A wounded person treatment simulation training system based on somatosensory interaction data collection

Abstract

The application provides a wounded person treatment simulation training system based on somatosensory interaction data collection, and relates to the field of education demonstration body interaction technology. The application obtains a desensitization skeleton joint sequence stripped of RGB color, respectively extracts a high-frequency micro-vibration aggregation index representing micro-muscle compensation, and action sequence information entropy representing macro decision load; constructs a body-brain two-dimensional dissipation phase space mapping matrix, orthogonally maps the above heterogeneous features into two-dimensional state coordinate points, generates an indication signal representing the underlying cause of decline by solving the phase vector included angle; finally analyzes the signal and generates a targeted control instruction to drive the output of matched correction teaching materials by external audio-visual teaching equipment, supplemented by confidence closed-loop degradation monitoring. The application effectively decouples body energy overdraft and cognitive overload state, overcomes the attribution limitation of existing single-dimensional space comparison, and realizes accurate targeted audio-visual interaction correction in a high-pressure simulation environment.

Description

Technical Field

[0001] This invention relates to the field of educational demonstration body interaction technology, specifically a simulated training system for treating wounded soldiers based on body-sensing interactive data acquisition. Background Technology

[0002] With the deep integration of virtual reality and sensor capture technologies, the field of electronic education and demonstration tools is undergoing a paradigm shift from "one-way information transmission" to "deep body interaction." In skills training involving high-risk operations, complex processes, and high psychological stress, it is not only necessary to evaluate the trainee's final operational results post-training, but also to capture and intelligently analyze the trainee's continuous body characteristics in real time during training. This dynamically triggers matching audiovisual teaching content, forming a closed-loop training mechanism of "perception-feedback-correction." Existing simulation training applications for casualty care have limitations in terms of deep interaction and attribution analysis. For example, existing patent CN117523936A proposes an interactive group training method based on evaluation feedback, which mainly generates a training list through user parameter selection and relies on post-training evaluation feedback. This rule-based action comparison and post-event evaluation mechanism faces a deep attribution bottleneck: when the subject experiences motion stagnation, distortion, or abnormal tremors during complex and high-pressure simulated emergency rescue tasks (continuous cardiopulmonary resuscitation or tactical bandaging for massive bleeding), existing technology can only broadly classify it as "non-standard action" based on deviations from spatial coordinate thresholds, failing to delve deeper into the underlying causes. This one-dimensional physical comparison logic struggles to decouple the two highly intertwined states of "physical exhaustion (physiological level muscle compensation and fatigue)" and "cognitive overload (psychological level decision chain disruption and high tension)" from their temporal entanglement. This attribution challenge directly leads to a lack of targeting in subsequent audiovisual instruction triggered by the system. For example, when a trainee pauses due to forgetting operational steps, the system may incorrectly push audiovisual content guiding physical recovery or force posture adjustment, thus increasing the trainee's cognitive load. Summary of the Invention

[0003] The purpose of this invention is to provide a simulation training system for casualty treatment based on somatosensory interactive data acquisition. It constructs a two-dimensional dissipative phase space mapping architecture for the body and brain, simultaneously extracting microscopic muscle compensation features and macroscopic decision information entropy, and performing orthogonal mapping and phase vector calculation in the two-dimensional phase space. The system can decouple physical decline and cognitive load trajectories in real time, thereby providing a deterministic basis for the precise scheduling of audiovisual teaching equipment. This addresses the problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides the following technical solution:

[0005] A simulation training system for treating wounded soldiers based on motion-sensing interactive data acquisition, specifically including:

[0006] The desensitized interactive acquisition module is configured to acquire a spatial depth coordinate stream and perform unstructured desensitization processing on the spatial depth coordinate stream to remove the RGB color dimension, so as to output a purely quantized skeleton joint sequence.

[0007] The micro-feature extraction module is configured to extract the high-frequency spatial displacement fluctuation rate of the skeleton joint sequence within a first time window and generate a high-frequency micro-vibration clustering index, which is configured to characterize the micro-muscle compensation state of the executing subject.

[0008] The macro logic extraction module is configured to perform cluster analysis on the frequency of action state transitions of the skeleton key point sequence within a second time window to generate action sequence information entropy, which is configured to characterize the macro decision load state of the executing subject.

[0009] The fusion decoupling module is configured to construct a two-dimensional dissipative phase space mapping matrix of body and brain, input the high-frequency micro-vibration aggregation index and the action sequence information entropy into the two-dimensional dissipative phase space mapping matrix of body and brain, and orthogonally map them into two-dimensional state coordinate points; and solve the slope of the tangent line of the phase vector vector pointing from the origin to the two-dimensional state coordinate points and the phase vector angle to generate an indication signal, which is configured to characterize the classification of the underlying causes of decay;

[0010] The audiovisual teaching material scheduling module is configured to parse the indication signal and generate a targeted control command based on the classification of underlying causes of decline. The targeted control command is configured to drive external teaching equipment to output corrective teaching materials that match the classification of underlying causes of decline.

[0011] The robust closed-loop module is configured to monitor the data confidence level of the skeleton joint sequence in real time, and when the data confidence level is lower than a preset degradation threshold, generate shielding control data to cut off the phase vector calculation path in the fusion decoupling module, and generate a backup scheduling instruction to trigger an environmental reset prompt.

[0012] Compared with the prior art, the beneficial effects of the present invention are:

[0013] By constructing a two-dimensional dissipative phase space mapping matrix for the body and brain, the "high-frequency micro-vibration aggregation index" output by the micro-feature extraction module and the "action sequence information entropy" output by the macro-logic extraction module are orthogonally fused. This method overcomes the dimensional entanglement problem caused by existing technologies that rely solely on absolute spatial coordinate thresholds. It decouples the two mutually masking deep states of "physical exhaustion" and "cognitive overload" from the system mechanism perspective, achieving high-precision, fine-grained quantitative classification of the underlying causes of action deformation.

[0014] The fusion decoupling module calculates the slope of the tangent line to the phase vector's time-series trajectory and the angle between the phase vectors to generate an indication signal, which is then used by the audiovisual teaching material scheduling module to generate targeted control commands. This mechanism addresses the pain point of traditional systems where teaching interventions are "off-topic" due to attribution errors. It ensures that external teaching equipment can output matching graphic or 3D corrective audiovisual teaching materials in real time and accurately based on the actual causes of decline (physiological or psychological), thereby improving the effectiveness and relevance of simulation training.

[0015] At the data acquisition front end, unstructured desensitization processing is used to remove the RGB color dimension, cutting off the risk of leakage of biological facial features at the physical source. Simultaneously, a robust closed-loop module is introduced to monitor the data confidence level of skeleton nodes in real time. When encountering boundary conditions such as physical occlusion, the phase vector calculation path is interrupted and backup instructions are triggered. This effectively resists optical noise and data incompleteness interference in complex interactive environments, ensuring the overall system architecture's anti-interference and degradation capabilities while guaranteeing data compliance. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the overall method flow of the present invention;

[0017] Figure 2 This is a schematic diagram of the technical route of the present invention;

[0018] Figure 3 Numerical verification diagram of the two-dimensional phase space dynamic decision boundary under composite high voltage disturbance. Detailed Implementation

[0019] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0020] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0021] Example 1:

[0022] Please see Figures 1 to 3 The present invention provides a technical solution:

[0023] A simulation training system for treating wounded soldiers based on motion-sensing interactive data acquisition includes:

[0024] The desensitized interactive acquisition module is configured to acquire a spatial depth coordinate stream and perform unstructured desensitization processing on the spatial depth coordinate stream to remove the RGB color dimension, so as to output a purely quantized skeleton joint sequence.

[0025] The micro-feature extraction module is configured to extract the high-frequency spatial displacement fluctuation rate of the skeleton joint sequence within a first time window and generate a high-frequency micro-vibration clustering index, which is configured to characterize the micro-muscle compensation state of the executing subject.

[0026] The macro logic extraction module is configured to perform cluster analysis on the frequency of action state transitions of the skeleton key point sequence within a second time window to generate action sequence information entropy, which is configured to characterize the macro decision load state of the executing subject.

[0027] The fusion decoupling module is configured to construct a two-dimensional dissipative phase space mapping matrix of body and brain, input the high-frequency micro-vibration aggregation index and the action sequence information entropy into the two-dimensional dissipative phase space mapping matrix of body and brain, and orthogonally map them into two-dimensional state coordinate points; and solve the slope of the tangent line of the phase vector vector pointing from the origin to the two-dimensional state coordinate points and the phase vector angle to generate an indication signal, which is configured to characterize the classification of the underlying causes of decay;

[0028] The audiovisual teaching material scheduling module is configured to parse the indication signal and generate a targeted control command based on the classification of underlying causes of decline. The targeted control command is configured to drive external teaching equipment to output corrective teaching materials that match the classification of underlying causes of decline.

[0029] The robust closed-loop module is configured to monitor the data confidence level of the skeleton joint sequence in real time, and when the data confidence level is lower than a preset degradation threshold, generate shielding control data to cut off the phase vector calculation path in the fusion decoupling module, and generate a backup scheduling instruction to trigger an environmental reset prompt.

[0030] Further specifying, the micro-feature extraction module is further configured to: perform high-pass filtering processing with a set cutoff frequency on the skeleton joint sequence to separate the heteroscedastic fluctuation component; input the heteroscedastic fluctuation component into the generalized autoregressive conditional heteroscedasticity model for time series feature fitting; extract the fluctuation aggregation metric value in the conditional variance sequence output by the generalized autoregressive conditional heteroscedasticity model, and calculate the high-frequency micro-vibration aggregation index.

