Immersive light and shadow art space interaction method and system based on somatosensory recognition

Abstract

This invention discloses an immersive light and shadow art space interaction method and system based on motion recognition. It collects user motion signals and analyzes them into posture features, then uses template comparison to filter valid action inputs. The system performs trajectory analysis on action sequences to extract rhythm distribution features and identifies emotional tendencies based on changes in action amplitude and speed, organizing the queue of actions to be processed according to the urgency of emotional response. For multi-limb synchronous movements, it partitions and decomposes the movements, identifying the temporal coordination relationship between dominant and subordinate limbs and establishing a cascading trigger mechanism. It maps action features to light and shadow control parameters and verifies rendering performance, generating optimized lighting effect execution commands. It monitors user interaction frequency and dwell time distribution to identify deviations from the immersive state, automatically triggering lighting adjustment strategies to restore user engagement. Through preloading and reuse scheduling of rendering tasks, it optimizes system resource utilization and outputs coordinated and unified spatial interaction control commands, achieving a natural fusion of action and light and shadow.

Description

Technical Field

[0001] This invention relates to the field of human-computer interaction technology, and in particular to an immersive light and shadow art space interaction method and system based on motion recognition. Background Technology

[0002] Immersive light and shadow art spaces, as a new form of exhibition that integrates digital art and interactive technology, create an immersive art experience for viewers through real-time interaction between light and shadow projection and user actions. In such spaces, accurately capturing users' body movements and transforming them into expressive light and shadow responses becomes a key factor in determining the quality of the interactive experience. Existing motion-sensing interaction solutions mostly adopt preset, fixed action-response mapping patterns, that is, triggering fixed light and shadow effects for specific actions. While this approach is simple to implement, it ignores the emotional state and rhythm inherent in user actions, resulting in a lack of depth and emotional resonance in the light and shadow responses, making it difficult to meet the high demands of art spaces for immersive experiences.

[0003] Furthermore, existing solutions typically employ a holistic recognition strategy when handling multi-limb coordinated movements, treating complex actions as a single input. This fails to fully exploit the temporal coordination and hierarchical linkage characteristics between different limb parts, making it difficult for lighting effects to present a hierarchical progression that harmonizes with natural human movement. Simultaneously, during continuous interaction between the user and the lighting space, user engagement and attention levels fluctuate over time. Existing solutions lack the ability to perceive and adaptively adjust to the user's immersion state in real time. When user attention wanders or interaction willingness declines, the lighting performance strategy cannot be adjusted promptly to rekindle user interest, impacting the integrity and consistency of the overall interactive experience. Summary of the Invention

[0004] This invention provides an immersive light and shadow art space interaction method and system based on motion recognition, aiming to solve problems such as the lack of emotional connection in motion responses, coarse processing of multi-limb linkage, and the inability to adaptively adjust the user's immersion state in existing light and shadow interaction solutions. This invention establishes a correlation mapping between motion intensity and light and shadow performance by extracting features from user limb movements and analyzing emotional tendencies; it performs limb partitioning and master-slave relationship identification for complex movements to achieve cascading triggering of light and shadow responses; and it introduces interactive behavior monitoring and deviation detection mechanisms to automatically adjust the light and shadow strategy when the user's immersion state fluctuates, providing an interactive control scheme for immersive light and shadow art spaces.

[0005] The first aspect of this invention proposes an immersive light and shadow art space interaction method based on motion perception recognition, comprising the following steps: User body sensory signals are collected and analyzed to generate body posture feature data. The body posture feature data is compared with a posture template library to determine the action recognition range. Based on the action recognition range, valid body sensory inputs are selected to establish a usable posture dataset. Based on the available posture dataset, motion trajectory analysis is performed to identify the action rhythm distribution. Emotion-related features are extracted from the action rhythm distribution to generate responsive action sequences. For composite postures in the body feature data, limb partitioning is performed to establish collaborative response groups. A preliminary mapping table is established based on the responsive action sequence and the coordinated response group to adapt lighting and shadow effects. The body feature data is converted into lighting and shadow configuration parameters. The preliminary mapping table is then rendered and verified using the lighting and shadow configuration parameters to generate an adaptation evaluation result. Based on the adaptation evaluation result, a scene-driven sequence for generating lighting and shadow projection is initiated. The interaction frequency and dwell time are analyzed using the scene-driven sequence to identify interaction deviation nodes. Based on the interaction deviation nodes, dynamic adjustment of lighting effects is triggered to generate experience optimization strategies. The parallelism of the scene-driven sequence and the experience optimization strategy is detected and the synchronous rendering stage is identified. The synchronous rendering stage and the responsive action sequence are cross-configured to generate an effect reuse mode. Based on the effect reuse mode, the collaborative response group is decomposed into light and shadow and outputs spatial interaction control commands.

[0006] A second aspect of this invention proposes an immersive light and shadow art space interaction system based on motion recognition, comprising: The signal acquisition module is used to acquire user body sensation signals, perform signal analysis to generate body posture feature data, compare the body posture feature data with a posture template library to determine the action recognition range, and filter valid body sensation inputs based on the action recognition range to establish a usable posture dataset; The motion analysis module is used to analyze motion trajectories and identify motion rhythm distribution based on the available posture dataset, extract emotion-related features from the motion rhythm distribution to generate responsive motion sequences, and decompose composite postures in the body feature data into limb partitions to establish coordinated response groups. The lighting and shadow mapping module is used to establish a preliminary mapping table based on the responsive action sequence and the coordinated response group to adapt lighting and shadow effects, convert the body feature data into lighting and shadow configuration parameters, perform rendering verification on the preliminary mapping table using the lighting and shadow configuration parameters to generate an adaptation evaluation result, and initiate lighting and shadow projection to generate a scene-driven sequence based on the adaptation evaluation result; The experience optimization module is used to analyze the interaction frequency and dwell time using the scene-driven sequence to identify interaction deviation nodes, and to dynamically adjust the lighting effects based on the interaction deviation nodes to generate experience optimization strategies. The output control module is used to detect and identify the parallelism of the scene-driven sequence and the experience optimization strategy, identify the synchronous rendering stage, cross-configure the synchronous rendering stage and the responsive action sequence to generate an effect reuse mode, and output spatial interaction control commands to the collaborative response group by decomposing light and shadow according to the effect reuse mode.

[0007] The beneficial effects of this invention are reflected in the following points: 1. By analyzing the skeletal key points and comparing posture templates of the user's somatosensory signals, effective action inputs that conform to the interaction specifications are selected. On this basis, motion features such as motion amplitude and speed change rate are extracted for emotional tendency mapping, so that the light effect can reflect the emotional state contained in the user's actions. At the same time, for multi-limb synchronous motion, partition recognition and master-slave relationship analysis are performed to establish a cascading trigger mechanism that conforms to the natural movement law of the human body, so that the light and shadow effect presents a sense of hierarchy from the dominant limb to the subordinate limb. 2. Establish the gradient mapping relationship between motion intensity and light and shadow brightness, perform spatial coverage analysis on the projection area to eliminate overlapping conflicts and response blind spots, and verify the frame rate and load performance of each mapping item through rendering instantiation test. Based on the pass rate, an adaptability assessment is given and a degradation strategy is adopted for the non-compliant items to ensure that the light effect achieves a balance between visual effect and system performance. 3. By statistically analyzing the frequency of interaction triggers and the distribution of time spent in different areas, a user behavior heatmap is constructed to identify attention drift segments and determine the state of immersion disengagement. When user engagement decreases, differentiated lighting and shadow adjustment schemes are automatically triggered to guide users to re-engage in the interaction. At the same time, task preloading and effect reuse configurations are performed for high-parallel rendering periods to reduce performance fluctuations caused by resource competition and ensure the smoothness of the interaction process and the integrity of the experience. Attached Figure Description

[0008] The accompanying drawings illustrate specific examples of the technical solutions described in this invention and, together with the detailed embodiments, form part of the specification, serving to explain the technical solutions, principles, and effects of this invention.

[0009] Figure 1 This is a flowchart illustrating the immersive light and shadow art space interaction method based on motion recognition according to the present invention.

[0010] Figure 2 This is a structural block diagram of the immersive light and shadow art space interaction system based on motion recognition of the present invention. Detailed Implementation

[0011] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods are omitted so as not to obscure the description of this application with unnecessary detail.

[0012] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0013] References to "one embodiment" or "some embodiments" as used in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0014] The technical solutions of the embodiments of this application will be described below.

[0015] like Figure 1 As shown, this embodiment of the invention provides an immersive light and shadow art space interaction method based on motion recognition, including the following steps S110-S150: Step S110: Collect user body sensation signals, perform signal analysis to generate body posture feature data, compare the body posture feature data with the posture template library to determine the action recognition range, and filter valid body sensation inputs based on the action recognition range to establish an available posture dataset.

