above technical solution are as follows: The boundary guiding field propagation module of this embodiment is used for the propagation of the nonlinear interferometric field within the dynamic boundary of the semantic block. The real-time deformation amplitude of the boundary generates a curvature guiding effect, which causes the energy flow of the interferometric field to be redirected along the curvature gradient of the dynamic boundary, forming an energy flow beam that converges towards future spatiotemporal coordinates; the historical baseline field filtering module is used to converge the energy flow beam through the historical abnormal field strength baseline stored in the historical field conduction mode database. The normal conduction mode in the historical field conduction mode database forms a field transmission impedance distribution, and the component in the energy flow beam that matches the impedance distribution... The anomalous energy flux density distribution, after absorption and attenuation, passes through the baseline. The trajectory phase delay compensation module is used to input the future extension point of the spatiotemporal trajectory of the anomalous energy flux density distribution. The topological path length of the future extension point is converted into a field propagation delay. The energy flux density distribution undergoes phase advance compensation based on the delay, generating a pre-synchronized energy flux hypersurface. The field strength peak collapse module is used to cause energy level collapse of the pre-synchronized energy flux hypersurface at the target spatiotemporal coordinate point. This collapse is modulated by the dynamic boundary deformation rate of the semantic block. The collapse time window is determined by the duration of the behavioral primitive segment, and the final output is the scalar value of the field strength at the collapse point, i.e., the comprehensive anomaly score. This scheme redirects the energy flux of the nonlinear interference field through the dynamic boundary curvature gradient, converging the originally diffuse field energy into a directional energy flux beam pointing towards the future spatiotemporal coordinates according to the curvature guidance effect; thus realizing the transformation of energy flux from isotropic distribution to spatiotemporal focusing. Using an impedance distribution field constructed from a historical field conduction mode database, mode-matched filtering is performed on the energy flux beam. Normal conduction modes form high-impedance regions that absorb matching components, retaining only the energy flux density distribution consistent with historical anomalous modes through the baseline to extract anomalous features. For the topological path delay characteristics of future extension points, phase-advance compensation is implemented for the anomalous energy flux density distribution. By converting path length into a time delay, an energy flux hypersurface synchronized with the target spatiotemporal point is generated, ensuring the temporal consistency of field strength superposition. Within the collapse time window defined by the duration of the behavioral primitive segment, the pre-synchronized energy flux hypersurface undergoes energy level collapse under dynamic boundary deformation rate modulation. The multidimensional field strength distribution is transformed into a single-point scalar value output, the numerical amplitude of which characterizes the comprehensive anomaly degree.

[0099] In summary, this embodiment achieves directional energy flow convergence through curvature guidance, filters anomalous modes using historical baselines, ensures spatiotemporal synchronization through phase compensation, and finally outputs quantified anomalous indicators through controlled collapse. The output values ​​reflect accurate anomaly measurement results verified by spatiotemporal calibration and historical modes.

[0100] Example 7: Based on Example 6, the field strength peak collapse module provided in this embodiment of the invention includes:

[0101] The deformation rate field constraint focusing submodule is used to pre-synchronize the energy flow hypersurface at the target spatiotemporal coordinate point by the dynamic boundary deformation rate of the semantic block. The positive value range of the boundary deformation rate enhances the curvature of the hypersurface, and the negative value range diffuses the energy flow distribution, generating a spatiotemporal focusing energy flow density kernel.

[0102] The sustained time window field truncation submodule is used to apply the duration of the behavior primitive segment to the focused energy flux density kernel. The duration length is mapped to the radius of the energy flux constraint loop, and the duration volatility is converted into the constraint loop permeability. The energy flux envelope is bound through the constraint loop truncation time window.

[0103] The baseline normalized field convergence submodule is used to bind the energy flow envelope input to the historical abnormal field strength baseline in the time window. The normal field strength value matched with the spatiotemporal coordinates in the baseline is used as the normalization reference. The field strength gradient flux of the energy flow envelope relative to the reference value is extracted to generate the normalized field strength manifold.

[0104] The peak scalar field generation submodule is used to normalize the field strength manifold as it undergoes energy level collapse at the target spatiotemporal coordinate point. The collapse process is controlled by the current deformation rate of the semantic block, and the output is the global field strength constraint scalar at the collapse point, which is the comprehensive anomaly score.