[0031] Further specifying, the macroscopic logic extraction module is further configured to: map the temporal features of the skeleton key point sequence to a preset discrete action semantic set to generate a discrete semantic state sequence; based on a hidden Markov model, statistically analyze the transition probabilities of the discrete semantic state sequence within the second time window to construct a state transition probability matrix; and output the action sequence information entropy by solving the Shannon entropy distribution of the state transition probability matrix.

[0032] Further specifying, the body-brain two-dimensional dissipative phase space mapping matrix in the fusion decoupling module is further configured as follows: The high-frequency micro-vibration aggregation index and the action sequence information entropy after extreme value normalization processing are obtained respectively; the normalized high-frequency micro-vibration aggregation index is used as the horizontal axis parameter, and the normalized action sequence information entropy is used as the vertical axis parameter to locate and generate the two-dimensional state coordinate point; the phase vector magnitude of the two-dimensional state coordinate point is calculated; the phase vector angle of the two-dimensional state coordinate point is calculated; when the phase vector angle is within a first preset interval, the underlying cause of decline is classified as physical exhaustion decline; when the phase vector angle is within a second preset interval, the underlying cause of decline is classified as cognitive overload decline.

[0033] Further specifying: the body-brain two-dimensional dissipative phase space mapping matrix in the fusion decoupling module is further configured to perform nonlinear relationship calculations based on data fitting:

[0034] Obtain the temporal penalty factor characterizing the duration of continuous treatment operations, and simultaneously obtain the environmental pressure factor characterizing the intensity of global external disturbances;

[0035] The first dimension normalization mapping coefficient and the second dimension normalization mapping coefficient, which are preset in the external data carrier, are invoked through the data reading interface. The first dimension normalization mapping coefficient and the second dimension normalization mapping coefficient are configured to perform the technical function of eliminating dimension differences, so as to map heterogeneous physical quantities into unified dimensionless values.

[0036] The fusion decoupling module performs a multiplication operation between the time-series penalty factor and the first dimension-normalized mapping coefficient to obtain a first dimensionless product, and performs a multiplication operation between the environmental pressure factor and the second dimension-normalized mapping coefficient to obtain a second dimensionless product;

[0037] The first dimensionless product and the second dimensionless product are added together to output the adaptive phase vector boundary bias. This dimensionless numerical processing logic is used to compensate for heterogeneous disturbances.

[0038] Extract the preset basic phase vector boundary angle, and perform an addition operation between the basic phase vector boundary angle and the adaptive phase vector boundary offset to generate the adaptive phase vector critical boundary angle;

[0039] When performing the action of configuring the underlying cause classification of decay, the static boundary benchmark between the original first preset interval and the second preset interval is replaced with the adaptive phase vector critical boundary angle, and the cause classification decision is performed based on the dynamically updated phase space topology interval.

[0040] Further specifying, an early warning decision on the degradation trend is made based on the slope of the tangent line of the time-series trajectory:

[0041] The slope of the tangent line of the time-series trajectory is obtained, and its absolute value is compared with a preset abrupt change slope threshold.

[0042] When the absolute value of the slope of the time-series trajectory tangent is greater than the abrupt change slope threshold, and the current two-dimensional state coordinate point is still in the safe range that has not been crossed, it is determined that the executing entity is in the transient abrupt change period of sliding towards the decay state, thereby generating an early warning indication signal that characterizes the deterioration trend in advance.

[0043] The safe zone is defined as the phase space topological region where the two-dimensional state coordinate point has not yet met the boundary conditions for triggering the first preset zone or the second preset zone to make a cause classification decision.

[0044] The advance warning indication signal is incorporated into the indication signal and sent out together, enabling the audiovisual teaching material scheduling module to trigger a light-weight voice reassurance or rhythmic prompt intervention a short time before the trainee experiences substantial motor paralysis. Through this mechanism, a two-dimensional deep joint adjudication is achieved, where "the absolute classification state is determined by the phase angle and the transient evolution trend is determined by the tangent slope," thus compensating for the lag in the temporal dimension of single-point determination.

[0045] Furthermore, the audiovisual teaching material scheduling module is further configured as follows:

[0046] In response to the underlying cause classification of the decline being identified as the physical exhaustion decline, a first addressing instruction is generated for invoking the 3D spatial perspective adjustment animation.

[0047] In response to the underlying cause classification of the decline being identified as cognitive overload decline, a second addressing instruction is generated to invoke the highlighted prompts of the two-dimensional flowchart.

[0048] In response to the parsing that the indication signal contains a pre-warning indication signal representing a deterioration trend, a third addressing instruction for invoking a lightweight audiovisual prompt is generated, and the execution priority of the third addressing instruction is higher than that of the addressing instruction generated based on the underlying cause classification of the degradation.

[0049] Further specifying, the robust closed-loop module is further configured to: obtain the tracking status identifier of each skeleton joint in the current frame through the hardware low-level interface, calculate the status as the ratio of the number of tracked joints to the total number of joints baseline value, and generate the data confidence score; when monitoring and determining that the low-frequency center of gravity absolute displacement of the skeleton joint sequence is less than the silent displacement threshold, and the cumulative state of the data confidence score being lower than the preset degradation threshold exceeds the preset tolerance range, trigger the shielding control data.

[0050] The boundary angle of the basic phase vector (denoted as) This represents the baseline critical angle for judgment when the physical and cognitive abilities of the executing subject are in a balanced state of consumption under ideal, interference-free conditions. In this embodiment, the specific value is... Radius. In this embodiment, in the standard normalized coordinate system, the diagonal (45 degrees) is the geometric equilibrium point that distinguishes the weights of physical fluctuations on the horizontal axis and logical chaos on the vertical axis. The method for determining the boundary angle of the basic phase vector is defined as follows: based on mapping the high-frequency micro-vibration aggregation index and the action sequence information entropy after extreme value normalization to an orthogonal two-dimensional state coordinate system. Since the data of the horizontal and vertical axes have been forcibly mapped to the unified dimensionless absolute value domain of [0,1], this phase space constitutes an isotropic standard square topological field. The offline calibration unit of the system obtains the maximum theoretical magnitude value of the horizontal axis vector; the offline calibration unit obtains the maximum theoretical magnitude value of the vertical axis vector; the offline calibration unit performs an arctangent function calculation on the ratio of the maximum value of the vertical axis to the maximum value of the horizontal axis. Since the ratio is constant at 1, the reference angle output by this calculation is equal to Radius. In terms of topological relationships, it constitutes an absolute geometric axis of symmetry that provides an equivalent hedging between physical decay potential energy and logical disorder potential energy, requiring no external subjective intervention.

[0051] Temporal penalty factor (its parameter symbol is) This parameter represents the cumulative time effect of the executing entity being continuously exposed to high-load operational tasks. It is calculated by recording the absolute time elapsed (in seconds) since the current simulation training task began; this time elapsed is then extracted and directly assigned as the time penalty factor.

[0052] Environmental stress factor (its parameter symbol is) This is a comprehensive scalar representing the acoustic and visual pressure intensity exerted by the rescue site environment on the implementing entity. It is achieved by acquiring the average decibel value of the background loudspeaker; performing logarithmic smoothing on this average decibel value, and outputting a continuous numerical value representing the magnitude of environmental pressure as the environmental pressure factor.

[0053] The time-series penalty factor characterizing the duration of continuous treatment operations is obtained, and the environmental pressure factor characterizing the intensity of global external disturbances is obtained simultaneously. The specific data acquisition and processing logic is as follows:

[0054] The specific logic for determining the timing penalty factor is as follows: The system, through its built-in high-precision timestamp recording module, reads and records the system's absolute time at the instant the current simulation training task initialization is completed and the start command is issued, using this as the initial timestamp (parameter symbol: ...). During any current processing cycle in the task execution process, the current system absolute time is read in real time and used as the current timestamp (parameter symbol is...). Calculate the current timestamp. With the initial timestamp The time difference between them; extract this time difference (in seconds) and directly assign it as the timing penalty factor. .

[0055] The acquisition of environmental stress factors includes the independent calculation of acoustic stress components and visual stress components, followed by weighted fusion. The specific steps are as follows:

[0056] For the calculation of acoustic pressure components: the acquisition of environmental pressure factors depends on an environmental acoustic sensing module, which is configured to have the ability to acquire dynamic sound pressure levels within a preset range (e.g., 40dB to 120dB) to ensure the undistorted capture of background noise in simulated environments such as explosions or distress calls.

[0057] According to the preset time window (acquiring the interval 5 seconds before the current moment), the average audio decibel value of the background loudspeaker within this interval is collected (parameter symbol is...). ).

[0058] average audio decibel value A preset anti-overflow constant of 1 is added to the summation to prevent the logarithmic function from becoming meaningless due to a minimum value; a base-10 logarithmic operation is performed on the summation result to simulate the non-linear perception of sound loudness by the human ear; the logarithmic result is then compared with a preset acoustic sensitivity adjustment coefficient (parameter symbol is...). Multiplying these components together, the output is the acoustic pressure component (parameter symbol: ...). ).

[0059] For the calculation of the visual stress component: the visual interference characteristics at the rescue site are obtained through the environmental visual perception module (in this embodiment, a light sensor or a strobe detector). The visual interference characteristics specifically include the average ambient illuminance (parameter symbol: ...). ) and the flicker frequency of the light source (parameter symbol is ). Ambient average illuminance Divide by the system's preset maximum illuminance threshold (parameter symbol is...) ), to obtain the illuminance compression ratio; similarly, the flicker frequency of the light source is... Divide by the system's preset maximum flicker frequency threshold (parameter symbol is...) ), thus obtaining the flicker compression ratio.

[0060] Multiply the illuminance compression ratio by the first preset visual weight coefficient (parameter symbol is...). The flicker suppression ratio is multiplied by a second preset visual weighting coefficient (parameter symbol is...). Adding these two products together, the output scalar value is the visual stress component (parameter symbol is...). ).