[0016] Specifically, user motion signals are collected and analyzed to generate body feature data. A depth camera array and infrared motion sensors are deployed within the immersive light and shadow art space. The depth cameras, employing Time-of-Flight (TOF) ranging technology, are positioned at five points—the four corners and the top of the interactive area—creating a comprehensive 3D perception zone without blind spots. The infrared motion sensors are equidistantly distributed along the boundaries of the interactive area, capturing the thermal radiation contours and movement directions of the user's limbs. The depth cameras collect user motion signals in real-time at a frequency of 30 frames per second, acquiring a 3D spatial coordinate point cloud of the user's body surface, with a point cloud density of 1200 sampling points per square meter. The collected motion signals are preprocessed, using Gaussian filtering to eliminate random noise from the sensors, and background subtraction to separate the user's contour from the static environment, filtering out ambient light interference and background clutter to extract the effective signal area of ​​the user's body. Human skeletal key points are analyzed from preprocessed somatosensory signals. A deep learning skeletal detection algorithm is used to locate these key points, which include 18 nodes: head, neck, left and right shoulders, left and right elbows, left and right wrists, left and right hips, left and right knees, and left and right ankles. Connection vectors and joint angles between adjacent skeletal key points are calculated, and limb extension range, trunk tilt angle, center of gravity coordinates, and limb movement velocity are extracted. The skeletal key point coordinates, joint angles, limb vectors, and movement velocities are integrated and encapsulated frame by frame to generate posture feature data. This data is organized using timestamps as indexes to represent the pose parameters of each frame, supporting temporal tracking and analysis of continuous movements.

[0017] The body posture feature data is compared with the posture template library to determine the action recognition range. A preset posture template library is loaded, storing standard action templates suitable for light and shadow art interaction. The template types cover five categories and 48 basic postures: waving, stretching, rotating, jumping, and static postures. Each posture template includes standard skeletal coordinates, joint angle ranges, and typical motion trajectory curves. The skeletal keypoint coordinates and joint angles of the current frame are extracted from the body posture feature data. The posture parameters of each frame of the body posture feature data are traversed, and the similarity is calculated one by one with the feature parameters of each template in the posture template library. The similarity calculation uses weighted Euclidean distance. First, the keypoint position distance and joint angle distance are normalized and mapped to the range of 0-1. Then, the normalized position distance and angle distance are weighted and summed. The weight of the keypoint position is set to 0.6, and the weight of the joint angle is set to 0.4. The smaller the weighted distance, the higher the similarity. The entirety of the 48 pose templates in the template library is traversed. The weighted Euclidean distance is converted into a similarity score ranging from 0 to 1. Templates with a similarity score higher than the 0.7 threshold are selected, and the corresponding action types constitute the initial matching range. Based on the similarity ranking of each action type within the initial matching range, the top 5 action types with the highest similarity are retained as a candidate set. The temporal continuity of the body feature data is used to verify the rationality of the motion trajectories of each action in the candidate set. The displacement of key points between adjacent frames is checked to see if it exceeds the physiological limits of motion. Candidate actions with abnormal trajectory jumps are eliminated, ultimately determining the action recognition range. The action recognition range is then sorted by confidence level in descending order of candidate action types, and each action is labeled with its corresponding template number and matching score.

[0018] A usable posture dataset is established by filtering valid somatosensory inputs based on the action recognition range. The confidence score of each action type within the action recognition range is read, and a lower limit of 0.75 is set for the confidence score of valid inputs. Action types with confidence scores below this threshold are considered uncertain and excluded from the valid input range. Action types with confidence scores above the lower limit are selected from the action recognition range, and the corresponding somatosensory inputs are marked as valid inputs. The frame number and action type code of the valid inputs are recorded. Time windows are aggregated for the valid inputs, with a time window length of 5 frames. Input sequences that maintain the same action type for 5 or more consecutive frames are considered stable postures, while actions appearing in a single frame or briefly are considered noise interference and filtered out. Action recognition range entries corresponding to stable postures are extracted, and the action type code, start and end timestamps, average confidence score, and duration (number of frames) are recorded. Transient actions lasting less than 200 milliseconds and unstable actions with confidence fluctuations exceeding 0.2 are removed. Transient actions are usually caused by unconscious limb movements, and unstable actions indicate ambiguity in the recognition results. The selected valid poses are arranged in chronological order, and a snapshot of the skeletal features of each pose is attached to create a usable pose dataset. The usable pose dataset is organized into a data structure based on action events, and each event fully records the pose evolution process from the start to the end of the action.

[0019] Step S120: Based on the available posture dataset, perform motion trajectory analysis to identify the action rhythm distribution, extract emotion-related features from the action rhythm distribution to generate a responsive action sequence, and decompose the composite postures in the body feature data into limb partitions to establish a collaborative response group.

[0020] Specifically, motion trajectory analysis is performed based on the available posture dataset to identify the distribution of action rhythm. Each effective posture in the available posture dataset carries a snapshot of skeletal features and time interval information, from which the start and end times and duration of each action event can be extracted. Adjacent action events in the available posture dataset are concatenated in chronological order, and the interval duration and switching frequency between actions are calculated to identify the user's action execution rhythm characteristics. The skeletal trajectories of each action event in the available posture dataset are smoothly fitted, and cubic Bézier curves are used to approximate the motion paths of key points to eliminate high-frequency jitter interference in the trajectory. The instantaneous velocity and acceleration at each moment are calculated along the fitted motion trajectory. The velocity is obtained from the displacement difference between adjacent sampling points, and the acceleration is obtained from the time difference of the velocity. The density distribution of action events throughout the entire interaction cycle is statistically analyzed. The time axis is divided into several equal-length intervals, and the number of action events and the average duration of each interval are counted. Dense and sparse action intervals are identified. Dense intervals correspond to periods of active user interaction, while sparse intervals correspond to periods of user pause or observation. The time distribution, velocity curve, and acceleration curve of action events are integrated to generate the action rhythm distribution. If a user performs 8 waving gestures within a 30-second interaction period, and completes 6 gestures in the first 15 seconds with an average interval of 1.5 seconds, but only completes 2 gestures in the last 15 seconds with an interval of more than 5 seconds, then the distribution of the gesture rhythm shows a characteristic of being denser at the beginning and sparser at the end, reflecting the process of the user's enthusiasm for interaction gradually declining from its peak.

[0021] In some embodiments, the step of extracting emotional association features from the action rhythm distribution to generate a responsive action sequence includes: extracting action amplitude and speed change rate from the action rhythm distribution; generating emotion labels by mapping emotion tendencies based on the action amplitude and speed change rate; establishing a priority response queue by sorting the action rhythm distribution by emotional response urgency using the emotion labels; and organizing a responsive action sequence according to the priority response queue.

[0022] The amplitude of movement and the rate of change of speed are extracted from the movement rhythm distribution. The skeletal trajectory data of each movement event in the movement rhythm distribution are traversed, and the maximum displacement distance of key points relative to their initial positions in each movement event is calculated. This maximum displacement distance is defined as the amplitude of movement. For movement events involving multiple key points, the weighted average of the maximum displacement distances of all moving key points is taken as the overall amplitude of movement, with weights allocated according to the dominance of the limb to which the key point belongs. The speed curve of the movement rhythm distribution contains the speed time series of each movement event. By traversing the speed data of each event in the movement rhythm distribution and calculating the difference between speed values ​​at adjacent moments, the instantaneous change in speed can be obtained. The rate of change of speed a = (v_t - v_(t-1)) / Δt, where v_t is the speed at the current moment, v_(t-1) is the speed at the previous moment, and Δt is the sampling interval. The maximum, minimum, and average values ​​of the rate of change of speed for each movement event within the execution cycle are statistically analyzed to identify the explosive and sustained characteristics of the movement. When a user performs a rapid waving motion, the amplitude of the movement can reach 0.8 meters, and the peak rate of change of speed reaches 15 meters per second. 2 The arm accelerates rapidly from a standstill and then stops abruptly, exhibiting typical explosive movement characteristics; when the user performs a slow extension movement, the range of motion is approximately 0.6 meters and the peak rate of change of speed is only 3 meters per second. 2 The arms extend and retract at a constant speed, exhibiting a smooth and continuous motion characteristic. The two types of movements are clearly distinguished in terms of the range of motion and rate of change of speed.

[0023] For example, the step of generating emotion tags by mapping emotion tendencies based on the amplitude of the action and the rate of change of speed includes: establishing an amplitude level sequence by performing energy decay gradient classification on the amplitude of the action; extracting change inertia features from the rate of change of speed to generate a rate feature identifier; fusing the amplitude level sequence and the rate feature identifier with emotion transfer paths to construct an emotion feature combination; and generating emotion tags by querying emotion mapping rules through the emotion feature combination.