[0105] The working principle and beneficial effects of the above technical solution are as follows: In this embodiment, the deformation rate field constraint focusing submodule is used to pre-synchronize the energy flow hypersurface at the target spatiotemporal coordinate point under the dynamic boundary deformation rate modulation of the semantic block. The positive value range of the boundary deformation rate enhances the hypersurface curvature, while the negative value range diffuses the energy flow distribution, generating a spatiotemporal focused energy flow density kernel. The duration window field truncation submodule is used to apply the duration of the behavioral primitive segment to the focused energy flow density kernel. The duration length is mapped to the radius of the energy flow constraint loop, and the duration fluctuation rate is converted into the constraint loop permeability. The constraint loop extracts the time window and binds the energy flow envelope; the baseline normalization field convergence submodule is used to bind the energy flow envelope to the historical anomalous field strength baseline. The normal field strength value matching the spatiotemporal coordinates in the baseline is used as the normalization benchmark. The field strength gradient flux of the energy flow envelope relative to the benchmark value is extracted to generate the normalized field strength manifold; the peak scalar field generation submodule is used to make the normalized field strength manifold undergo energy level collapse at the target spatiotemporal coordinate point. The collapse process is controlled by the current deformation rate of the semantic block to control the shrinkage rate. The output is the global field strength constraint scalar at the collapse point, which is the comprehensive anomalousness score. The above scheme's peak field strength collapse module achieves precise conversion from pre-synchronized energy flow hypersurface to comprehensive anomaly scalar output through the cascaded processing of four sub-modules. The deformation rate field constraint focusing sub-module modulates the pre-synchronized energy flow hypersurface with the deformation rate of the semantic block's dynamic boundary, forming a spatiotemporal focused energy flow density kernel at the target spatiotemporal coordinate point. Positive deformation rate values ​​enhance the hypersurface curvature, promoting energy flow convergence; negative values ​​diffuse the energy flow distribution, avoiding excessive concentration; ensuring optimal focusing of the energy flow kernel in the spatiotemporal dimension. The continuous time window field truncation sub-module maps the duration of the behavioral primitive segment to the radius of the energy flow constraint loop, and its volatility determines the constraint loop penetration rate, forming a dynamic truncation window. This window is bound to the energy flow envelope, ensuring that anomaly calculation is based only on the effective energy flow interval of the target spatiotemporal point, avoiding irrelevant interference. The baseline normalized field convergence submodule matches the energy flux envelope with historical anomalous field strength baselines via a time window, using the normal field strength value as the normalization benchmark. It generates a normalized field strength manifold by extracting the field strength gradient flux of the energy flux envelope relative to the benchmark, thus eliminating the impact of baseline fluctuations on anomaly assessment. The peak scalar field generation submodule causes the normalized field strength manifold to undergo controlled energy level collapse at the target spatiotemporal coordinate point, with the collapse rate dynamically adjusted by the current deformation rate. Finally, it outputs the global field strength constraint scalar at the collapse point, i.e., the comprehensive anomaly score, which reflects the precise anomaly intensity after normalization calibration and spatiotemporal constraints.

[0106] In summary, this embodiment optimizes energy flow focusing through deformation rate modulation, ensures accurate calculation range by utilizing time window constraints, and eliminates system bias by combining baseline normalization. Finally, it generates a highly reliable anomaly scalar during controlled collapse. The output value integrates the quantization results of spatiotemporal focusing, historical calibration, and dynamic constraints, providing an accurate field strength index for anomaly detection.

[0107] Example 8: Based on Example 7, the peak scalar field generation submodule provided in this embodiment of the invention includes:

[0108] The deformation rate field dynamic loading unit is used to apply the current deformation rate of the semantic block to the normalized field strength manifold. A positive deformation rate generates radial contraction field dynamics, and a negative deformation rate generates tangential shear field stress. The radial contraction field dynamics and tangential shear field stress are superimposed to form a dynamic collapse force field.

[0109] The constraint ring field boundary solidification unit is used to control the collapse range by the energy flow constraint ring parameters. The constraint ring radius limits the collapse action area, the constraint ring permeability determines the field strength dissipation threshold, and the dynamic collapse force field is compressed into a closed collapse cavity by the constraint ring.