[0061] Further obtain the preset acoustic prior experience weights (parameter symbol is...) ) and visual prior experience weights (parameter symbol is The sum of the two is 1. The acoustic pressure components calculated above... Multiplied by acoustic prior experience weights Visual stress components Multiplied by visual prior experience weights The product of the two factors is weighted and summed to achieve dynamic fusion of multimodal environmental disturbances. The continuous weighted sum output serves as the current environmental pressure factor. .

[0062] The parameters involved in this embodiment are explained as follows: and These are the absolute system times at the moment the task starts and the current moment, respectively, used to quantify the exposure time of the executing entity under high pressure. It is the raw environmental noise intensity assessment value collected by the environmental acoustic sensing module. It is the acoustic sensitivity adjustment coefficient, an empirical constant used to adjust the model's sensitivity to acoustic stimuli so that it conforms to the actual human ear's perception threshold (preferred range is 1.0~2.5, preferred example value is 1.2). and These are the maximum illuminance limit achievable in the physical space (in this embodiment, the critical value for blindness caused by strong light) and the maximum flicker frequency (in this embodiment, the critical frequency that causes visual fatigue or dizziness). The value range is set to 10Hz to 30Hz, with a preferred example value of 15Hz. ), used to map the acquired absolute physical quantities to the standardized interval [0,1]. and It is a sub-weight within visual features, used to balance the relative contributions of strong light and flicker to the visual pressure exerted on the subject. The value range is from 0.4 to 0.8. In this embodiment, since the interference of sudden strong light in the environment on operation is greater than that of flicker, the value is set to... The preferred example value is 0.6. The preferred example value is 0.4. and These are global modal weights, used to characterize whether auditory or visual interference is dominant in a specific simulation task (e.g., nighttime tasks). High, in noisy battlefield missions (Higher); In the typical high-noise treatment scenario of this embodiment, the following is set The preferred example value is 0.6. The preferred example value is 0.4.

[0063] Dimensionally normalized mapping coefficient set (including first-dimensionally normalized mapping coefficients) Mapping coefficients with second dimension normalization It eliminates the physical difference between the time dimension (seconds) and the sound intensity dimension (decibels), mapping heterogeneous physical quantities into dimensionless pure numerical weights that can directly participate in the angle offset calculation. This value ensures that the total offset does not exceed the basic geometric quadrant ( Under the hard constraints of [the algorithm], the optimal parameters are iteratively derived based on an expert sample library. All the configurable parameters mentioned above are predefined and stored in an external structured JSON spreadsheet file. Before executing the algorithm, a data loading module running in a Python interpreter environment first parses the spreadsheet file and loads the data into the working memory, achieving decoupling between configuration and logic. In this step, the fusion decoupling module needs to fuse the temporal penalty factor (dimension: seconds) representing time and the environmental pressure factor (dimension: decibels) representing sound pressure intensity to output an adaptive phase vector boundary bias (dimension: radians) used to correct the angle. A first dimensional normalized mapping coefficient and a second dimensional normalized mapping coefficient are introduced and defined only when the computation involves parameters with different physical properties or different units of measurement. The two mapping coefficients mentioned above are used to map heterogeneous physical quantities into unified dimensionless values. By performing multiplication operations, the inherent dimensions of the input parameters are eliminated. The calculation logic in the fusion decoupling module, which multiplies the dimension-normalized mapping coefficients with their corresponding factors and performs summation to generate the phase vector boundary bias, does not follow the laws of conservation of physical dimensions in classical mechanics. The specification explicitly defines it as "a nonlinear relationship based on data fitting" and "a preset empirical algorithm model."

[0064] For the first dimension normalized mapping coefficients Mapping coefficients with second dimension normalization How to obtain:

[0065] A multi-dimensional regression training reference dataset was constructed. Specifically, subjects were recruited and, under different decibel levels of environmental noise, the time elapsed from the start of simulation training to the occurrence of motion distortion, the rate of change of the high-frequency micro-vibration aggregation index, and the rate of change of the motion sequence information entropy were recorded. The offline training server extracted the actual phase space angular offset corresponding to the moment when the subject experienced cognitive overload as the dependent variable matrix; and extracted the corresponding time-series values ​​and environmental decibel values ​​as the independent variable matrix. The offline training server obtained the dependent and independent variable matrices and performed a multiple linear regression operation on them. The offline training server extracted the output weight vector generated by the regression operation. The offline training server truncated and solidified the scalar elements corresponding to the time-series dimension of the weight vector as first-dimensional normalized mapping coefficients, and truncated and solidified the scalar elements corresponding to the environmental dimension as second-dimensional normalized mapping coefficients. The calculated first-dimensional and second-dimensional normalized mapping coefficients were defined as read-only parameters and stored in a local structured spreadsheet file.

[0066] In traditional simulation training systems for casualty care based on two-dimensional phase space judgment, the distinction between physical exhaustion and cognitive overload relies heavily on static mapping at a fixed angle (using...). (Dividing the time into intervals). This design has a fatal blind spot: as the time for high-energy activities such as CPR increases, the accumulation of lactic acid in the body increases non-linearly. If environmental disturbances are added at this time, it can easily cause short-term cognitive stagnation in the operator. If a fixed interval is maintained... In the aforementioned complex situation, the event that should have been attributed to "physical fitness approaching its limit, leading to movement distortion" is mistakenly assigned to the "cognitive overload" quadrant due to a brief period of logical confusion. This causes the audiovisual equipment to incorrectly push "operation flowcharts" instead of "force exertion posture replacement animations" to the exhausted trainees, thus undermining the effectiveness of the training guidance.

[0067] This embodiment introduces the "dynamic feedforward compensation theory" from adaptive control theory. Treating training time and environmental noise as disturbance sources, it proactively intervenes in the topological shape of the decision boundary using a weighted bias mechanism to pre-compromise the asymmetry of physiological decay. At the implementation layer, the following logical loop is executed:

[0068] In the unstructured desensitization acquisition phase, the desensitization interactive acquisition module, specifically the depth sensor array and pre-processing chip coupled to the system motherboard, is configured to perform the following unstructured desensitization acquisition operations: The depth sensor array emits infrared structured light and receives reflected signals to acquire the spatial depth coordinates of the 3D point cloud, which is represented by the physical coordinate stream of the original 3D point cloud containing the simulation training environment and the subject being executed; The pre-processing chip performs distance-based depth threshold truncation to filter out static background point clouds in the original 3D point cloud physical coordinate stream whose depth (absolute physical distance on the Z-axis) is greater than the preset effective interactive distance, thus obtaining the foreground human point cloud data; The pre-processing chip physically blocks the underlying data bus calls of the RGB image sensor to forcibly perform unstructured desensitization processing on the foreground human point cloud data, stripping the RGB color dimension, ensuring that the data packets transmitted to the downstream working memory only contain 3D spatial geometric coordinate attributes and do not contain any pixel textures; This pure geometric coordinate foreground human point cloud data is input into a preset skeleton node localization empirical algorithm model to extract and output a pure quantized skeleton joint sequence containing the 3D coordinates of multiple joint points. The preset effective interaction distance is defined as the radius of the spherical effective spatial detection boundary with the optical center of the depth sensor as the origin; its physical unit is meters (m); in this embodiment, it is preferably 1.5 meters, but in practical applications, this parameter can be any real number between 0.5 meters and 3.0 meters. By introducing physical blocking of RGB bus calls and depth threshold truncation, sensitive biometric features such as facial texture or skin color of the executing subject are avoided from being cached in memory even briefly, thus eliminating the risk of privacy leakage at the underlying hardware level.

[0069] The micro-feature extraction module is configured to perform micro-feature extraction operations: receiving a purely quantized skeleton joint sequence; performing a high-pass filtering operation on the sequence with a cutoff frequency parameter configured to 10Hz to filter out the low-frequency smooth trajectory representing macroscopic rescue actions, and extracting the heteroscedastic fluctuation component at the micro level, which is essentially the high-frequency spatial displacement fluctuation rate of the skeleton joint sequence. The micro-feature extraction module uses the heteroscedastic fluctuation component representing the high-frequency spatial displacement fluctuation rate as a one-dimensional discrete input vector, imports it into the generalized autoregressive conditional heteroscedasticity (GARCH) model configured in its internal memory, and performs a nonlinear fitting operation; specifically, the fitting operation includes using the heteroscedastic fluctuation component as input, calculating the current conditional variance estimate through a variance update equation composed of a preset autoregressive order and a moving average order, thereby extracting the discrete conditional variance sequence array generated after model fitting; quantifying the degree of fluctuation in the discrete digital system; then setting a discrete-time sliding window containing a fixed number of sampling frames, and defining it as the first time window (set to 0.5 seconds in this embodiment). (A short observation window of 1.0 second); traversing all discrete conditional variance values ​​within the discrete-time sliding window, performing an accumulation operation on each read conditional variance value to obtain the absolute sum of the window variance; to eliminate the time-domain dimension differences caused by different sampling rate devices and ensure the stability of subsequent calculations, a normalization parameter is introduced, which is defined as the total number of sampling frames within the discrete-time sliding window; the absolute sum of the window variance is divided by the total number of sampling frames to perform an averaging operation, thereby eliminating the frame frequency dimension differences, and the fluctuation aggregation metric value is calculated; this fluctuation aggregation metric value is mapped to a unified dimensionless value, and the value is output as the high-frequency micro-vibration aggregation index (denoted as V) characterizing the micro-compensation state of muscles. agg ).