[0024] An amplitude level sequence is established by classifying the amplitude of movements using energy decay gradient grading. The numerical sequence of amplitude values ​​for each movement event reflects the change in extension during the movement execution process. Calculating the difference in amplitude between adjacent moments yields the instantaneous change in amplitude. A negative change in amplitude indicates the movement is in the contraction phase, while a positive change indicates the movement is in the extension phase. Dividing the change in amplitude by the current amplitude value yields the energy decay gradient, which characterizes the relative rate of energy decay. Based on the numerical range of the energy decay gradient, the amplitude of movements is divided into five levels: a gradient close to zero corresponds to a stable amplitude level; a large positive gradient corresponds to a rapidly increasing amplitude level; a large negative gradient corresponds to a rapidly decreasing amplitude level; a moderately positive gradient corresponds to a slowly increasing amplitude level; and a moderately negative gradient corresponds to a slowly decreasing amplitude level. During the process of a user's arm moving from full retraction to full extension, the amplitude increases from 0.1 meters to 0.7 meters, with the energy decay gradient remaining consistently positive. The amplitude level sequence is recorded as three levels: slowly increasing, rapidly increasing, and stable. During the process of the user's arm rapidly retracting from full extension, the range of motion decreases from 0.7 meters to 0.15 meters, the energy decay gradient remains negative, and the amplitude level sequence is recorded as two levels: rapid decay and slow decay, which characterizes the energy release process of the movement.

[0025] The inertia feature of the velocity change rate is extracted to generate rate feature identifiers. The time series of velocity change rates for each action event exhibits different fluctuation characteristics and persistence features. The cumulative change is obtained by summing the velocity change rate time series; the cumulative change reflects the degree of inertia in velocity change. A large cumulative change indicates continuous acceleration or deceleration, while a small cumulative change indicates frequent and repetitive velocity changes. Peak and trough points in the velocity change rate time series are identified, and the time interval and amplitude difference between peaks and troughs are statistically analyzed. Short peak-trough intervals and large amplitude differences correspond to explosive inertia, while long peak-trough intervals and small amplitude differences correspond to gradual inertia. Based on the cumulative change and peak-trough characteristics, the inertia is classified into four types: explosive, gradual, oscillating, and steady, and corresponding rate feature identifiers are generated for each. When a user performs a sudden punch, the velocity change rate jumps from 0 to a peak and then falls back within 0.1 seconds, with extremely short peak-trough intervals and an amplitude difference of 20 meters per second. 2 The rate characteristic is identified as explosive; when the user performs the Tai Chi push palm movement, the rate of change of speed fluctuates slowly within 2 seconds, with long intervals between peaks and troughs and a drop of only 5 meters per second. 2 The rate characteristic is identified as progressive; when the user performs a hesitant and tentative action, the rate of change of speed fluctuates repeatedly, and the rate characteristic is identified as oscillating. The inertial characteristics of the three actions form a sharp contrast.

[0026] An emotion feature combination is constructed by fusing amplitude level sequences and rate feature identifiers to construct an emotion transfer path. A temporal correspondence exists between the amplitude level sequences and rate feature identifiers. The amplitude level labels at each moment in the amplitude level sequence are traversed, and the moments when amplitude levels change abruptly are identified and marked as emotion transfer nodes. These emotion transfer nodes correspond to the moments when the user's interaction state changes. The rate feature identifier corresponding to each emotion transfer node is extracted. The direction of amplitude level changes in the amplitude level sequence is combined with the inertia type of the rate feature identifier to form an emotion transfer path description. An amplitude change from low to high level, combined with an explosive rate feature identifier, describes an emotion arousal path; an amplitude change from high to low level, combined with a gradual rate feature identifier, describes an emotion decay path; and a stable amplitude level, combined with a steady rate feature identifier, describes an emotion maintenance path. The emotion transfer paths of each action event are concatenated chronologically to generate an emotion feature combination, which depicts the trajectory of the user's emotional evolution during the interaction process. The user's action sequence of slowly raising their hand and then suddenly waving it at an accelerated speed is recorded as a migration path of "gradual rise → explosive activation" in terms of emotional characteristics, reflecting the complete process of emotion from brewing to release; the user's action sequence of quickly raising their hand and then slowly lowering it is recorded as a migration path of "explosive activation → gradual decline" in terms of emotional characteristics, reflecting the process of emotion from heightened to calm.

[0027] Emotional tags are generated by querying emotion mapping rules through combinations of emotional features. A pre-defined emotion mapping rule base stores the correspondence between emotion transfer path patterns and emotion types, with rule entries covering single-node, two-node, and multi-node sequence patterns. The number of nodes and path type sequence in the emotion transfer path of the emotion feature combination are matched with the pattern entries in the rule base. Emotional feature combinations containing emotion arousal paths and high node density are matched to output excited emotions; emotional feature combinations containing emotion extinction paths and sparse nodes are matched to output calm emotions; and emotional feature combinations containing emotion maintenance paths and evenly distributed nodes are matched to output focused emotions. The matching degree score between the emotional feature combination and each candidate rule entry is calculated, and the rule entry with the highest matching degree is selected to determine the emotion type and output the emotion tag. When a user performs three rapid hand-waving actions in a light and shadow art space, the emotional feature combination is a three-node sequence of "explosive activation → explosive activation → stable maintenance," which has the highest matching degree with the "stable after continuous activation" pattern in the rule base. The output emotional label is pleasant with a confidence level of 0.85. When a user performs only one slow stretch and then remains still, the emotional feature combination is a two-node sequence of "gradual rise → stable maintenance," and the output emotional label is calm with a confidence level of 0.72.

[0028] An emotional response urgency ranking system is established based on emotion tags used to rank the action rhythm distribution. Each action event is accompanied by an emotion tag containing an emotion type and a confidence score. Different emotion types correspond to different intensity weights: excitement is weighted at 1.0, pleasure at 0.8, focus at 0.6, calm at 0.4, and hesitation at 0.3. The emotional response urgency score is calculated by multiplying the confidence score of the emotion tag by the emotional intensity weight, using the formula U = C × W, where U is the urgency score, C is the confidence score, and W is the emotional intensity weight. All action events in the action rhythm distribution are traversed, and the timestamps and action types of each event are extracted. The urgency score for each event is calculated using the emotion tags, and the events are ranked from highest to lowest based on their emotional response urgency scores. An entry threshold of 0.5 is set for the urgency score; action events below this threshold are categorized as delayed responses and not included in the priority response queue, while those above the threshold are entered into the priority response queue in order of their scores. Of the eight action events identified by a user within a 30-second interaction period, three events labeled with the emotion of excitement and with an urgency level of 0.85 are prioritized and enter the first priority response queue. Two events labeled with the emotion of pleasure and with an urgency level of 0.64 are then added to the queue. Two events labeled with the emotion of calm and with an urgency level below 0.4 are not added to the queue, ensuring that interactions with strong emotions receive priority light and shadow feedback.

[0029] A sequence of responsive actions is formed based on a priority response queue. The action event at the head of the priority response queue is processed first. Next, the time interval between adjacent action events in the priority response queue is checked to see if it meets the minimum response period constraint, which is set to 100 milliseconds. Adjacent action events with time intervals shorter than the minimum response period are discarded; events with higher urgency scores are retained, while lower-urgency events are delayed or merged. Conflict detection is performed on the action events in the priority response queue to identify situations where there is competition for light and shadow response resources. Resource competition refers to multiple action events needing to simultaneously drive the same light and shadow area. Action events with resource competition are arbitrated according to their urgency scores; higher-urgency events gain priority use of the resource, while lower-urgency events enter a waiting state. For example, a user simultaneously performs two actions: waving their left hand and waving their right hand. Both events point to the central light and shadow area of ​​the canopy, creating resource competition. The left-hand waving event (0.88 urgency score) is higher than the right-hand waving event (0.72 urgency score), so the left-hand waving event gains priority, and the right-hand waving event is delayed by 50 milliseconds. Action events that are re-arranged according to their execution sequence through conflict detection and resource arbitration are used to form a responsive action sequence. The order of each action event in the responsive action sequence takes into account both the urgency of emotional response and the availability of lighting resources. The triggering order, response area and effect parameters of each event are recorded using timestamps as an index.

[0030] In some embodiments, the step of decomposing the composite postures in the body feature data into limb partitions to establish a collaborative response group includes: identifying multi-limb linkage actions from the composite postures and extracting a set of linked limbs; performing motion correlation analysis on the set of linked limbs to identify the dominant limb and the subordinate limb; establishing a delay compensation trigger chain based on the response timing difference between the dominant limb and the subordinate limb; and organizing limb response units into a collaborative response group according to the delay compensation trigger chain.