[0110] The manifold equipotential surface folding unit is used to normalize the field strength manifold in a closed collapse cavity. Along the field strength gradient flux direction, the manifold is driven by the dynamic collapse force field to generate abrupt changes in the curvature of the equipotential surface, triggering continuous folding of the field strength equipotential surface.

[0111] The global constraint scalar condensation unit is used to solidify the field strength gradient flux along the axis of the collapsed cavity when the curvature of the folded equipotential surface reaches the permeability limit of the constraint ring at the target spatiotemporal coordinate point. The output is an irreversible field strength constraint scalar, which is the comprehensive anomaly score.

[0112] The working principle and beneficial effects of the above technical solution are as follows: In this embodiment, the deformation rate field dynamic loading unit is used to apply the current deformation rate of the semantic block to the normalized field strength manifold. A positive deformation rate generates radial contraction field dynamics, and a negative deformation rate generates tangential shear field stress. The radial contraction field dynamics and tangential shear field stress are superimposed to form a dynamic collapse force field. The constraint ring field boundary solidification unit is used to control the collapse range using energy flow constraint ring parameters. The constraint ring radius limits the collapse domain, and the constraint ring permeability determines the field strength dissipation threshold, dynamically... The collapse force field is compressed into a closed collapse cavity by the constraint loop. The manifold equipotential surface folding element is used to normalize the field strength manifold within the closed collapse cavity. Driven by the dynamic collapse force field, the equipotential surface curvature abruptly changes along the field strength gradient flux direction, triggering continuous folding of the field strength equipotential surface. The global constraint scalar condensation element is used at the target spatiotemporal coordinate point of the folded equipotential surface. When the curvature abrupt change reaches the constraint loop permeability limit, the field strength gradient flux directionally solidifies along the collapse cavity axis, outputting an irreversible field strength constraint scalar, i.e., the comprehensive anomaly score. This scheme, through a physical field dynamics mechanism driven by the semantic block deformation rate, transforms the normalized field strength gradient flux into an irreversible anomaly scalar output under the constraint of the behavior duration window. The semantic block dynamic boundary deformation rate is directly converted into a collapse force field; the radial contraction field dynamics / tangential shear field stress provide the physical driving force for the collapse of the field strength manifold. The constraint loop parameters, analyzed by the duration of the behavior primitive segment, construct the closed collapse cavity, locking the scope and dissipation conditions of the scalar generation. Within the collapsed cavity, the field strength gradient flux extracted along the historical baseline is used to achieve continuous folding of equipotential surfaces, completing the energy form transformation. When the fold curvature exceeds the constraint ring penetration threshold, the gradient flux solidifies directionally along the cavity axis, generating anomaly quantifications with time irreversibility.

[0113] In summary, this embodiment forms an end-to-end field physics transformation chain from dynamic boundary deformation to anomalous scalar output. The deformation rate serves as the power source and scalar solidification controller, the duration of the behavior determines the spatiotemporal constraint framework for scalar generation, the historical field strength gradient flux guides the energy folding direction, and the constraint loop penetration threshold triggers an irreversible scalar phase transition. This module constitutes the ultimate physical convergence node for anomaly score generation, and its output scalar serves as the core input parameter for the downstream intent deduction system.

[0114] Example 9: Based on Example 8, the global constrained scalar condensed unit provided in this embodiment of the invention includes:

[0115] The curvature-permeability critical triggering subunit is used to match the field strength scale between the curvature mutation value of the equipotential surface and the permeability threshold of the constraint ring. When the curvature mutation value exceeds the permeability threshold, the axial constraint field of the collapse cavity is activated, generating the flux solidification induction axis.

[0116] The gradient flux axial alignment sub-unit is used when the direction of the field strength gradient flux is constrained by the induced axis. When the angle between the flux vector and the axis is greater than the critical angle, it is forced to correct by the radial contraction field dynamics to form an axially co-directional gradient flux bundle.

[0117] The dual-field dynamic crystallization compression subunit is used to compress the flux distribution along the axial direction of the co-axial gradient flux bundle in the collapse cavity, while the radial contraction field compresses the flux distribution along the axial direction and the tangential shear field stress strips away the non-axial flux components to generate flux crystallization nuclei.

[0118] The constrained scalar phase transition output subunit is used to guide the crystallization nucleus at the target spatiotemporal coordinate point. The phase transition space is limited by the radius of the constrained loop. The crystallization rate is modulated by the semantic block deformation rate, and the output is an irreversible field strength constrained scalar.