[0070] To further elaborate on the above, the calculation logic for the high-frequency micro-vibration aggregation index, which quantifies the continuous muscle compensation state and eliminates the time dimension bias caused by the sampling rate of heterogeneous sensors, is defined as follows: ;in, Defined as the normalized high-frequency micro-vibration aggregation index, it characterizes the physical aggregation potential energy level of high-frequency muscle tremors within a specific time window; in this embodiment, based on experimental calibration, the preferred normalized value is between 0.0 and 1.0. Defined as the total number of sampled frames contained within the discrete-time sliding window, it serves as a normalization parameter to eliminate temporal sampling rate differences; its data attribute is an unsigned integer scalar; in this embodiment, it is preferably 150 frames (corresponding to a 2.5-second window under 60FPS hardware), but in practical applications, this parameter takes any positive integer between 30 and 300 frames. t is a discrete-time frame index boundary variable, physically representing the current data frame number within the sliding window, and its value ranges from 1 to... A closed interval of positive integers. Defined as the discrete conditional variance value of the model output corresponding to the position of the t-th frame, it characterizes the intensity of displacement heteroscedasticity at that instant. It is a general mathematical summation operator, physically bound here to the microprocessor for all summations within a time window. An iterative operation of the accumulator register is performed on each discrete conditional variance value.

[0071] By introducing a division-average normalization method based on the total number of sampled frames, the baseline drift of the absolute sum of variance caused by accessing depth sensing hardware with different refresh rates (30Hz and 60Hz) is avoided, ensuring that the extracted high-frequency micro-vibration aggregation index has absolute numerical consistency and cross-platform robustness on heterogeneous hardware devices.

[0072] For the "microscopic feature extraction module," a cutoff frequency (10Hz in the example) and the model lag order of the generalized autoregressive conditional heteroscedasticity model (GARCH) are set. The underlying data input for the above parameters comes from the spatial depth sensing module. This spatial depth sensing module is required to have a sampling frequency of no less than 60 frames / second and a spatial ranging error of no more than ±0.5%. This functional attribute is a key prerequisite to ensure that microscopic muscle compensation features (high-frequency jitter) are not overwhelmed by sampling aliasing.

[0073] The cutoff frequency was determined as follows: When distinguishing between macroscopic rescue actions (including regular CPR compressions) and microscopic muscle compensation (including physiological exhaustion and tremors), this cutoff frequency was calibrated through controlled objective experiments. An experimental subject sample set was constructed, covering emergency responders with different physical fitness levels. Subjects were instructed to perform standard CPR compressions until muscle exhaustion was achieved. Throughout the entire cycle, the wrist and elbow skeletal joint sequences of the subject were continuously acquired using the aforementioned spatial depth sensing module. The acquired time-domain coordinate sequences were converted to the frequency domain using a Fast Fourier Transform algorithm. The spectral energy difference vector between the standardized operation period and the exhaustion and tremor period was calculated. In this embodiment, the macroscopic action energy of standardized compressions is concentrated in the 0.5Hz to 2.5Hz frequency band, while the heteroscedastic fluctuation energy generated by irregular motor unit recruitment caused by exhaustion shows a statistically significant surge with a confidence level of up to 95% concentrated in the frequency band above 10Hz. The cutoff frequency was quantized and configured to 10Hz. This constitutes the optimal physical isolation zone for filtering macroscopic action interference and fully preserving high-frequency fluctuations of muscle compensation.

[0074] The method for determining the model lag order is as follows: The GARCH model includes autoregressive order parameters and moving average order parameters. High-frequency microscopic spatial displacement components above 10Hz from the aforementioned experiments are extracted to construct a reference dataset. Partial autocorrelation function (PACF) and autocorrelation function (ACF) calculations are performed on this reference dataset. Based on the truncation characteristics of the calculated output, the autoregressive order parameter and the moving average order parameter are preferentially configured to 1, representing the GARCH(1,1) structure. The Akaike Information Criterion (AIC) is calculated based on this reference dataset. In this embodiment, the AIC value reaches the global minimum when the order is configured as (1,1). This represents a mathematically optimal balance between "ensuring that heteroscedastic fluctuation characteristics are fully captured and fitted" and "avoiding the introduction of computational redundancy that could overload the edge processor."

[0075] The macro-logic extraction phase involves performing macro-logic extraction operations to reconstruct the logical coherence of action sequences and measure disorder. The macro-logic extraction module's processor uses a dynamic time warping algorithm to calculate the relative distance between the skeleton keypoint sequence features and a preset discrete action semantic set (containing semantic markers such as 'preparing for rescue,' 'effective compression,' 'interruption and hesitation,' and 'disorderly wandering'), obtaining a set of distance values. The minimum distance value is extracted from this set, and the semantic marker corresponding to this minimum distance value is assigned to the current sequence, completing the hard-decision semantic mapping. The macro-logic extraction module uses the maximum likelihood estimation algorithm for explicit transition probabilities in Hidden Markov Models to construct a two-dimensional state transition probability matrix, specifically executing the following calculation logic:

[0076] Initialize a two-dimensional frequency matrix with dimension K×K in memory, where K is the total number of semantic states of the preset discrete action semantic set, and assign all initial elements in the matrix to a very small non-zero smoothing constant 0.001 to prevent mathematical anomalies caused by division by zero in subsequent calculations.

[0077] A time observation scale for macroscopic semantic behavior statistics is introduced, namely a second time window (the duration of which is preferably set to a fixed value between 1.0 seconds and 3.0 seconds); the discrete semantic state sequence after mapping is traversed by sliding according to the set second time window step size. Whenever a transition from the i-th semantic state to the j-th semantic state is detected, the frequency element in the i-th row and j-th column of the two-dimensional frequency matrix is ​​incremented by one until the traversal is completed, and an absolute transition matrix containing global statistical frequencies is obtained.

[0078] For each row of the absolute transition matrix, extract all frequency elements in that row and perform an accumulation operation to obtain the sum of frequencies in the current row; divide each frequency element in that row by the sum of frequencies in the current row to obtain the corresponding transition probability value, which is the transition probability of the system switching between adjacent states; fill these transition probabilities into a new matrix in sequence, thereby constructing a two-dimensional state transition probability matrix that satisfies the condition that the sum of the probabilities of each row is 100%.

[0079] The matrix processing unit of the macroscopic logic extraction module is configured to perform scalarization and dimensionality reduction information entropy calculation logic on the transition probability matrix. This calculation logic essentially solves for the Shannon entropy distribution of the state transition system: the matrix processing unit obtains the constructed state transition probability matrix; it performs matrix eigenvalue decomposition to find the left eigenvector corresponding to the eigenvalue, and performs probability normalization on the left eigenvector to obtain a one-dimensional vector of steady-state probability distribution; it iterates through each probability element in the one-dimensional vector of steady-state probability distribution, extracts the current probability element, and calculates the product of the current probability element and its natural logarithm; it accumulates all the products obtained from the iterative calculation to generate a negative sum; it multiplies the negative sum by negative one to complete the mathematical solution of the Shannon entropy distribution, thereby extracting a single scalar result, which is output as the action sequence information entropy (denoted as E) representing the decision disorder. seq ).

[0080] In a preferred embodiment, the mapping model for the entropy of the output action sequence information is shown in the following equation:

[0081]

[0082] in Defined as the entropy of the action sequence, it characterizes the degree of uncertainty and logical confusion in the transition between multiple discrete first aid action states by the executing subject; in this embodiment, it is preferably a value between 0.0 and 2.5. K is defined as the total number of semantic states in a preset discrete action semantic set; in this embodiment, it is preferably 5 (mapped to: prepare, press, blow air, check, loiter), and in practical applications, this parameter takes any positive integer between 3 and 10. k is a state index boundary variable, which physically represents the current counter for traversing discrete action semantic states, and its value ranges from 1 to K, which are positive integers. Defined as the steady-state probability value of the system in the k-th semantic state in the one-dimensional steady-state probability distribution derived from the normalization of the left eigenvector; its value range is limited to the closed interval [0,1] and satisfies the probability normalization conservation condition: . It is the natural logarithm operator, physically bound here as a logarithmic series expansion approximation calculation performed by the floating-point unit of the microprocessor. By introducing a dimensionality reduction method to extract a one-dimensional vector of the steady-state probability distribution by solving the left eigenvector of the transition matrix, overfitting of the local instantaneous action switching probability is avoided. Thus, by utilizing ergodicity, the disorder of long-term macroscopic behavior is accurately compressed into a scalar entropy value, improving the mathematical rigor and computational efficiency of measuring complex decision loads.

[0083] During the fusion decoupling and dynamic topology adjudication phase, the fusion decoupling module is configured to perform fusion decoupling and dynamic topology adjudication operations: the fusion decoupling module performs extreme value normalization operations on the generated high-frequency micro-vibration aggregation index and action sequence information entropy respectively, and uniformly constrains the heterogeneous feature values ​​to the standard interval of [0,1] to eliminate amplitude differences; the fusion decoupling module initializes a two-dimensional orthogonal coordinate system in memory, and sets the high-frequency micro-vibration aggregation index after extreme value normalization as the first orthogonal axis coordinate value (horizontal axis coordinate value), and sets the action sequence information entropy after extreme value normalization as the second orthogonal axis coordinate value (vertical axis coordinate value), thereby depicting the orthogonal two-dimensional state coordinate points at the current moment.

[0084] The fusion decoupling module performs the following orthogonal mapping and phase vector calculation steps: It obtains the high-frequency micro-vibration aggregation index after extreme value normalization as the first orthogonal axis coordinate value, and obtains the action sequence information entropy after extreme value normalization as the second orthogonal axis coordinate value; it performs a squaring operation on the first orthogonal axis coordinate value to obtain the first squared value, and performs a squaring operation on the second orthogonal axis coordinate value to obtain the second squared value; it adds the first squared value and the second squared value to generate a coordinate square sum, and performs a square root operation on the coordinate square sum to obtain the phase vector magnitude representing the absolute severity level of decay. Simultaneously, the fusion decoupling module obtains the second orthogonal axis coordinate value and the first orthogonal axis coordinate value, divides the second orthogonal axis coordinate value by the first orthogonal axis coordinate value to obtain the tangent quotient, and performs an arctangent function mapping operation on the tangent quotient to obtain the phase vector angle representing the relative proportion of physical and logical decay.