[0031] This method identifies multi-limb coordinated movements from complex postures and extracts a set of coordinated limbs. In the sequence of motion events from body feature data, events where three or more limb regions experience significant displacement simultaneously are classified as complex postures. After statistically analyzing the skeletal keypoint motion data of each complex posture event according to anatomical regions, the amplitude and direction of movement for each limb can be obtained. The limb regions include six areas: head, trunk, left arm, right arm, left leg, and right leg. The ratio of the direction angle and amplitude between the motion vectors of each limb region is calculated. Limb regions with a direction angle less than 30 degrees and an amplitude ratio between 0.5 and 2.0 are identified as coordinated limbs. Coordinated limb identification is performed on each complex posture event, and the limb region numbers that meet the coordination criteria are recorded in a set to form a coordinated limb set. When a user performs a jumping jack in the light and shadow art space, the angle between the movement directions of the left and right arm areas is 25 degrees, and the amplitude ratio is 1.1, indicating a linkage relationship. The angle between the movement directions of the left and right leg areas is 20 degrees, and the amplitude ratio is 0.95, also indicating a linkage relationship. The linkage limb sets are recorded as two groups: {left arm area, right arm area} and {left leg area, right leg area}. When the user performs a single-arm waving motion, only the right arm area shows significant displacement, which does not meet the composite posture determination conditions. The linkage limb set is empty, and this action is processed as a single-limb mode.

[0032] Motion correlation analysis is performed on the linked limb set to identify dominant and subordinate limbs. The motion time series of each limb partition in the linked limb set includes start and peak time information. After sorting by motion start time, the limb partition with the earliest start time is marked as a candidate dominant limb. The motion correlation coefficient between the candidate dominant limb and other limb partitions in the linked limb set is calculated using Pearson correlation, with the displacement time series of each limb partition in the linked limb set as input. Partitions with a correlation coefficient higher than 0.7 and whose motion start time lags behind that of the candidate dominant limb are identified as subordinate limbs of the candidate dominant limb. When the user performs a double-arm extension action, the start time of the right arm partition is t=0.5 seconds, and the start time of the left arm partition is t=0.65 seconds. The correlation coefficient between the two is 0.92, so the right arm partition is identified as the dominant limb, and the left arm partition is identified as the subordinate limb. The time series difference of 0.15 seconds will be used as the delay parameter for the light effect response. When a user performs a torso-driven arm rotation movement, the starting time for the torso area is t=0.3 seconds, the starting time for the right arm area is t=0.42 seconds, and the starting time for the left arm area is t=0.48 seconds. The torso area is identified as the dominant limb, while the right and left arm areas are both identified as subordinate limbs, forming a hierarchical relationship of one dominant and two subordinate limbs.

[0033] A delay-compensated trigger chain is established based on the response timing difference between the dominant and subordinate limbs. The difference in the start time of movement between the dominant and subordinate limbs in each pair is defined as the response timing delay, which reflects the time difference in limb coordination during natural human movement. The time taken for the dominant limb to reach its peak displacement from the start of movement is statistically analyzed and compared with the response timing delay to determine whether the movement of the subordinate limb is initiated within the movement cycle of the dominant limb. The moment when the dominant limb triggers the photosensitive response is recorded as the reference moment, and the moment when the subordinate limb triggers the photosensitive response is set to the reference moment plus the response timing delay. The trigger delay parameters of each pair are connected in series according to the master-slave relationship to form a delay-compensated trigger chain. When there are multiple levels of subordinate relationships, the delay-compensated trigger chain is cascaded, with the trigger delay of the second-level subordinate limbs accumulated based on the first-level subordinate limbs. When the torso area is the primary dominant limb, the right arm area is the primary subordinate limb with a delay of 0.1 seconds, and the right wrist area is the secondary subordinate limb with a delay of 0.08 seconds, the delay compensation trigger chain is recorded as follows: torso area (0 seconds) → right arm area (0.1 seconds) → right wrist area (0.18 seconds). The light effect should be activated sequentially in this order to create a visual flow effect that spreads from the core of the body to the extremities.

[0034] Based on the delay compensation trigger chain, limb response units are organized into collaborative response groups. Each level of the delay compensation trigger chain is traversed, creating an independent response unit for each limb partition. The response unit records the limb partition number and trigger delay duration, and automatically associates the light and shadow response area number according to a preset limb-light and shadow area correspondence rule: the torso corresponds to the central area of ​​the canopy, the left and right arm areas correspond to the areas on either side of the canopy, the left and right leg areas correspond to the ground area, and the head area corresponds to the top area. The response effect type is determined according to the hierarchical role of the limb partition; the dominant limb uses a diffusion effect, and subordinate limbs use a follow-up effect. Response units with a direct master-slave relationship in the delay compensation trigger chain establish an association reference, and the trigger signal of the dominant limb response unit is transmitted to the subordinate limb response units to achieve cascade activation. All response units belonging to the same composite posture event are grouped and encapsulated. Response units in the same group share the action type code and overall confidence score, and after encapsulation, they are integrated according to the master-slave hierarchical relationship to establish collaborative response groups. When a user performs a jumping jack, the coordinated response group contains four response units: the torso unit is the trigger root node, and the arm and leg units are child nodes that are respectively associated with the light and shadow areas on both sides and the ground. The cascade triggering mechanism ensures that the changes in light and shadow are synchronized with the rhythm of human movement.

[0035] Step S130: Based on the responsive action sequence and the collaborative response group, a preliminary mapping table is established to adapt the lighting and shadow effects. The body feature data is converted into lighting and shadow configuration parameters. The preliminary mapping table is rendered and verified using the lighting and shadow configuration parameters to generate an adaptation evaluation result. Based on the adaptation evaluation result, the lighting and shadow projection is initiated to generate a scene-driven sequence.

[0036] In some embodiments, the step of establishing a preliminary mapping table for light and shadow effect adaptation based on the responsive action sequence and the cooperative response group includes: extracting action intensity gradient features from the responsive action sequence to establish intensity-brightness candidate associations; performing spatial coverage analysis on the cooperative response group to identify projection coverage aberration areas; performing conflict resolution on the intensity-brightness candidate associations based on the projection coverage aberration areas to generate a conflict-free mapping scheme; and constructing a preliminary mapping table based on the conflict-free mapping scheme.

[0037] Action intensity gradient features are extracted from responsive action sequences to establish intensity-luminance candidate associations. Each action event in the responsive action sequence is traversed, and the action amplitude and velocity change rate of each event are used to calculate the overall action intensity. The intensity calculation formula is I = 0.6 × A_norm + 0.4 × V_norm, where A_norm is the normalized action amplitude, V_norm is the normalized velocity change rate, and I is the action intensity value ranging from 0 to 1. The intensity difference between adjacent action events in the responsive action sequence reflects the gradient change characteristics of action intensity; a positive gradient corresponds to a gradually increasing action intensity, while a negative gradient corresponds to a gradually decreasing action intensity. The peak point of the intensity gradient corresponds to the moment of action energy burst, and the trough point corresponds to the moment of action energy drop. The candidate mapping relationship between action intensity level and light / shadow brightness level adopts a same-level correspondence strategy. Both intensity and brightness are divided into 5 levels: intensity level 1 corresponds to brightness level 1, intensity level 5 corresponds to brightness level 5, and the correspondence between each intensity level and brightness level is recorded as an intensity-luminance candidate association. In the responsive action sequence, a fast punch with an intensity value of 0.85 corresponds to level 4, and the intensity-brightness candidate association is configured with a brightness level of 4, which is 80% of the light and shadow output power; a slow stretching action with an intensity value of 0.35 corresponds to level 2, and the intensity-brightness candidate association is configured with a brightness level of 2, which is 40% of the light and shadow output power, ensuring that the force of the action and the intensity of the light and shadow form an intuitive perceptual correspondence.

[0038] Spatial coverage analysis is performed on the coordinated response group to identify projection coverage anomalies. The light and shadow projection areas associated with each limb response unit in the coordinated response group have physical spatial coverage coordinates. The projection area boundary parameters of each unit are extracted by traversing all response units in the coordinated response group. The projection coverage areas of all response units in the coordinated response group are plotted in a unified spatial coordinate system, and the overlap between coverage areas is analyzed. Significant overlap is defined as the ratio of the overlapping area of ​​any two projection areas to the smaller area exceeding 30%. Significantly overlapping projection areas may cause light and shadow overlay conflicts during linkage triggering, resulting in overexposure or color confusion. Coverage blind spots may also exist in the coordinated response group. Blind spots refer to spatial areas within the user's activity range that are not covered by any projection area; actions within blind spots will not receive light and shadow response feedback. Overlapping areas are labeled with the anomaly type "conflict," and coverage blind spots are labeled with the anomaly type "missing." Both types of areas are uniformly marked as projection coverage anomalies. When the user performs a jumping jack action, the right arm response unit and the left arm response unit are simultaneously associated with the top light band area and the coverage overlap rate reaches 75%. This area is identified as a projection coverage anomaly area and marked as a conflict type. There is a 1.5 square meter projection blind zone to the right rear of the user's activity area. This area is identified as a projection coverage anomaly area and marked as a missing type. Both types of anomalies need to be resolved in subsequent steps.