[0119] The working principle and beneficial effects of the above technical solution are as follows: The curvature-permeability critical triggering subunit of this embodiment is used to match the field strength scaling between the curvature mutation value of the equipotential surface and the permeability threshold of the constraint ring. When the curvature mutation value exceeds the permeability threshold, the axial constraint field of the collapse cavity is activated, generating the flux solidification induction axis; The gradient flux axial alignment subunit is used to constrain the field strength gradient flux direction by the induction axis. When the angle between the flux vector and the axis is greater than the critical angle, it is forced to correct by the radial contraction field dynamics, forming an axial co-directional gradient flux bundle; The dual-field dynamic crystallization compression subunit is used to compress the flux distribution along the axial direction in the collapse cavity by the radial contraction field dynamics, and the tangential shear field stress strips away the non-axial flux component, generating a flux crystallization nucleus; The constraint scalar phase transition output subunit is used to output an irreversible field strength constraint scalar at the target spatiotemporal coordinate point of the flux crystallization nucleus, which is limited by the constraint ring radius in the phase transition space, modulates the crystallization rate through the semantic block deformation rate. The above scheme achieves directional control and scalar output of field strength energy through multi-level physical field coupling. By real-time monitoring of the curvature abrupt change characteristics of the equipotential surface and dynamic comparison with a preset permeability threshold, the axial constraint field generation function is immediately activated when the curvature disturbance exceeds the tolerance value, constituting the primary trigger condition of the system and ensuring that the energy disturbance starts the subsequent processing flow within a controllable range. A dynamic forced correction algorithm is used to perform real-time angle correction on the flux vector deviating from the axis. When the flux-axis angle exceeds the critical parameter, the radial contraction field automatically intervenes to calibrate the vector direction, ensuring that all gradient fluxes strictly maintain axial propagation characteristics. Energy purification is achieved through the synergistic effect of orthogonal field forces: the radial contraction field generates an axial compression effect, the tangential shear field performs transverse component stripping, and the dual-field coupling forms a three-dimensional filtering mechanism; this process ultimately produces flux crystallization nuclei with high directional consistency. A dynamic constraint loop is set at the target coordinate point, and the phase transition space range is precisely controlled by the radius parameter; combined with the deformation rate parameter of the semantic block, the crystallization process is rate-modulated, and a stable scalar value that meets the spatiotemporal constraint conditions is finally output.

[0120] In summary, this embodiment constructs a complete field strength energy directional control chain, forming a closed-loop control physical quantity conversion system from initial disturbance detection to final scalar output. Through the cascading action of four processing modules, the original field strength energy is transformed into a constraint scalar with definite spatiotemporal characteristics, meeting the application requirements of high-precision energy control. The collaborative work of each module ensures the irreversibility and directional stability of the output results, making it suitable for scenarios requiring precise control of field strength distribution.

[0121] Example 10: As Figure 5 As shown, based on Embodiment 1, the intent-based triggering system provided in this embodiment of the invention includes:

[0122] The state chain field phase subsystem is used to compile the behavioral state chain into a temporal field phase source. Each behavioral primitive segment is transformed into a field strength oscillation phase angle, and the state transition intensity is mapped to the phase synchronization intensity to form a spatiotemporal phase field wavefront.

[0123] Anomaly field intensity modulation subsystem is used to integrate the anomaly score applied to the phase field wavefront. The anomaly score is converted into a field strength attenuation weight. The phase field attenuation is enhanced in high anomaly regions, while the phase field propagation is maintained in low anomaly regions, generating an anomaly-modulated phase field distribution.

[0124] The semantic gradient field reconstruction subsystem is used for dynamic region semantic-driven field distribution reconstruction. The semantic block boundary deformation rate generates the semantic barrier gradient, which is interfered with by the anomalously modulated phase field to reconstruct the semantically constrained future field strength gradient map.

[0125] The target path field convergence subsystem is used to allow the future field strength gradient map to naturally evolve along the descent direction of the field strength gradient when it propagates in the topological network. The path continuity is locked by the phase synchronization intensity, and finally converges to the optimal path field streamline under the constraint of semantic barrier. The convergence point of the optimal path field streamline is the target region.