[0085] Furthermore, the fusion decoupling module is further configured to perform discretization of the slope of the tangent line of the time-series trajectory in order to capture the transient degradation trend of the decay state;

[0086] The fusion decoupling module constructs a circular cache queue of state coordinates in memory to continuously store historical orthogonal two-dimensional state coordinate points within a preset time window.

[0087] Extract the current state coordinates of the current time step and extract the historical state coordinates of the backtracking time difference step from the cache queue;

[0088] Calculate the difference between the vertical axis coordinate value of the current state coordinate point (representing the information entropy of the current action sequence) and the vertical axis coordinate value of the historical state coordinate point, and simultaneously calculate the difference between the horizontal axis coordinate value of the current state coordinate point (representing the current high-frequency micro-vibration aggregation index) and the horizontal axis coordinate value of the historical state coordinate point;

[0089] The system determines whether the absolute value of the horizontal axis difference is less than a preset zero floating-point tolerance. If it is less, the horizontal axis difference is forcibly replaced with the value of the zero floating-point tolerance. The vertical axis difference is divided by the horizontal axis difference, and a division operation is performed to generate a differential quotient. This differential quotient is output as the slope of the time-series trajectory tangent representing the current direction of decay evolution. The zero floating-point tolerance is defined as a minimal positive real number that prevents arithmetic overflow caused by the denominator approaching zero; in this embodiment, it is preferably a small positive real number. .

[0090] The fusion decoupling module acquires the predefined timing penalty factor and environmental pressure factor in parallel; the fusion decoupling module retrieves the first dimension normalized mapping coefficient and the second dimension normalized mapping coefficient loaded in memory respectively; and performs multiplication and accumulation operations to obtain the dimensionless adaptive phase vector boundary bias.

[0091] The fusion decoupling module is configured to perform a dynamic phase vector topology boundary adaptive mechanism for joint compensation of spatiotemporal environment. The computational logic described below, which biases geometric angles by fusing time and acoustic parameters, aims at empirically compensating for errors. Specifically, the physical mapping model for generating the adaptive phase vector critical boundary is shown in the following equation:

[0092]

[0093]

[0094] in, Defined as the time-series penalty factor, which is the absolute operation time accumulated from the start of training to the current moment, representing the noise floor of natural physical fatigue decay; the physical unit is seconds (s). The environmental pressure factor is defined as the average background audio intensity emitted by an environmental amplification device after logarithmic smoothing; its physical unit is decibel (dB). Defined as the first dimension normalized mapping coefficient, used to map seconds to a dimensionless bias; in this embodiment, the preferred value is 0.0005, and in practical applications, it is a real number between 0.0001 and 0.001. Defined as the second dimension normalization mapping coefficient, used to map decibels to a dimensionless bias; in this embodiment, the preferred value is 0.002, and in practical applications, it is a real number between 0.0005 and 0.005. Defined as the dimensionless adaptive phase vector boundary bias. Defined as the basic phase vector boundary angle characterizing the static boundary reference under ideal, undisturbed conditions; in this embodiment, the preferred value is... The 45-degree angle is represented. Defined as the critical boundary angle of the final generated adaptive phase vector; used to dynamically delineate the first preset interval. With the second preset interval By introducing a time-series and environment-based joint weighted summation mechanism with dimensional mapping coefficients, the use of a fixed [method / mechanism] is avoided. Static geometric boundaries, when physical energy is misjudged as cognitive overload under extreme fatigue and high noise superposition, give the fusion decoupling system an adaptive topology decision-making ability to sense environmental pressure, reducing the cause misjudgment rate under extreme edge conditions.

[0095] The logic adjudication unit within the fusion decoupling module is configured to perform dynamic classification configuration adjudication actions: the logic adjudication unit compares the calculated real-time phase vector angle with the adaptive phase vector critical boundary angle. When it is determined that the real-time phase vector angle is less than the critical boundary angle (falling within the updated first preset interval dynamically contracted or expanded by the adaptive critical boundary angle), the logic adjudication unit confirms that physical decay potential energy is dominant, thereby outputting an indication signal characterizing physical overexertion decay; conversely, when it is determined that the phase vector angle is greater than or equal to the critical boundary angle (falling within the updated second preset interval), the logic adjudication unit confirms that logical disorder is dominant, thereby outputting an indication signal characterizing cognitive overload decay.

[0096] The backtracking time difference step size is defined as the time span parameter when calculating the discrete state derivative; in this embodiment, it is preferably 30 frames to achieve a balance between "filtering transient high-frequency coordinate jitter noise" and "maintaining the real-time sensitivity of trend calculation".

[0097] The abrupt change slope threshold is defined as the critical slope value for determining the uncontrollable abrupt change in the decline state; it is a dimensionless scalar; in this embodiment, it is preferably 5.0 (characterizing that the deterioration rate of the vertical axis is more than 5 times that of the horizontal axis, and the cognitive load collapses precipitously in a very short time).

[0098] Targeted audiovisual teaching material scheduling phase: The audiovisual teaching material scheduling module is configured to perform targeted audiovisual teaching material scheduling operations: analyze the indication signal generated by the fusion decoupling module; and enter the internal conditional addressing branch network based on the analysis results;

[0099] The audiovisual teaching material scheduling module determines whether the indication signal contains a pre-warning indication signal triggered by the slope of the tangent of the time trajectory; in response to the presence of a pre-warning indication signal, the audiovisual teaching material scheduling module pushes a third addressing instruction into the system communication bus to drive the external audio device to output a lightweight voice soothing or press rhythm synchronization metronome, thereby performing non-invasive intervention before the subject of execution suffers substantial paralysis.

[0100] In parallel, in response to the underlying cause being analyzed as physical exhaustion and decline, the audiovisual teaching material scheduling module pushes a first addressing instruction onto the communication bus to accurately match and call the three-dimensional spatial perspective adjustment animation in the external teaching material library involving "changing the rescue posture or adjusting the muscle exertion point"; in response to the cause being analyzed as cognitive overload and decline, the audiovisual teaching material scheduling module pushes a second addressing instruction onto the communication bus to drive the external screen to display a highlighted two-dimensional flowchart in the form of a "next emergency operation tree diagram" with voice guidance.

[0101] Furthermore, the audiovisual teaching material scheduling module pre-configures and maintains a structured teaching material mapping table. This mapping table is securely stored in the system's local non-volatile memory or in a cloud communication interaction node. The mapping table pre-defines the corresponding mapping relationships between various decline-causing state identifiers (including physical exhaustion decline, cognitive overload decline, and deterioration trend warning) and the physical storage addresses of multimedia files in the external teaching material library.

[0102] The specific execution logic of the audiovisual teaching material scheduling module for generating various addressing instructions is as follows: using the currently determined classification of underlying causes of degradation or the deterioration warning status as the index key, the module queries and retrieves the teaching material mapping table, extracts the specific multimedia file address that matches the index key, and encapsulates the address according to a preset communication protocol format, thereby outputting the machine-readable first addressing instruction, second addressing instruction, or third addressing instruction. Through an index encapsulation mechanism based on a preset structured form, the deterministic transformation of abstract physiological / psychological diagnostic results into the driving level of physical external devices is achieved.

[0103] The robust closed-loop module is configured to execute dynamic monitoring logic for underlying data quality to ensure a minimum performance degradation: During the desensitized interactive acquisition process, the underlying spatial depth sensor generates a hardware-level tracking status identifier (including tracked or lost) for each independent spatial skeleton joint in the captured human skeleton. The robust closed-loop module is configured to obtain the tracking status identifiers of all skeleton joints in the current frame in real time through the hardware application programming interface (API); count the number of joints with tracked status and obtain a preset baseline value for the total number of human joints; divide the number of tracked joints by the baseline value for the total number of joints to generate an effective tracking ratio value within the closed interval [0,1], which is defined as the data confidence level used to characterize the point cloud quality of the current frame. In parallel, the robust closed-loop module is configured to continuously capture the low-frequency absolute displacement of the basic skeleton centroid from the main coordinate stream. Its specific calculation logic is as follows: for the skeleton joint sequence of the current frame, calculate the average three-dimensional coordinates of all joints to obtain the current frame centroid coordinates; input the current frame centroid coordinates into a preset low-pass filter to filter out high-frequency jitter noise, obtaining smoothed centroid coordinates; calculate the Euclidean space distance between the smoothed centroid coordinates and the smoothed centroid coordinates of the previous adjacent frame, and output this distance value as the low-frequency absolute displacement; the robust closed-loop module sends this low-frequency absolute displacement value to a comparator and performs a safety threshold comparison operation.

[0104] When the robust closed-loop module determines that the data confidence level is lower than the preset degradation threshold of 0.4, and simultaneously detects that the absolute value of the low-frequency displacement is less than the aforementioned silent displacement threshold, it starts the internal timer for abnormal states; when it determines that the number of missing frames accumulated continuously in the time dimension of this dual abnormal state reaches or exceeds the preset tolerance range set by offline calculation (indicating that extreme physical occlusion or sensor loss has caused the center of gravity drift to be lower than the system's background noise, i.e., confidence avalanche has occurred), the robust closed-loop module confirms the triggering of the boundary condition;

[0105] At this point, the robust closed-loop module is configured to forcibly block and truncate the output channels of the phase space matrix's magnitude and bias, and forcibly overwrite the main output terminal with a backup scheduling instruction to drive the external speaker to trigger a pure audio environment reset prompt, thereby completely blocking the distribution of error correction instructions.