[0039] Based on the projection coverage anomaly area, conflict resolution is performed on intensity-luminance candidate associations to generate a conflict-free mapping scheme. The type label of each anomaly region in the projection coverage anomaly area determines the resolution strategy, with conflict-type and missing anomalies handled separately. For conflict-type projection coverage anomaly areas, the mapping entries of all response units involved in the intensity-luminance candidate associations need to be prioritized. The entries related to the conflict area in the intensity-luminance candidate associations are traversed, and the mapping entry with the highest intensity level or the highest response priority is retained as the primary mapping for that area. The remaining intensity-luminance candidate association entries are adjusted to alternative mappings and are only activated when the primary mapping is idle. Missing projection coverage anomaly areas are filled by expanding the projection coverage range of intensity-luminance candidate association entries near the blind zone boundary, with the expansion aiming to cover 80% of the blind zone area. After conflict resolution, it is checked whether there is still overlap in the projection areas of each mapping entry. A residual overlap rate of less than 10% is considered acceptable. The conflict-resolved mapping entries and the expanded coverage mapping entries are integrated to generate a conflict-free mapping scheme. The conflict-free mapping scheme ensures that only one mapping rule is activated in the same projection area at the same time, and there are no response blind zones within the user's activity range. After the conflict in the top light strip area is resolved, the right arm swing action is the primary mapping and is triggered first, while the left arm swing action is the alternative mapping and is triggered after a delay of 100 milliseconds. The timing constraint condition is recorded for the conflict-free mapping scheme.

[0040] A preliminary mapping table is constructed based on a conflict-free mapping scheme. The correspondence between each action event and the corresponding light effect region in the conflict-free mapping scheme includes the action type code, response region number, and trigger condition parameters. All mapping entries in the conflict-free mapping scheme are traversed, and a preliminary mapping table record is created for each entry. The action input information of the record includes the action type, trigger threshold, and effective time window, while the light effect output information includes the target projection area, light effect type, and response intensity level. Entries with the same action type in the conflict-free mapping scheme are aggregated. The same action type may correspond to multiple light effect regions. After aggregation, a one-to-many mapping relationship is formed and written into the preliminary mapping table. The aggregated mapping relationship uses the action type code as the primary key to establish a fast retrieval structure, supporting the rapid location of all associated light effect regions based on the input action. Each entry in the preliminary mapping table must pass an integrity check to verify that each action type is associated with at least one valid light effect region and that the response parameters of each light effect region are within the device's supported range. The waving action is associated with two response targets in the preliminary mapping table: the top light strip area and the side wall projection area. The jumping action is associated with two response targets in the preliminary mapping table: the ground light halo area and the ambient light area. The rotating action is associated with the surrounding beam area in the preliminary mapping table. Different action types drive different spatial areas of light effects.

[0041] Body posture data is converted into lighting and shadow configuration parameters. The real-time coordinates and movement speed of skeletal key points in the body posture data can calculate the user's current spatial position and orientation angle, with position coordinate accuracy down to the centimeter level and orientation angle accuracy down to 5 degrees. The body contour dimensions determined by the distribution range of key points in the body posture data are used to scale the coverage area of ​​the lighting projection to adapt to users of different heights. For example, when a user 1.8 meters tall triggers a hand gesture, the lighting coverage radius is set to 1.2 meters; when a user 1.5 meters tall triggers the same gesture, the lighting coverage radius is scaled down to 1.0 meter, ensuring that tall and short users receive proportionally coordinated lighting effects. The numerical range of joint angles in the body posture data reflects the user's movement amplitude level; large movements correspond to high-intensity lighting output, and small movements correspond to low-intensity lighting output. The temporal changes in the body posture data include movement rhythm characteristics; fast-paced periods are configured with rapid flashing or flowing lighting effects, while slow-paced periods are configured with slow gradation or breathing lighting effects. The spatial distance and angular relationship between the user's position and each lighting projection device determines the pointing parameters and focal length parameters of the projection device. The user's position coordinates are mapped to the center point coordinates of the projection area, the movement amplitude is mapped to the light and shadow brightness level, and the movement speed is mapped to the light and shadow change frequency. After integration, the light and shadow configuration parameters are output. The light and shadow configuration parameters include five types of parameters: projection center coordinates, coverage radius, brightness level, change frequency, and color temperature mode, which fully describe the rendering requirements of a single light and shadow effect.

[0042] In some embodiments, the step of generating a compatibility evaluation result by rendering the preliminary mapping table using the lighting and shadow configuration parameters includes: instantiating the preliminary mapping table to construct a rendering test frame based on the lighting and shadow configuration parameters; performing frame rate and rendering load detection on the rendering test frame to generate a performance index set; performing deviation analysis between the performance index set and a preset performance benchmark to obtain the rendering pass rate; and determining the compatibility evaluation result based on the rendering pass rate.

[0043] The initial mapping table is instantiated based on the lighting configuration parameters to construct rendering test frames. The mapping entries in the initial mapping table, combined with the projection position, brightness level, and change frequency parameters in the lighting configuration parameters, convert abstract mapping rules into specific rendering instructions. Each mapping entry in the initial mapping table undergoes parameter filling; the brightness level value from the lighting configuration parameters is assigned to the light effect intensity field of the mapping entry, and the change frequency value is assigned to the light effect refresh rate field. The projection center coordinates and coverage radius in the lighting configuration parameters are used to calculate the rendering position and size of each light effect element. The light effect type encoding in the initial mapping table is parsed into specific rendering primitives: light strip type is parsed into rectangular gradient primitives, light ring type into circular diffusion primitives, and light beam type into conical volumetric light primitives. The position, size, color, and animation parameters of all rendering primitives are combined to generate a complete single-frame rendering description. When generating multi-frame continuous rendering descriptions according to the change frequency parameters, the primitive parameters between frames are interpolated according to the animation curve. The encapsulated rendering test frame sequence covers three test scenarios: static lighting effects, looping animations, and trigger responses. The rendering test frames cover typical application scenarios of all mapping entries in the initial mapping table. The rendering test frames for the waving action contain a 20-frame animation sequence of the top light band flowing from left to right, and the rendering test frames for the jumping action contain a 15-frame animation sequence of the ground halo spreading outwards.

[0044] A performance metric set is generated by detecting frame rate and rendering load on the rendering test frames. The rendering test frame sequence is submitted frame by frame to the rendering engine for actual rendering operations. The time difference between the rendering command issuance and the completion of the screen output for each frame is defined as the rendering time per frame. After statistically analyzing the rendering time sequence of 100 consecutive rendering test frames, the average time and standard deviation can be calculated. The average time reflects steady-state rendering capability, and the standard deviation reflects rendering stability. The actual frame rate is calculated from the average time using the formula F = 1000 / T_avg, where T_avg is the average rendering time (milliseconds) and F is the actual frame rate (frames / second). During the rendering test frames, hardware resource usage data is collected synchronously, and the monitored metrics include GPU memory usage, GPU core utilization, and CPU rendering thread load. High-load rendering test frames containing numerous particle effects and low-load rendering test frames containing only basic lighting effects are recorded to facilitate the identification of rendering bottlenecks. After integrating frame rate, latency statistics, peak resource usage, and bottleneck analysis, a performance metric set is output. This set includes five core metrics: actual frame rate, average latency, latency standard deviation, peak GPU utilization, and bottleneck type. The performance metric set for high-load scenarios shows an actual frame rate of 52 frames per second and a GPU utilization of 92%, while the performance metric set for low-load scenarios shows an actual frame rate of 75 frames per second and a GPU utilization of 45%. Comparing these two sets of data helps pinpoint performance bottlenecks.

[0045] The rendering pass rate is obtained by performing deviation analysis between the performance metric set and the preset performance benchmark. The preset performance benchmark parameters include the target frame rate lower limit, frame rate fluctuation tolerance, GPU utilization upper limit, and single-frame latency upper limit. The actual frame rate value in the performance metric set is compared with the target frame rate lower limit of 60 frames / second, which ensures visual smoothness in lighting and shadow changes. The percentage deviation between the actual frame rate and the target frame rate is calculated using the formula D = (F_actual - F_target) / F_target × 100%, where F_actual is the actual frame rate, F_target is the target frame rate lower limit, and D is the frame rate deviation percentage. A positive deviation indicates performance surplus, and a negative deviation indicates insufficient performance. The peak GPU utilization in the performance metric set is compared with the upper limit of 85%; test scenes exceeding the upper limit are marked as high-load risk. The standard deviation of latency in the performance metric set is compared with the frame rate fluctuation tolerance; an excessively large standard deviation indicates unstable rendering performance. The pass / fail status of each test scene is determined by a combination of three indicators: frame rate deviation, GPU usage, and latency fluctuation. Scenes that meet all three indicators are considered to have passed the benchmark. The ratio of the number of passing scenes to the total number of test scenes is the rendering pass / fail rate, calculated as: Rendering pass / fail rate = Number of passing scenes / Total number of scenes × 100%. A certain light and shadow art space contains 25 test scenes. Among them, 23 scenes meet all three indicators, and 2 scenes have excessive GPU usage, resulting in a rendering pass / fail rate of 92%.