[0126] The working principle and beneficial effects of the above technical solution are as follows: The state chain field phase transformation subsystem of this embodiment is used to compile the behavioral state chain into a temporal field phase source. Each behavioral primitive segment is transformed into a field strength oscillation phase angle, and the state transition intensity is mapped to the phase synchronization intensity to form a spatiotemporal phase field wavefront. The anomaly field intensity suppression subsystem is used to integrate the anomaly score and apply it to the phase field wavefront. The anomaly score is converted into a field strength attenuation weight. The phase field attenuation is enhanced in high anomaly regions, and the phase field propagation is maintained in low anomaly regions, generating an anomaly-modulated phase field distribution. The semantic gradient field reconstruction subsystem is used to reconstruct the dynamic region semantic-driven field distribution. The semantic block boundary deformation rate generates a semantic barrier gradient, which interferes with the anomaly-modulated phase field to reconstruct a semantically constrained future field strength gradient map. The target path field convergence subsystem is used to allow the future field strength gradient map to evolve naturally along the field strength gradient descent direction when propagating in the topological network. The path continuity is locked by the phase synchronization intensity, and finally converges to the optimal path field streamline under the semantic barrier constraint. The convergence point of the optimal path field streamline is the target region. The above scheme achieves accurate prediction and convergence control of future behavior paths through multi-level field phase modulation and semantic constraint optimization. Discrete behavioral state chains are transformed into a continuous spatiotemporal phase field, where each behavioral primitive segment is mapped to a field strength oscillation phase angle, and the state transition intensity corresponds to the phase synchronization intensity, forming a dynamically evolving phase field wavefront; this subsystem provides the basic field distribution for subsequent inference. Dynamic attenuation modulation of the phase field is applied based on anomaly scores, with enhanced field strength attenuation in high-anomaly regions and maintained field propagation in low-anomaly regions; this mechanism ensures that anomalous interference factors are suppressed, preserving a phase field distribution with high credibility. A semantic barrier gradient is generated through the deformation rate of semantic block boundaries and interferes with the anomaly-modulated phase field to reconstruct a future field strength gradient map that conforms to semantic constraints; this subsystem ensures that the inference results conform to logical and semantic rules. The future field strength gradient map evolves along the gradient descent direction in the topological network, maintaining path continuity under the constraint of phase synchronization intensity, and converges to the optimal path field streamline under the constraint of the semantic barrier; the final convergence point is the target region that conforms to semantic constraints and has the lowest anomaly.

[0127] In summary, this embodiment constructs a complete intent inference and path optimization system, forming a closed-loop predictive control process from the field phase transformation of the behavior state chain to the convergence of the final target path. Through anomaly screening, semantic constraint interference, and gradient descent optimization, it ensures that the inference results not only conform to logical rules but also possess optimal path characteristics. It is suitable for application scenarios requiring high-precision behavior prediction and autonomous decision-making.

[0128] Example 11: As Figure 6 As shown, based on Examples 1-10, the automated control method for the AI ​​human motion capture sensor provided in this embodiment of the invention includes the following steps:

[0129] S100: Users set area monitoring conditions and expected actions through an intuitive interface to form an original instruction set; the user-preset rules are matched with the semantic block information output in real time by the dynamic area identification system; when changes in physical space layout cause semantic block deformation, the scope of the rules is automatically corrected to ensure that the instructions are always anchored on the dynamically changing semantic area, forming an executable instruction set.

[0130] S200: Input the behavioral state chain output in real time by the anomaly assessment system and the corresponding comprehensive anomaly score into the prediction model; high-score abnormal events will temporarily increase the prediction weight of related behavioral patterns; based on the anomaly fusion result of the state chain, the intention inference triggers the system to recalculate the future behavioral path and target area of ​​the target human body, and the abnormal data significantly changes the inference direction, generating the prediction intention of anomaly perception.

[0131] S300: Combines the predicted intent of anomaly perception with the current environmental state and compiles it into a device operation pre-sequence; before the operation pre-sequence takes effect, it continuously compares the conformity between the human state in the latest semantic block output by the dynamic region identification system and the predicted intent; when the verification is passed, the pre-sequence is released; when the deviation exceeds the threshold, the sequence is discarded and re-prediction is triggered, and the final execution instruction is output.