[0106] In this embodiment, the determination of the silent displacement threshold and the preset tolerance range is as follows: the robust closed-loop module truncates the phase vector calculation path under occlusion conditions. A statistical continuous determination is introduced by the offline calibration unit: during the absolutely static idle period in the simulation training environment without human interaction, the offline calibration unit activates the desensitized interactive acquisition module to continuously acquire a reference spatial depth coordinate stream for a specified duration; the offline calibration unit extracts the low-frequency absolute displacement sequence of the centroid of the virtual skeleton joints generated by the background thermal noise of the optical devices during this period (wherein, the coordinates of the centroid are obtained by calculating the mean of the coordinates of all relevant nodes); the offline calibration unit performs statistical quantile analysis on this absolute displacement sequence to calculate the 99th percentile value of its probability density distribution; the offline calibration unit obtains the 99th percentile value, multiplies it by a preset safety margin coefficient, and the final product output is configured as the silent displacement threshold.

[0107] The offline calibration unit introduces a preset safety margin coefficient to prevent false triggering caused by minor environmental disturbances. The safety margin coefficient is determined by being predefined and stored in an external data carrier in JSON format deployed on a cloud server or a local independent file system. Before the offline calibration unit executes the final threshold generation logic, the data loading subunit is configured to parse the external data carrier through a read interface and extract the value of the safety margin coefficient into the current working memory. In this preferred embodiment, the specific value of the safety margin coefficient is set to 1.15. This value effectively covers the 15% of abnormal optical jitter peaks outside the 99th percentile, achieving the best technical balance between avoiding frequent false positive degradation and ensuring timely blocking of misleading commands in cases of extreme occlusion. This value is then permanently stored in the static threshold register of the online robust closed-loop module for real-time access.

[0108] For the preset tolerance range (representing the cumulative number of frames indicating consecutive missing data confidence indicators): the offline calibration unit obtains the calibration refresh rate parameter of the aforementioned spatial depth perception module; the offline calibration unit multiplies the average time of a single emergency rescue action (including the compression and rebound phase) with the calibration refresh rate parameter to obtain the theoretical number of frames of the characteristic period of a single action; the offline calibration unit extracts half of the theoretical number of frames of the characteristic period, rounds it down, and establishes it as the preset tolerance range. This ensures that the system can tolerate brief limb self-occlusion, but once the occlusion duration cuts off the smallest semantically recognizable unit of a single action, the backup scheduling instruction generated by the robust closed-loop module is immediately triggered to ensure that the feedback adjustment mechanism has a clear temporal closed-loop direction.

[0109] Under prolonged high-noise environment boundary conditions (continuous pressing for 15 minutes accompanied by extremely high decibel external interference), the dynamic bias is pushed to the maximum constraint threshold by the algorithm. At this point, the adaptive critical boundary angle will shift significantly towards the vertical axis. The system behavior verification shows that even if noise causes significant student hesitation (abnormal increase in vertical axis information entropy), as long as the micro-vibration aggregation index remains at an extremely high level, the system can still accurately encompass the coordinate point by relying on the expanded offset "first preset interval," stably outputting the "physical exhaustion" command, achieving highly robust anti-interference physiological state recognition.

[0110] In this invention, the core output values ​​of the fusion decoupling module are the "phase vector angle" and "phase vector magnitude" in a two-dimensional orthogonal phase space. Since the high-frequency micro-vibration aggregation index and the action sequence information entropy at the input are both constrained to a closed interval by the system's extreme value normalization operation, the theoretical range of the phase vector angle is limited to the range of zero to half a π radians (i.e., 0 degrees to 90 degrees). The technical state of a minimum trend (phase vector angle approaching zero): When the calculated phase vector angle approaches the theoretical minimum of zero degrees, it indicates that the microscopic physical jitter representing the horizontal axis component within the system is approaching its maximum, while the macroscopic logical disorder representing the vertical axis component is approaching its minimum. This quantitatively represents the extreme physical exhaustion state exhibited by the executing subject during emergency operations, where "severe failure of the muscle recruitment unit leads to high-frequency physiological spasms, but the subconscious memory of the rescue process remains clear and continuous." At this point, the fusion decoupling module confirms that the current action deformation is purely caused by the exhaustion of the physical entity (representing physical strength), rather than by obstruction of the brain's decision-making center.

[0111] The technical state of maximum trend (phase vector angle approaching 90 degrees): When the calculated phase vector angle approaches the theoretical maximum of 90 degrees, it indicates that the vertical axis component (disorder degree of the action sequence) is absolutely dominant. This quantitatively characterizes the severe break in the decision-making chain and behavioral hesitation caused by battlefield high pressure, panic, or memory loss in the executing subject, even before irreversible compensatory tremors occur in the physiological muscles. At this time, the fusion decoupling module confirms that the root cause of the action deformation lies in the logical paralysis caused by cognitive overload.

[0112] The high-frequency micro-vibration aggregation index is positively correlated with the phase vector modulus (characterizing the severity of decline) and also positively correlated with the cosine component of the phase vector angle (leading the judgment focus to the area of ​​physical exhaustion). When human skeletal muscles approach the fatigue threshold, the synchronization of motor neuron impulses declines, macroscopically manifested as high-frequency, uncontrollable heteroscedastic physiological tremors at specific cutoff frequencies (above 10Hz) in the extremities. By setting the spatial displacement fluctuation rate in the micro-dimensional dimension as a positively correlated input variable of the system, it is ensured that the system can capture and quantify weak precursor signals of muscle compensation in advance within the time window before the macro-level emergency actions (including the depth of CPR compressions) of the executor completely fail, directly supporting the technical effect of the "proactive fatigue warning" of this invention.

[0113] The entropy of action sequence information is positively correlated with the phase vector magnitude and also positively correlated with the sine component of the phase vector angle (causing the judgment focus to tend towards the cognitive overload region). This design originates from information theory and Markov decision process theory. Skilled first aid operations are characterized by deterministic, high-probability state transitions, while psychological panic or fuzzy step memory inevitably leads to a uniform distribution of the action state transition matrix, resulting in a large number of invalid step regressions or hesitations (manifested as high entropy). By reducing the dimensionality of macroscopic action transition frequencies to a single scalar entropy, the system can objectively quantify the psychological high-pressure load on the executor without infringing on facial biometric privacy (without collecting sensitive data such as eye movements and expressions), directly supporting the high-precision intent recognition capability of this invention under unstructured desensitization conditions.

[0114] The value of the adaptive phase vector boundary bias is positively correlated with the product and sum of the time-series penalty factor (continuous operation time) and the environmental pressure factor (external noise intensity). The fusion decoupling module directly adds this bias to the basic phase vector boundary angle, thereby dynamically expanding the first preset interval (physical exhaustion judgment interval) within the system. As the logic execution entity is exposed to excessive environmental noise and continuous heavy physical output tasks for a period of time, the rate at which its tolerance to physical fatigue decreases is much greater than the rate at which its mechanical motion memory decays. This dynamic bias design based on heterogeneous parameter fusion endows the fusion decoupling module with the ability to topologically adapt and expand in response to external physical environmental pressure, ensuring that the accuracy of underlying cause attribution does not plummet under extremely severe superimposed disturbance environments.

[0115] The absolute value of the tangent slope of the temporal trajectory is positively correlated with the probability of the system triggering an early warning intervention. When the absolute value of this slope exceeds a preset abrupt change slope threshold, a pre-warning indicator signal with higher priority than the absolute state classification is forcibly triggered. This design is based on catastrophe theory in dynamic systems. In complex emergency high-pressure environments, the physiological or psychological breakdown of trainees often does not occur linearly, but rather exhibits a "nonlinear slippage period." Capturing the derivative of the phase space coordinate motion trajectory (representing the tangent slope) can detect this precipitous change earlier than relying solely on absolute coordinate values. Introducing the discretized temporal trajectory tangent slope into the adjudication logic enables the system to predict the state evolution trend within a very short future time window, thereby providing lightweight intervention before the trainee experiences substantial motor paralysis, directly supporting the technical effect of the "transient abrupt change early warning" of this invention.

[0116] This embodiment is configured in a battlefield casualty treatment digital twin monitoring and verification scenario to demonstrate the data flow patterns of the system's computational kernel. The pre-parameterized conditions are preset as follows: the spatial depth coordinate stream sampling rate of the desensitized interactive acquisition module is locked at 60 frames per second; the sliding window length of the micro-feature extraction module is set to 150 frames per second; the clustering semantic set of the macro-logic extraction module includes five standard discrete first aid action states; in the fusion decoupling module, the basic phase vector boundary angle is fixed at π / 4 radians (equivalent to 45 degrees in degrees); the first dimension normalized mapping coefficient determined by the offline calibration unit is 0.0005 radians / second, and the second dimension normalized mapping coefficient is 0.002 radians / decibels; the degradation threshold of the robust closed-loop module is set to 0.4; simultaneously, to filter out normal physiological micro-movements and reasonable pauses in action thinking of the executing subject, a preset module length activation threshold is set, which in this embodiment is preferably set to 0.3.

[0117] Table 1 shows an example of the logical deduction and calculation of the system's internal state variables and execution instructions when the physiological and psychological input parameters of the executing entity change.