[0046] The compatibility assessment results are determined based on the rendering pass rate. The rendering pass rate is compared with a preset compatibility assessment threshold system, using a three-tier threshold: above 95% is considered fully compatible, 85% to 95% is considered basically compatible, and below 85% is considered insufficiently compatible. The output status for a fully compatible level is "Pass," meaning all entries in the initial mapping table are ready for production without adjustment. The output status for a basically compatible level is "Conditionally Pass," with an appendix listing the item numbers of non-compliant scenes and the specific performance gap values ​​for each item. The output status for an insufficiently compatible level is "Fail," requiring a return to the mapping scheme stage for light effect complexity reduction or response density lowering. Common characteristics of non-compliant scenes in the test details corresponding to the rendering pass rate need to be extracted and analyzed to determine whether the performance bottleneck is concentrated in GPU computing, memory bandwidth, or CPU scheduling. A rendering pass rate of 91% is considered basic compatibility. If the seven scenes that fail to meet the standard all involve triggering more than three projection areas simultaneously, the compatibility assessment suggests limiting the maximum number of concurrent projection areas from five to three. A rendering pass rate of 78% is considered insufficient compatibility, and the compatibility assessment result will output a failing status, suggesting reducing the particle density by 50% and retesting.

[0047] Based on the adaptability assessment results, a scene-driven sequence for generating light and shadow projection is initiated. The light and shadow projection process begins when the adaptability assessment results indicate a pass or conditional pass. In the case of conditional pass, the mapping numbers of the non-compliant items in the adaptability assessment results are extracted, and a downgraded rendering strategy is applied to these items during projection, either halving the particle count or reducing the refresh rate to 30 frames per second. The usable mapping items confirmed by the adaptability assessment results are used to extract a sequence of light and shadow instructions to be executed according to the trigger timing. Each light and shadow instruction is sorted by execution time and assigned a unique sequence number and expected execution duration, with necessary transition intervals inserted between adjacent instructions. The rendering load distribution recorded in the adaptability assessment results guides the dilution of instruction density during high-load periods, avoiding a sudden drop in frame rate caused by executing multiple high-load instructions consecutively. The sorted and adjusted light and shadow instructions are encapsulated into executable driving units, outputting the scene-driven sequence. When a user performs a continuous waving and jumping motion, the scene-driven sequence sequentially includes three commands: at time T0, the top light strip moves to the right; at T0+200ms, the ground halo expands outwards; and at T0+350ms, the ambient light color temperature gradually changes from cool white to warm yellow. These three commands create a coherent narrative effect of light and shadow. When a user performs a rotating and extending arm motion, the scene-driven sequence includes three commands: at time T0, the central light pillar rotates; at T0+150ms, the side beams expand; and at T0+300ms, the ground halo rotates synchronously. The multiple commands in the scene-driven sequence work together to create a surround-like effect of light and shadow.

[0048] Step S140: Use scene-driven sequences to analyze interaction frequency and dwell time to identify interaction deviation nodes, and trigger dynamic adjustment of light and shadow effects based on interaction deviation nodes to generate experience optimization strategies.

[0049] In some embodiments, the step of using the scene-driven sequence to analyze interaction frequency and dwell time to identify interaction deviation nodes includes: generating a frequency distribution curve by counting the number of interaction triggers per unit time in the scene-driven sequence; performing regional dwell time analysis on the frequency distribution curve to obtain a behavior heatmap; identifying attention drift segments from the behavior heatmap and marking them as candidate deviation areas; and determining interaction deviation nodes by performing immersion de-locking on the candidate deviation areas.

[0050] A frequency distribution curve is generated by counting the number of interaction triggers per unit time in the scene-driven sequence. The complete time range covered by the scene-driven sequence is divided into continuous statistical windows, with a window length of 5 seconds. Adjacent windows have a 50% overlap rate to improve temporal resolution. The number of instructions contained in the scene-driven sequence within each window is the number of interaction triggers for that window. The trigger count for each window is counted by traversing all instructions in the scene-driven sequence, and a curve is plotted with the window center time as the x-axis and the trigger count as the y-axis. The original curve is filtered using a moving average to eliminate random fluctuations in a single window. The filter window width is set to 3 statistical windows. The overall mean and standard deviation of the filtered curve are used to standardize the frequency values ​​to a relative deviation factor. A deviation factor greater than 1.5 indicates abnormally active interaction at that moment, and less than 0.5 indicates abnormally low interaction. The trend of the frequency distribution curve intuitively reflects changes in the user's interaction state. A continuous rise in the frequency distribution curve corresponds to increased user interest, a continuous decline corresponds to decreased interest, and violent oscillations correspond to the user being in an exploratory and trial phase. During a 60-second interaction period, a frequency distribution curve indicating active interaction is maintained between 1.2 and 1.5 deviation factors for the first 30 seconds, while a deviation factor of 0.4 to 0.6 is observed for the next 30 seconds, indicating less active interaction. The output frequency distribution curve is stored as a time-series array, with each array element containing three fields: time, original frequency value, and standardized deviation factor.

[0051] A behavioral heatmap is obtained by analyzing the dwell time in different areas of the frequency distribution curve. The temporal data of the frequency distribution curve is combined with the projection area numbers associated with each instruction in the scene-driven sequence to establish a two-dimensional analysis framework of time and space. The light and shadow interaction space is divided into grid cells with a granularity of 0.5m × 0.5m. The data points at each moment of the frequency distribution curve are traversed, and the activation count of each grid cell within each time window is counted. The frequency value at the corresponding moment of the frequency distribution curve is used as the weight for weighted accumulation. The dwell time of a user in each grid cell is calculated by accumulating the continuous time when the user's position coordinates fall within that grid cell's range; a longer dwell time indicates a higher level of user attention to that area. The activation count, frequency weight, and dwell time are normalized and then fused into a heat value H using the formula H = 0.4 × N_norm + 0.3 × F_norm + 0.3 × T_norm, where N_norm is the normalized activation count, F_norm is the normalized frequency weight, and T_norm is the normalized dwell time. The heat values ​​of each grid cell are mapped to color levels to generate a behavior heatmap. This heatmap presents the distribution of attention in the interaction space from a top-down perspective. Highlighted red areas indicate frequent user interaction, while muted blue areas indicate less user attention. After a user performs multiple wave gestures in the center of the interaction area, the heatmap shows that the central 3×3 grid has a heat value above 0.85, appearing dark red, while the edge grids have a heat value of only around 0.2, appearing light blue. This spatial distribution of heat gradients reveals that user behavior preferences are concentrated in the central area.

[0052] Attention drift segments are identified and marked as candidate deviation areas from the behavior heatmap. After dividing the complete interaction cycle into multiple consecutive time segments, each segment corresponds to a local behavior heatmap. Comparing the behavior heatmaps of adjacent segments reveals the dynamic migration process of attention. By traversing the behavior heatmaps of each time segment, the center coordinates and peak heat of hotspot areas are extracted. Segments where the positional shift of the hotspot area exceeds 1.5 meters or the peak heat drops by more than 40% indicate a significant change in user attention during that period. When the heat of three or more consecutive segments is below 80% of the overall average, the corresponding time interval is determined as an attention drift segment. The migration of hotspots from the center to the edge of the interaction area in the behavior heatmap indicates a shift in user interest outwards; the dissipation of hotspots without the formation of new ones indicates an overall decline in user engagement. The start and end times and drift type of attention drift segments are recorded and marked as candidate deviation areas. Drift types are categorized into three types: outward migration, dissipation, and oscillation. During the period from 45 seconds to 60 seconds, the behavior heatmap showed that the hotspot moved 1.8 meters from the central area to the left edge. This period was marked as an outward-moving candidate deviation area. During the period from 80 seconds to 95 seconds, the behavior heatmap showed that the heat of the entire area dropped to below 0.3 and there were no obvious hotspots. This period was marked as a dissipating candidate deviation area. Different types correspond to different subsequent processing strategies.

[0053] For example, the step of determining the interaction deviation node by locking the immersion detachment in the candidate deviation area includes: determining an observation time window based on the candidate deviation area; generating a recovery quality index by counting the number of interaction recoverys and the recovery depth according to the observation time window; determining the immersion detachment state when the recovery quality index is lower than a preset recovery threshold; and locking the start time of the immersion detachment state as the interaction deviation node.

[0054] The observation time window is determined based on the candidate deviation zone. The duration D of the candidate deviation zone is calculated by subtracting the start time from the end time. A longer duration indicates a more persistent user deviation, requiring a wider observation range for analysis. The initial extension of the observation time window is set to 30% of the duration, with a minimum of 5 seconds, and the subsequent extension is set to 50% of the duration, with a minimum of 8 seconds. This dynamic adjustment mechanism ensures that both short-term and long-term deviations have appropriate observation ranges. The starting boundary is obtained by subtracting the initial extension from the start time of the candidate deviation zone, and the ending boundary is obtained by adding the subsequent extension to the end time of the candidate deviation zone. Boundary values ​​require out-of-bounds verification; the starting boundary must not be earlier than the start time of the interaction session, and the ending boundary must not be later than the current analysis time. The time range after successful verification is the observation time window. The observation time window defines the effective interval for evaluating changes in the user's immersion state, and the behavioral data within the window will be used to calculate recovery quality indicators. The candidate deviation zone lasts for 30 seconds, with an initial extension of 9 seconds and a subsequent extension of 15 seconds. The total observation time window spans 54 seconds, covering the period from the normal state before deviation to the potential recovery period after deviation.