[0132] The working principle and beneficial effects of the above technical solution are as follows: In this embodiment, the user first sets the area monitoring conditions and expected actions through an intuitive interface to form an original instruction set; the user-preset rules are matched with the semantic block information output in real time by the dynamic area identification system; when changes in physical space layout cause semantic block deformation, the scope of the rules is automatically corrected to ensure that the instructions are always anchored on the dynamically changing semantic area, forming an executable instruction set; secondly, the behavior state chain output in real time by the anomaly assessment system and the corresponding comprehensive anomaly score are input into the prediction model; high-score anomaly events will temporarily increase the prediction weight of related behavior patterns; based on the anomaly fusion result of the state chain, the intention-driven deduction triggers the system to recalculate the future behavior path and target area of ​​the target human body, and the abnormal data significantly changes the deduction direction, generating anomaly perception prediction intention; finally, the anomaly perception prediction intention is combined with the current environmental state and compiled into a device operation pre-sequence; before the operation pre-sequence takes effect, the conformity between the human body state in the latest semantic block output by the dynamic area identification system and the prediction intention is continuously compared; when the verification is passed, the pre-sequence is released; when the deviation exceeds the threshold, the sequence is discarded and re-prediction is triggered, and the final execution instruction is output. The above scheme achieves an adaptive decision-making mechanism based on dynamic semantic space anchoring through multi-system closed-loop collaboration. Its technical integration effects are manifested in: dynamic rule-space coupled execution, establishing a dynamic mapping relationship between the rule's scope and physical space deformation through continuous matching of user instruction sets and real-time semantic blocks; when the three-dimensional topology undergoes non-rigid deformation due to human activity, the system automatically reconstructs the instruction's boundary, ensuring that the control logic is always bound to the semantic region rather than fixed coordinates, thus solving the problem of rule failure caused by dynamic environmental changes in traditional spatial partitioning control. Anomaly-driven prediction correction uses the anomaly score of the behavioral state chain as a dynamic weight adjustment parameter for the probabilistic graphical model. Through real-time reconstruction of the state transition matrix, intention inference becomes anomaly sensitive; when a high-anomaly behavioral pattern is detected, the system automatically enhances the transition probability calculation weight of relevant state nodes, causing the inference path to bias towards the anomaly-associated region, achieving a shift from passive anomaly detection to proactive prediction and defense. The prediction-perception dual-verification execution relies on both the predicted intent and the real-time semantic block state for the generation of the operation pre-sequence. A dual verification mechanism is established through continuous comparison during the pre-execution phase. This mechanism uses the time-varying characteristics of the human state within the semantic block as a verification benchmark. When the deviation between the predicted path and the actual observation exceeds the topological connectivity tolerance, the current control flow is immediately interrupted and a re-inference based on the latest spatial state is initiated to ensure that the final execution instruction always conforms to the dynamic environment constraints.

[0133] In summary, this embodiment achieves spatial adaptability of control rules through a semantic block deformation compensation mechanism, dynamically reconstructs the state transition space of the probabilistic graphical model using anomaly scores, and constructs a mutual feedback verification architecture between predicted output and real-time perception. Ultimately, it achieves self-stabilizing control based on human behavioral intent in unstructured environments, without relying on a preset environmental coordinate system or a fixed behavioral rule library.

[0134] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of equivalents of this invention, this invention is also intended to include these modifications and variations.

Claims

1. An AI human motion capture sensor, characterized in that, Include: The anomaly assessment system tracks the movement path and dwelling nodes of a human body within a dynamically labeled topology network, forming a coordinate-independent spatiotemporal trajectory sequence. Based on trajectory velocity, direction change rate, and regional dwelling pattern, it automatically segments and labels these segments as coherent behavioral primitives, stringing them together into a time-ordered behavioral state chain. The system calculates the comprehensive anomaly score in real time using state transition probability deviation, duration deviation, and regional taboo triggering. The comprehensive anomaly score, behavioral state chain, and corresponding spatiotemporal trajectory sequence are then encoded and multiplexed, and transmitted to a remote monitoring center via wired or wireless communication interfaces.