[0118] Table 1: Examples of system response calculations under different environmental disturbances

[0119] Experimental scenario description Operation time (s) Noise intensity (dB) High-frequency micro-vibration aggregation index Action information entropy Dynamic critical boundary angle (degrees) Real-time phase angle (degrees) Cause offset Final target control command output Scenario 1: Initial Stabilization Period 30 50 0.15 0.1 51.59 33.69 -17.9 None (phase vector mode length did not reach activation threshold) Scenario 2: Short-term memory loss 60 60 0.2 0.85 53.59 76.76 23.17 Highlighting prompts when calling a 2D flowchart Scenario 3: Physical exhaustion 300 70 0.9 0.25 61.62 15.52 -46.1 Calling 3D perspective adjustment animation Scenario 4: Dual Composite High Voltage A 300 110 0.88 0.85 66.2 43.99 -22.21 Calling 3D perspective adjustment animation Scenario 5: Dual Composite High Voltage B 600 110 0.8 0.85 74.79 46.74 -28.05 Calling 3D perspective adjustment animation Scenario 6: Sensor-based avalanche blocking N / A N / A Confidence level < 0.4 Displacement < Static value N / A N / A N / A Truncation of phase vector calculation triggers reset prompt

[0120] In Scenario 6, the robust closed-loop module detects that the confidence level of the input skeleton sequence data is lower than the preset 0.4 and the low-frequency absolute displacement is lower than the silence threshold. At this time, the hardware-level blocking logic is triggered, and the processor directly intercepts and discards all subsequent floating-point operations on the phase space matrix to avoid the unnecessary computational power consumption of erroneous data and the distribution of erroneous instructions.

[0121] The absolute offset of the causal decay classification boundary, or simply the causal offset, is denoted as... The logic adjudication unit within the fusion decoupling module acquires the angular value of the real-time phase vector angle calculated from the solution, extracts the angular value of the dynamically generated adaptive phase vector critical boundary angle, and then performs a subtraction operation between the two to obtain this parameter. This parameter is used to objectively and quantitatively characterize the direction and intensity of the deviation of the current executing entity's state from the system's dynamic equilibrium benchmark boundary. Its polarity directly determines the resource allocation direction of the audiovisual teaching material scheduling module: when the value is negative, it quantitatively indicates that the coordinate point falls into the physical decay domain; when the value is positive, it quantitatively indicates that the coordinate point falls into the logical decay domain.

[0122] The modulus activation threshold is denoted as The threshold is derived from the two-dimensional spatial geometric distance after extreme value normalization. Under normal and stable emergency rescue conditions (Scenario 1 in Table 1), the high-frequency micro-vibration aggregation index (0.15) and action information entropy (0.10) of the executing subject are both at extremely low background levels. The fusion decoupling module performs spatial distance calculation: the squares of 0.15 and 0.10 are calculated respectively (0.0225 and 0.01), and the sum is 0.0325. The real-time phase vector mode length obtained after taking the square root is approximately 0.18. To ensure that the system does not "over-intervene" due to normal physiological micro-movements, the offline statistical unit extracts the maximum background phase vector mode length from a large number of qualified emergency rescue samples (the maximum background phase vector mode length in this embodiment is distributed between 0.15 and 0.25), and adds a preset anti-shake constant of 0.05 on the basis of its upper limit, and calculates and solidifies the mode length activation threshold of 0.3. This is used to objectively define the "tolerance error range" and the "functional failure boundary". Only when the real-time phase vector magnitude is greater than or equal to the activation threshold of that magnitude does it indicate that the severity of the decay has reached a level that requires intervention, thereby effectively blocking invalid alarms in low-hazard states.

[0123] The "Scenario 4 (Dual Composite High Voltage A)" and "Scenario 5 (Dual Composite High Voltage B)" in Table 1 were extracted to form a limit theory control group for mechanism analysis.

[0124] Under these two extremely high-pressure conditions, the actuators face extremely high-intensity micro-physical vibrations (high-frequency micro-vibration aggregation indices reaching 0.88 and 0.80 respectively) and extremely high-intensity macro-motion disorder (information entropy values ​​soaring to a high level of 0.85). The system input exhibits a multimodal phenomenon of "dual collapse of body and brain." According to coordinate system calculations, the real-time phase vector angles of the module output at this time are respectively... and .

[0125] Compared to the benchmark scheme that does not apply the technology of this invention (the system does not introduce spatiotemporal environment joint compensation and is highly dependent on...), As a static mapping boundary benchmark: the real-time angle between state coordinate points in scenario five ( The baseline system will forcibly determine that the "logically disordered load is dominant," causing the audiovisual teaching material scheduling module to incorrectly distribute "operation flowcharts" to trainees whose physical strength is completely and irreversibly exhausted. This rigid static threshold mapping, lacking dynamic environmental compensation, leads to errors in the correction instructions. Conversely, as can be seen from the computational data flow shown in Table 1 of this invention: in the same scenario five, due to the activation of the nonlinear relationship calculation logic based on data fitting within the fusion decoupling module, the product operation and cumulative compensation of the time-series penalty factor (600 seconds) and the environmental pressure factor (110dB) are introduced, so that the system's "adaptive phase vector critical boundary angle" is legally and accurately dynamically extended to the theoretical algorithm. With the intervention of this adaptive expansion mechanism, the real-time phase angle ( The data is still effectively and stably contained within the updated first preset interval by the logic decision unit (at this time, the cause offset remains stably negative at -28.05). Therefore, the audiovisual teaching material scheduling module still accurately generates the first addressing instruction for calling the "3D spatial perspective adjustment animation".

[0126] This invention transforms environmental pressure into a dimensionless phase vector boundary bias. Under severe working conditions at the edge of compound high pressure, by actively reconstructing the tolerance boundary of the topological judgment interval, it corrects the underlying cause misjudgment that inevitably occurs in the baseline system and restores it to a precise perception of physical exhaustion.

[0127] Through the dual constraints of the fusion decoupling module and the robust closed-loop module, three application zones with clear physical meanings and independent resource scheduling strategies are constructed from the ground up. Their boundary delineation is not subjectively set, but rather a unique technical solution derived from heterogeneous data fusion and underlying hardware feedback:

[0128] Interval 1, Physical Exhaustion Physical Intervention Interval (First Preset Interval): Core Output Value Range: Real-time phase vector angle is less than the adaptive phase vector critical boundary angle. Furthermore, the phase vector magnitude is greater than or equal to the system's preset magnitude activation threshold. Its boundary is formed by the extreme value normalization mapping result based on multi-source sensor input, superimposed with the dynamic angle after nonlinear empirical compensation by temporal and acoustic environmental factors. This topological interval represents the energy weight of the high-frequency abnormal electrical activity of the muscle nerve unit of the executing subject in the system vector space, and has an absolute mathematical statistical advantage in the probability entropy of the disordered state transition of macroscopic actions. Once the state coordinates fall into this interval, the logic adjudication unit confirms that the physical decay potential energy is dominant, and controls the audiovisual teaching material scheduling module to generate the first addressing instruction. This instruction wakes up the external teaching equipment to retrieve and play the three-dimensional spatial perspective adjustment animation, forcing the system to intervene in the dynamic correction process of the rescuer's physical force exertion posture.

[0129] Interval 2, the second preset interval to which the cognitive overload logic reconstruction interval belongs, has a real-time phase vector angle greater than or equal to the adaptive phase vector critical boundary angle. And the phase vector length The threshold value is greater than or equal to the system's preset activation threshold. This boundary breach signifies that the real-time phase angle has exceeded the adaptive tolerance extreme value after extreme expansion by the environmental penalty factor. The surge in information entropy caused by the macroscopic movement deformation sequence substantially overshadows the background noise of basic muscle tremors resulting from continuous physical exertion. At this point, the system confirms that the core fault node causing functional failure has been completely transferred from the underlying muscle actuators to the decision-making logic unit in the brain's central nervous system. In response to this classification decision, the audiovisual teaching material scheduling module generates a second addressing instruction. This instruction retrieves a highlighted two-dimensional flowchart containing a tree diagram of emergency procedures, utilizing dimensionality reduction to strip away environmental interference and purely visual logical information to reconstruct and organize the subject's confused short-term working memory chain.

[0130] Interval 3, the confidence avalanche hardware shielding interval, is the core output value range of the underlying fallback interval: the confidence of the extracted point cloud data is lower than the preset fallback threshold, and the calculated absolute displacement of the low-frequency centroid is less than the silent displacement threshold generated by the offline calibration unit, and the cumulative duration of this abnormal state exceeds the tolerance range. Its boundary defense is directly constructed by the underlying hardware tracking state identifier missing rate of the spatial depth sensor, combined with the limit product of the background thermal noise of the idle system at the 99th statistical quantile. In the system priority matrix, the decision-making power of this interval always has the highest priority. Its underlying basis is the basic axiom of Shannon's information theory: once a large-scale missing source input data destroys the completeness of the system's information entropy, any phase vector solution result built on the incomplete matrix completely loses its mathematical legitimacy. At the moment this interval is triggered, the robust closed-loop module physically cuts off the output data bus of the phase space matrix operation, blocking the propagation of abnormal data. Simultaneously, shielding control data is generated and the main output is forcibly overwritten as a backup scheduling command, driving the underlying sound unit to play a pure audio environment reset prompt that is detached from complex graphic analysis, thus completely cutting off the control link that causes secondary teaching hazards to the system.

[0131] Interval four is the transient abrupt change early warning interval, representing the intervention interval for deterioration trends. The absolute value of the slope of the calculated time-series trajectory tangent is greater than the preset abrupt change slope threshold, and the real-time phase vector angle is still within the safe interval (not triggering the judgment of interval one or interval two), while the phase vector magnitude is greater than or equal to the magnitude activation threshold. Its boundary is established by the difference quotient of the two-dimensional phase space coordinates within the continuous time step. Once the state characteristics match the conditions of this interval, the logic decision unit generates a command containing the early warning indication signal; the audiovisual teaching material scheduling module responds to this signal first, generating and pushing the third addressing command. This command directionally wakes up the external audio device to output a lightweight voice soothing or press the rhythm synchronization metronome, thereby completing non-invasive advance intervention before the subject of execution suffers substantial paralysis.