[0055] Recovery quality metrics are generated based on the number of interaction recovery attempts and recovery depth within the observation time window. Light and shadow-driven commands within the observation time window are sorted by execution time to form a command sequence within the window. Frequency data from each statistical sub-window within the observation time window is traversed; a recovery attempt occurs when the interaction frequency in the sequence rises from a trough. A recovery is defined as a frequency increase of more than 50% compared to the previous window, with an absolute value exceeding 1.5 times per 10 seconds. The depth of each recovery, R_depth, is calculated as (F_peak - F_valley) / (F_baseline - F_valley), where F_peak is the peak frequency after recovery, F_valley is the frequency deviating from the trough, and F_baseline is the frequency deviating from the baseline. The recovery depth ranges from 0 to 1. The average depth of all effective recovery attempts within the observation time window is used as the comprehensive recovery depth; a higher comprehensive depth indicates a stronger ability for the user to regain immersion. The recovery quality index, Q = 0.4 × N_recovery / N_expected + 0.6 × R_depth, is derived by weighting and integrating the number of recoveries with the overall depth. N_expected represents the expected number of recoveries estimated based on the observation time window, and N_recovery represents the number of effective recoveries. The recovery quality index ranges from 0 to 1; a higher value indicates a greater likelihood that the user will autonomously recover from their immersive state.

[0056] A user is considered disengaged when their recovery quality index falls below a preset recovery threshold. The recovery threshold is set at 0.45, derived from extensive user interaction data. When the recovery quality index is below this value, the probability of a user spontaneously recovering from immersion is less than 30%. Candidate deviation zones with recovery quality indices above the threshold represent temporary attention fluctuations that users can recover from spontaneously; these zones are marked as non-disengagement and excluded from subsequent processing. A recovery quality index below or equal to the threshold indicates a substantial interruption of immersion, requiring external intervention for recovery. Disengagement severity is graded according to the recovery quality index value: 0.3 to 0.45 is mild disengagement, 0.15 to 0.3 is moderate disengagement, and below 0.15 is severe disengagement. Different severity levels correspond to different intensity adjustment strategies. The determination of immersion disengagement and its severity level are linked to the corresponding candidate deviation zone record for use in subsequent node locking processes. A user's recovery quality index is 0.28, which is lower than the threshold of 0.45. This indicates that the user is in a state of immersion detachment and the severity is moderate detachment. Mild detachment triggers ambient light fine-tuning, moderate detachment triggers directional lighting guidance, and severe detachment triggers all-around light and shadow awakening.

[0057] The starting moment of the immersion-disengagement state is identified as the interaction deviation node. The candidate deviation zone record for immersion-disengagement states includes the interval's starting moment, which serves as the initial time anchor for the deviation. During the immersion-disengagement state determination process, the various components of the recovery quality index are calculated. The moment when the frequency first significantly decreases is extracted as the basis for precision correction; this moment reflects the true starting point of the deviation better than the initial anchor. The corrected precise moment is marked as the interaction deviation node. The interaction deviation node record includes a millisecond-level timestamp, disengagement severity, recovery quality index value, and associated spatial region number. Multiple interaction deviation nodes in the same session are arranged chronologically, and the interval between adjacent nodes is checked. Adjacent interaction deviation nodes with an interval of less than 30 seconds are merged into a single deviation event to avoid frequent triggering of adjustment strategies. During merging, the earlier node's timestamp and the higher immersion-disengagement severity level are used to ensure that the adjustment strategy is triggered early enough and with sufficient intensity. For example, a user experiences two interaction deviation nodes at 52 seconds and 68 seconds, with an interval of 16 seconds (less than the 30-second threshold). These are merged into a single deviation event, with 52 seconds as the trigger moment, and the higher of the two (moderate disengagement level) is selected. After all interactive deviation nodes are processed and output, a deviation node list is formed. The list includes four fields for each node: timestamp, severity, recovery quality index, and spatial region.

[0058] A dynamic adjustment strategy for lighting effects triggered by interaction deviation nodes is generated to optimize the user experience. Interaction deviation node records include a severity level of deviation. Differentiated lighting adjustment schemes are configured based on the severity: for mild deviation, a subtle flowing light effect is added to the edge of the user's field of vision to draw attention; for moderate deviation, a gradient light spot is projected directly in front of the user, with a color temperature transitioning from cool to warm; for severe deviation, an all-around surround lighting effect is activated to wake the user up. The spatial area number in the interaction deviation node record indicates the user's focus direction when deviating, and the adjusted lighting effect is projected to the vicinity of that direction to naturally integrate into the user's line of sight. The duration and intensity progression curve of the lighting adjustment are set: 8 seconds for mild deviation, 12 seconds for moderate deviation, and 20 seconds for severe deviation. The severity of deviation, adjustment scheme configuration, and progression curve parameters are integrated to generate the experience optimization strategy. When a user experiences moderate disengagement, the adjustment scheme for the experience optimization strategy configuration is as follows: First, project a warm-colored gradient light spot in the user's attention direction for 4 seconds, then activate a slowly flowing light band for 5 seconds, and finally project ambient light with a breathing rhythm for 3 seconds. The three-stage progression forms a complete attention re-engagement process.

[0059] Step S150: Parallelism detection and identification of synchronous rendering links are performed on the scene-driven sequence and experience optimization strategy. The synchronous rendering links and responsive action sequences are cross-configured to generate effect reuse modes. Based on the effect reuse modes, the collaborative response group is decomposed into light and shadow to output spatial interaction control commands.

[0060] Specifically, parallelism detection is performed on scene-driven sequences and experience optimization strategies to identify synchronous rendering stages. Each lighting and shadow driving instruction in the scene-driven sequence and each adjustment scheme in the experience optimization strategy carries execution time and rendering resource requirement information. All instructions in the scene-driven sequence are traversed to extract the execution time window and GPU resource requirements of each instruction. Simultaneously, all adjustment schemes in the experience optimization strategy are traversed to extract the trigger time and rendering complexity parameters of each scheme. Arranging the two types of instructions on the same timeline according to their execution time allows observation of the temporal overlap of resource requirements. Instruction combinations with overlapping execution time windows exceeding 50 milliseconds are considered overlapping. Overlap detection covers three scenarios: instruction pairing within the scene-driven sequence, instruction pairing within the experience optimization strategy, and instruction pairing across sources. The cumulative rendering resource requirements of each overlapping instruction group include GPU computation, memory usage, and the number of rendering pipeline channels. Overlapping groups with cumulative requirements exceeding 80% of the single-frame rendering capacity are marked as high-parallelism groups. The execution period corresponding to the high-parallelism group is the synchronous rendering stage, which records the start and end times, the list of involved instruction numbers, and the resource contention level. Resource contention is categorized into critical and overload levels. Critical level refers to cumulative demand between 80% and 95%, while overload level refers to demand exceeding 95%. During a user's continuous hand-waving action, the scene-driven sequence's light ribbon flow instructions and the experience optimization strategy's attention-guided lighting effects overlap between the 15th and 18th seconds, with cumulative GPU demand reaching 92%. This period is identified as the critical level synchronous rendering phase.

[0061] In some embodiments, the step of cross-configuring the synchronous rendering phase with the responsive action sequence to generate an effect reuse mode includes: extracting the resource occupancy characteristics of each rendering task in the synchronous rendering phase; aligning and analyzing the resource occupancy characteristics with the trigger timing of the responsive action sequence to identify resource idle windows; using the resource idle windows to perform elastic preloading configuration of rendering tasks to generate a reuse schedule table; and establishing an effect reuse mode based on the reuse schedule table.

[0062] Resource consumption characteristics of each rendering task in the synchronous rendering process are extracted. In the list of rendering instructions involved in the synchronous rendering process, each instruction is associated with one or more rendering tasks. The resource consumption parameters of each task are extracted by traversing the instruction list of the synchronous rendering process. GPU computation is measured by the number of shader calls and texture sampling times; particle effect tasks typically have a computational cost 3 to 5 times that of basic lighting effect tasks. Memory usage includes frame buffer, texture mapping, and vertex data; the memory requirement for full-screen lighting effects is approximately 2 to 3 times that of local lighting effects. The number of rendering pipeline channels varies from 1 to 4; particle systems and volumetric lighting effects usually require dedicated channels to avoid rendering order errors. The execution time of each task is predicted based on complexity and historical data; the execution time of complex tasks can be more than 10 times that of simple tasks. The four indicators—GPU computation, memory usage, channel usage, and execution time—are integrated to form a resource consumption feature vector. In the synchronous rendering process, the resource usage characteristics of a particle burst light effect show that the GPU computation is 4.2 times that of the basic light effect, the video memory usage is 380MB, the channel usage is 2, and the execution time is 28 milliseconds; the resource usage characteristics of a gradient light strip show that the GPU computation is 1.1 times that of the basic light effect, the video memory usage is 85MB, the channel usage is 1, and the execution time is 6 milliseconds. The difference in resource usage characteristics between the two types of tasks provides a basis for subsequent task scheduling.