2. The AI ​​human motion capture sensor as described in claim 1, characterized in that, Anomaly assessment system, including: The topology trajectory aggregation subsystem is used to record the movement path between nodes and mark the stationary nodes by continuously tracking the propagation direction and intensity attenuation ratio of field distortion caused by the human body between topology nodes based on a dynamically labeled topology map. The movement path and the stationary nodes together constitute a spatiotemporal trajectory sequence; The behavioral primitive field decomposition subsystem is used to automatically cut the spatiotemporal trajectory sequence into behavioral primitive segments when the rate of change of the propagation delay of the spatiotemporal trajectory sequence exceeds the adaptive threshold, the field strength gradient direction angle is continuously reversed, the propagation intensity attenuation ratio is abruptly triggered to cut the path, or the cumulative amount of field distortion reaches the saturation threshold. Each behavioral primitive segment carries its internal field propagation uniformity characteristics and endpoint field distortion intensity value. The state chain field coupling subsystem is used to trigger semantic block matching by the field distortion intensity of the endpoints of the behavioral primitive segments: the behavioral primitive segments are connected in series through field transmission continuity verification to form a behavioral state chain weighted by block semantics. The state transition intensity is determined by the field distortion transmission loss rate of adjacent behavioral primitive segments. The abnormal field distortion assessment subsystem is used for three-dimensional anomaly detection driven by the field transmission characteristics of the behavioral state chain. It generates a spatiotemporally fused anomaly score by integrating three field distortion indicators: state transition probability deviation, duration deviation, and regional taboo triggering. The score weight is adaptively adjusted by the dynamic boundary deformation amplitude of the semantic block.

3. The AI ​​human motion capture sensor as described in claim 2, characterized in that, State transition probability deviation in the abnormal field distortion assessment subsystem: the probability density deviation between the actual conduction loss rate and the historical field conduction mode between adjacent behavioral primitive segments; duration deviation: the statistical deviation between the field conduction uniformity within a single behavioral primitive segment and the historical uniformity distribution of the corresponding semantic block; regional taboo triggering: when a taboo semantic block suddenly appears at the endpoint of a behavioral primitive, it triggers a backtracking verification of field distortion conduction based on topological edge weights.

4. The AI ​​human motion capture sensor as described in claim 2, characterized in that, Anomaly field distortion assessment subsystem, comprising: The conduction loss field interference component is used to generate conduction path interference patterns. When the actual loss rate causes distortion of the historical interference pattern on a specific topology path, the field strength coherence attenuation of the specific topology path is calculated as a physical representation of the state transition probability deviation. A uniform field resonant component is used to output the field resonant energy dissipation rate as a physical measure of duration deviance when the uniform distribution of the behavioral primitive segment detunes from the cavity characteristic frequency by matching the inherent resonant frequency spectrum of the historical uniform field resonant cavity. The taboo backtracking field verification component is used to drive the topological edge weight network to start reverse field propagation triggered by the behavioral primitive endpoints of taboo semantic blocks; backtracking from the endpoints along the ingress propagation path, the propagation trajectory is reconstructed through the edge weight decay gradient; when the reconstructed trajectory conflicts with the field distortion phase of the taboo block dynamic boundary, the length of the illegal propagation path is generated. The distortion level fusion component is used to convert the field strength coherence attenuation into topological path interference level, the field resonance energy dissipation rate into resonant cavity dissipation level, and the illegal conduction path length into phase conflict level. The phase conflict level generates a spatiotemporal fusion anomalous field through the field distortion superposition principle, and its field strength peak value is the comprehensive anomalousness score.

5. The AI ​​human motion capture sensor as described in claim 4, characterized in that, The distortion level fusion component includes: The boundary deformation modulator component is used to convert the field strength coherence attenuation into a topological path interference energy level through nonlinear scaling of the deformation amplitude; the field resonance energy dissipation rate is modulated by the deformation amplitude frequency and converted into a resonant cavity dissipation energy level; the illegal conduction path length is converted into a phase conflict energy level by deformation amplitude curvature renormalization. The spatiotemporal coupling sub-component of energy levels is used to constrain the topological path interference energy levels in the spatial dimension by the topological edge weights of the original conduction path; the resonant cavity escaping energy levels are bound to the oscillation period in the temporal dimension by the duration of the behavioral primitive segment; the phase conflict energy levels are distributed along the topological nodes of the taboo backtracking trajectory; energy level gradient diffusion is performed; and energy level carriers with separated spatiotemporal properties are formed. The distortion field superposition exciton component is used to generate a spatial interference ripple field by superimposing field distortions on the energy level carrier after spatiotemporal coupling. The topological path interference energy level excites the spatial interference ripple field, the resonant cavity escape energy level generates a time-decayed oscillation field, and the phase conflict energy level forms a forbidden gradient vortex field, which generates nonlinear interference in the four-dimensional field coupler. The anomalous field strength convergence sub-component is used to propagate the nonlinear interference field within the dynamic boundary of the semantic block. Guided by boundary deformation, it forms an energy flow convergence and is filtered by the historical anomalous field strength baseline. Finally, the field strength peak is generated at the future extension point of the spatiotemporal trajectory, which is the comprehensive anomaly score.