[0132] Figure 3This is a schematic diagram illustrating the results of a numerical verification experiment on the dynamic topological boundary adaptive adjudication mechanism in the fusion decoupling module, as described in one embodiment of the present invention. This numerical verification aims to test the theoretical response and topological deflection characteristics of the present invention under preset standardized boundary conditions (i.e., the dual composite high-pressure input test vectors of scenarios four and five in Table 1). In the figure, the horizontal axis represents the normalized high-frequency micro-vibration aggregation index output by the micro-feature extraction module, and the vertical axis represents the normalized action sequence information entropy output by the macro-logic extraction module. Different line types represent the constant static baseline and the adaptive critical boundary calculated based on the logic of the present invention, respectively, and the scatter points represent the two-dimensional orthogonal state coordinate points calculated under specific test scenarios. The numerical calculation results clearly reveal that the dynamic adaptive critical boundary undergoes a significant positive topological deflection compared to the static baseline. The spatiotemporal environment joint compensation mechanism proposed in this invention can effectively correct the rigidity defects in the judgment under extreme conditions. Specifically, within the upper right test area representing the dual extreme loads, the calculated coordinates of a specific test state representing scenario five show a spatial distribution effectively contained within the physical decay interval (first preset interval) defined by the dynamically extended boundary. In stark contrast, if the static baseline is used under the same test vector, this coordinate point crosses the boundary, appearing incorrectly segmented within the logical decay interval (second preset interval). This difference in relative position determination directly confirms that introducing an adaptive phase vector boundary offset allows for the legitimate reconstruction of the tolerance boundary of the topology determination interval, avoiding attribution misjudgments and outputting correct targeted control commands.

[0133] It should be noted that all calculation formulas in this application employ regression analysis, including but not limited to machine learning algorithms, to deeply analyze the collected parameters and identify their natural trends and interrelationships. Specialized software, such as Python's Scikit-learn library or the R language, is used to automatically generate mathematical models that match the data. Then, cross-validation and other methods are used to objectively evaluate the model performance, and continuous feedback and optimization are combined to ensure that the created formulas truly reflect the inherent laws of the data, thereby guaranteeing their effectiveness and accuracy. In all calculation formulas in this application, the parameters in each formula undergo dimensionless processing within a consistent range to ensure that different physical quantities are compared on the same scale; dimensionless processing techniques include, but are not limited to, min-max-normalization and Z-score standardization.

[0134] The technical solution of this invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random-access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of various embodiments of this invention.

[0135] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0136] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A simulation training system for treating wounded soldiers based on motion-sensing interactive data acquisition, characterized in that, Specifically, it includes: The desensitized interactive acquisition module is configured to acquire a spatial depth coordinate stream and perform unstructured desensitization processing on the spatial depth coordinate stream to remove the RGB color dimension, so as to output a purely quantized skeleton joint sequence. The micro-feature extraction module is configured to extract the high-frequency spatial displacement fluctuation rate of the skeleton joint sequence within a first time window and generate a high-frequency micro-vibration aggregation index. The macroscopic logic extraction module is configured to perform cluster analysis on the frequency of action state transitions of the skeleton joint sequence within a second time window to generate action sequence information entropy. The fusion decoupling module is configured to construct a two-dimensional dissipative phase space mapping matrix of body and brain, input the high-frequency micro-vibration aggregation index and the action sequence information entropy into the two-dimensional dissipative phase space mapping matrix of body and brain, and orthogonally map them into two-dimensional state coordinate points; and solve the slope of the tangent line of the phase vector vector pointing from the origin to the two-dimensional state coordinate points and the phase vector angle to generate an indication signal, which is configured to characterize the classification of the underlying causes of decay; The audiovisual teaching material scheduling module is configured to parse the indication signal and generate a targeted control command based on the classification of underlying causes of decline. The targeted control command is configured to drive external teaching equipment to output corrective teaching materials that match the classification of underlying causes of decline. The robust closed-loop module is configured to monitor the data confidence level of the skeleton joint sequence in real time, and when the data confidence level is lower than a preset degradation threshold, generate shielding control data to cut off the phase vector calculation path in the fusion decoupling module, and generate a backup scheduling instruction to trigger an environmental reset prompt.

2. The Casualty Treatment Simulation Training System Based on Somatosensory Interaction Data Acquisition according to claim 1, characterized in that: The micro-feature extraction module is further configured to: perform high-pass filtering processing with a set cutoff frequency on the skeleton joint sequence to separate the heteroscedastic fluctuation component; input the heteroscedastic fluctuation component into the generalized autoregressive conditional heteroscedasticity model for time series feature fitting; extract the fluctuation aggregation metric value in the conditional variance sequence output by the generalized autoregressive conditional heteroscedasticity model, and calculate the high-frequency micro-vibration aggregation index.

3. The Casualty Treatment Simulation Training System Based on Somatosensory Interaction Data Acquisition according to claim 2, characterized in that: The macroscopic logic extraction module is further configured to: map the temporal features of the skeleton key point sequence to a preset discrete action semantic set to generate a discrete semantic state sequence; based on a hidden Markov model, calculate the transition probability of the discrete semantic state sequence within the second time window to construct a state transition probability matrix; and output the action sequence information entropy by solving the Shannon entropy distribution of the state transition probability matrix.

4. The Casualty Treatment Simulation Training System Based on Somatosensory Interaction Data Acquisition according to claim 3, characterized in that: The body-brain two-dimensional dissipative phase space mapping matrix in the fusion decoupling module is further configured as follows: the high-frequency micro-vibration aggregation index and the action sequence information entropy after extreme value normalization processing are obtained respectively; the normalized high-frequency micro-vibration aggregation index is used as the horizontal axis parameter, and the normalized action sequence information entropy is used as the vertical axis parameter to locate and generate the two-dimensional state coordinate point; the phase vector mode length of the two-dimensional state coordinate point is calculated. Calculate the phase angle of the two-dimensional state coordinate point; when the phase angle is within a first preset interval, classify the underlying cause of decline as physical exhaustion decline; when the phase angle is within a second preset interval, classify the underlying cause of decline as cognitive overload decline.

5. The Casualty Treatment Simulation Training System Based on Somatosensory Interaction Data Acquisition according to claim 4, characterized in that: The body-brain two-dimensional dissipative phase space mapping matrix in the fusion decoupling module is further configured to perform nonlinear relationship calculations based on data fitting: Obtain the temporal penalty factor characterizing the duration of continuous treatment operations, and simultaneously obtain the environmental pressure factor characterizing the intensity of global external disturbances; The preset first-dimensional normalized mapping coefficient and second-dimensional normalized mapping coefficient are invoked via the data reading interface. The fusion decoupling module performs a multiplication operation between the time-series penalty factor and the first dimension-normalized mapping coefficient to obtain a first dimensionless product, and performs a multiplication operation between the environmental pressure factor and the second dimension-normalized mapping coefficient to obtain a second dimensionless product; The first dimensionless product and the second dimensionless product are added together to output the adaptive phase vector boundary bias. This dimensionless numerical processing logic is used to compensate for heterogeneous disturbances. Extract the preset basic phase vector boundary angle, and perform an addition operation between the basic phase vector boundary angle and the adaptive phase vector boundary offset to generate the adaptive phase vector critical boundary angle; When performing the action of configuring the underlying cause classification of decay, the static boundary benchmark between the original first preset interval and the second preset interval is replaced with the adaptive phase vector critical boundary angle, and the cause classification decision is performed based on the dynamically updated phase space topology interval.

6. The Casualty Treatment Simulation Training System Based on Somatosensory Interaction Data Acquisition according to claim 5, characterized in that: Based on the slope of the time-series trajectory tangent, an early warning decision on the degradation trend is made: The slope of the tangent line of the time-series trajectory is obtained, and its absolute value is compared with a preset abrupt change slope threshold. When the absolute value of the slope of the time-series trajectory tangent is greater than the abrupt change slope threshold, and the current two-dimensional state coordinate point is still in the safe range that has not been crossed, it is determined that the executing entity is in the transient abrupt change period of sliding towards the decay state, thereby generating an early warning indication signal that characterizes the deterioration trend in advance. The safe zone is defined as the phase space topology region where the two-dimensional state coordinate point has not yet met the boundary conditions for triggering the first preset zone or the second preset zone to make a cause classification decision.

7. The Casualty Treatment Simulation Training System Based on Somatosensory Interaction Data Acquisition according to claim 6, characterized in that: The audiovisual teaching material scheduling module is further configured as follows: In response to the underlying cause classification of the decline being identified as the physical exhaustion decline, a first addressing instruction is generated for invoking the 3D spatial perspective adjustment animation. In response to the underlying cause classification of the decline being identified as cognitive overload decline, a second addressing instruction is generated to invoke the highlighted prompts of the two-dimensional flowchart. In response to the parsing that the indication signal contains a pre-warning indication signal representing a deterioration trend, a third addressing instruction for invoking a lightweight audiovisual prompt is generated, and the execution priority of the third addressing instruction is higher than that of the addressing instruction generated based on the underlying cause classification of the degradation.

8. The Casualty Treatment Simulation Training System Based on Somatosensory Interaction Data Acquisition according to claim 7, characterized in that: The robust closed-loop module is further configured to: obtain the tracking status identifier of each skeleton joint in the current frame through the hardware low-level interface, calculate the status as the ratio of the number of tracked joints to the total number of joints baseline value, and generate the data confidence score; when the low-frequency center of gravity absolute displacement of the skeleton joint sequence is less than the silent displacement threshold and the data confidence score is lower than the preset degradation threshold, the shielding control data is triggered.

Images