6. The AI ​​human motion capture sensor as described in claim 5, characterized in that, The anomalous field strength convergence sub-component includes: The boundary guidance field propagation module is used to propagate nonlinear interference fields within the dynamic boundary of semantic blocks. The real-time deformation amplitude of the boundary generates a curvature guidance effect, which redirects the energy flow of the interference field along the curvature gradient of the dynamic boundary, forming an energy flow beam that converges towards future spatiotemporal coordinates. The historical baseline field filtering module is used to converge the energy flow beam through the historical abnormal field strength baseline stored in the historical field conduction mode database. The normal conduction mode in the historical field conduction mode database forms the field transmission impedance distribution. The component in the energy flow beam that matches the impedance distribution is absorbed and attenuated, and the remaining abnormal energy flow density distribution passes through the baseline. The trajectory phase delay compensation module is used to input the future extension points of the spatiotemporal trajectory of the abnormal energy flux density distribution. The topological path length of the future extension points is converted into the field propagation delay. The energy flux density distribution performs phase advance compensation based on the delay to generate a pre-synchronized energy flux hypersurface. The field strength peak collapse module is used to cause energy level collapse of the pre-synchronized energy flow hypersurface at the target spatiotemporal coordinate point. It is modulated by the dynamic boundary deformation rate of the semantic block, and the collapse time window is determined by the duration of the behavior primitive segment. Finally, the field strength scalar value of the collapse point is output, which is the comprehensive anomaly score.

7. The AI ​​human motion capture sensor as described in claim 1, characterized in that, Data transmission employs time-division multiplexing, allocating anomaly scores, behavioral state chains, and spatiotemporal trajectory sequences to different time slots for transmission.

8. The AI ​​human motion capture sensor as described in claim 1, characterized in that, It also includes a dynamic region labeling system, which continuously captures multi-dimensional field changes caused by human activity within the covered space through an intelligent sensor array; inputs the multi-dimensional field changes into an adaptive spatial modeling engine, and automatically constructs a three-dimensional topological relationship map describing the relative positions and connectivity between spatial points in the covered area through continuous field analysis without a preset grid; and dynamically labels semantic blocks in the topological map based on user-defined semantic rules and the output of the real-time three-dimensional topological relationship map.

9. The AI ​​human motion capture sensor as described in claim 1, characterized in that, It also includes an intent-based triggering system, which integrates behavioral state chains, anomaly scores, and dynamic region semantics to predict the most likely behavioral path and target area of ​​the target human body in the next few seconds using a probability graph.

10. An automated control method for an AI human motion capture sensor, characterized in that, Includes the following steps: Users set regional monitoring conditions and expected actions through an intuitive interface to form an original instruction set; the user-preset rules are matched with the semantic block information output in real time by the dynamic regional identification system; when changes in physical space layout cause semantic block deformation, the scope of the rules is automatically corrected to ensure that the instructions are always anchored on the dynamically changing semantic region, thus forming an executable instruction set. The behavioral state chain output in real time by the anomaly assessment system and the corresponding comprehensive anomaly score are input into the prediction model; high-score anomaly events will temporarily increase the prediction weight of related behavioral patterns; Based on the state chain anomaly fusion results, the intent-based inference triggering system recalculates the target human body's future behavior path and target area. Anomaly data significantly changes the inference direction, generating anomaly perception and predictive intent. The predicted intent of anomaly perception is combined with the current environmental state and compiled into a device operation pre-sequence. Before the operation pre-sequence takes effect, the consistency between the human state in the latest semantic block output by the dynamic region identification system and the predicted intent is continuously compared. When the verification is successful, the pre-sequence is released. When the deviation exceeds the threshold, the sequence is discarded and re-prediction is triggered, and the final execution instruction is output.

Images