[0063] Resource occupancy characteristics are aligned with the trigger timing of responsive action sequences to identify resource idle windows. The trigger times of each action event in the responsive action sequence form a discrete set of trigger points on the timeline. The trigger timestamps and expected response durations of each event are extracted by traversing the responsive action sequence. Each rendering task is labeled with its occupancy period based on the execution duration of its resource occupancy characteristics, forming a task distribution map. The resource load rate at each moment is calculated based on the GPU computation and memory usage of the resource occupancy characteristics. After aligning the trigger timing with the task distribution, the GPU load rate, memory usage, and channel occupancy rate at each moment can be calculated. Time intervals where all three indicators are below 60% are considered resource idle periods, and periods with continuous idle durations exceeding 20 milliseconds are marked as resource idle windows. The positional relationship between a resource idle window and adjacent trigger times in the responsive action sequence determines its utilization value; resource idle windows closer to their trigger times are more suitable for preloading tasks about to be executed. The start and end times, duration, available resource balance, and adjacent trigger distances of each resource idle window are recorded and output. Three resource idle windows were identified within a certain interaction cycle, located at 8-12 seconds (lasting 40 milliseconds, GPU availability 55%), 22-25 seconds (lasting 30 milliseconds, GPU availability 48%), and 35-38 seconds (lasting 35 milliseconds, GPU availability 52%). These three windows can be used to preload high-load rendering tasks.

[0064] A reused schedule table is generated by configuring elastic preloading of rendering tasks using resource idle windows. The available resource margin of each resource idle window determines its capacity for preloaded tasks. The task capacity of each window is evaluated by traversing the list of resource idle windows. Tasks suitable for preloading must meet three conditions: input parameters can be determined in advance, output can be cached and reused, and execution duration is less than the duration of the resource idle window. After sorting the tasks that meet the conditions from shortest to longest execution duration, the shorter-duration tasks are prioritized for allocation to resource idle windows to improve window utilization. The allocation process checks whether the resource requirements of each task exceed the window margin. Tasks that exceed the margin are retained for execution at their original trigger time instead of being forcibly preloaded. The original trigger time, preload execution time, and cache identifier of each preloaded task form an allocation record. The preload results are written to the rendering cache for direct access at the original time to save real-time rendering overhead. All preload allocations are integrated with the original schedules of non-preloaded tasks to output a reused schedule table. The reused schedule table records the task number, execution type, and cache reference relationship to be executed at each time using the time axis as an index. The reuse schedule table for a certain interactive scene shows that: the 8-second window preloads the light strip gradient task of the 14-second window, and the 22-second window preloads the halo diffusion task of the 28-second window. The two preloadings reduce the peak GPU load of the 14-second and 28-second windows by 35% and 28%, respectively.

[0065] An effect reuse mode is established based on the reuse schedule table. The distribution ratio of preloaded tasks and real-time rendering tasks in the reuse schedule table reflects the coverage of resource optimization. The proportion of preloaded tasks is counted by traversing the reuse schedule table; the higher the preload ratio, the more significant the peak load reduction effect. Tasks with the same lighting effect type and similar trigger times are grouped to evaluate the feasibility of sharing rendering results. Task groups with similar parameters are extracted from the reuse schedule table. Task groups with the same lighting effect type, parameter differences of less than 10%, and adjacent target areas are suitable for sharing single rendering results. Reuse rules are configured for task groups that meet the sharing conditions. The first task in the group performs actual rendering, and the remaining tasks are configured as cached references. The three-layer strategy of preload configuration, shared reuse rules, and real-time rendering scheduling priority is integrated to form the effect reuse mode. The first layer of the effect reuse mode distributes the peak load to idle periods through preloading. The second layer reduces redundant calculations through rendering result sharing. The third layer ensures the response time of critical tasks through priority scheduling. With the synergistic effect of the three-layer strategy, the peak GPU load during high-parallelism periods can be reduced by 30% to 50%. After applying the effect reuse mode to a certain light and shadow art space, the overload period when the GPU load was originally 95% was reduced to 62%, and the rendering frame rate stabilized from a fluctuating 45-72 frames / second to 65-70 frames / second.

[0066] Based on the effect reuse mode, the collaborative response group is decomposed into light and shadow to output spatial interaction control commands. The task merging rules and scheduling priority configuration in the effect reuse mode are read, and the associated light effect information of each limb response unit in the collaborative response group is loaded. According to the output distribution mapping of the effect reuse mode, it is determined whether the light effect source of each response unit in the collaborative response group is independent rendering or reused and shared, and the reuse rules of the effect reuse mode are traversed to match the light effect type of each response unit. For response units using reused and shared methods, light and shadow are decomposed, and the shared light effect is split into multiple segments according to spatial regions, with each segment corresponding to the projection range of different limb response units. The position offset, brightness scaling factor, and hue offset angle of each segment are calculated to make the combined segments present a visual effect similar to the original independent rendering. The activation sequence of each segment is configured according to the delay compensation trigger chain of the collaborative response group, with the segment corresponding to the dominant limb activated first, and the segments corresponding to subordinate limbs activated sequentially according to the set delay. The light effect segment parameters, activation sequence, and target projection device number are encapsulated into executable control commands, and spatial interaction control commands are output. The spatial interaction control command includes four types of fields: device location, lighting effect parameters, trigger time, and duration. When the user performs the arm rotation action, the spatial interaction control command sequentially drives the top tracking beam to follow the user's rotation direction, the side wall projection to present a spiral diffusion texture, and the ground halo to rotate synchronously and change color. The three sets of lighting effects work together to complete the complete spatial virtual interaction within 2.5 seconds.

[0067] To implement the immersive light and shadow art space interaction method based on motion recognition corresponding to the above method embodiments, and to achieve the corresponding functions and technical effects. See also Figure 2 , Figure 2 This diagram illustrates the structural block diagram of an immersive light and shadow art space interaction system 200 based on motion sensing recognition, as provided in an embodiment of this application. For ease of explanation, only the parts relevant to this embodiment are shown. The immersive light and shadow art space interaction system 200 based on motion sensing recognition provided in this embodiment includes: The signal acquisition module 201 is used to acquire user body sensation signals, perform signal analysis to generate body posture feature data, compare the body posture feature data with a posture template library to determine the action recognition range, and filter valid body sensation inputs based on the action recognition range to establish an available posture dataset; The motion analysis module 202 is used to perform motion trajectory analysis and identify motion rhythm distribution based on the available posture dataset, extract emotion-related features from the motion rhythm distribution to generate responsive motion sequences, and perform limb partitioning and decomposition of composite postures in the body feature data to establish coordinated response groups; The light and shadow mapping module 203 is used to establish a preliminary mapping table based on the responsive action sequence and the cooperative response group to adapt the light and shadow effects, convert the body feature data into light and shadow configuration parameters, perform rendering verification on the preliminary mapping table through the light and shadow configuration parameters to generate an adaptation evaluation result, and initiate light and shadow projection to generate a scene-driven sequence based on the adaptation evaluation result; The experience optimization module 204 is used to analyze the interaction frequency and dwell time using the scene-driven sequence to identify interaction deviation nodes, and to dynamically adjust the lighting and shadow effects based on the interaction deviation nodes to generate an experience optimization strategy. The output control module 205 is used to detect and identify the parallelism of the scene-driven sequence and the experience optimization strategy, identify the synchronous rendering stage, cross-configure the synchronous rendering stage and the responsive action sequence to generate an effect reuse mode, and output spatial interaction control commands to the collaborative response group by decomposing light and shadow according to the effect reuse mode.

[0068] The aforementioned immersive light and shadow art space interaction system 200 based on motion recognition can implement the immersive light and shadow art space interaction method based on motion recognition in the above method embodiments. The options in the above method embodiments are also applicable to this embodiment, and will not be detailed here. The remaining content of this application embodiment can be referred to the content of the above method embodiments, and will not be repeated in this embodiment.

[0069] The purpose of the above embodiments is to reproduce and derive the technical solution of the present invention by way of example, and to fully describe the technical solution, purpose and effect of the present invention. The purpose is to enable the public to have a more thorough and comprehensive understanding of the disclosure of the present invention, and not to limit the scope of protection of the present invention.

[0070] The above embodiments are not an exhaustive list based on the present invention, and there may be many other embodiments not listed. Any substitutions and improvements made without departing from the concept of the present invention are within the protection scope of the present invention.

Claims

1. An immersive light and shadow art space interaction method based on motion recognition, characterized in that... ,include: User body sensory signals are collected and analyzed to generate body posture feature data. The body posture feature data is compared with a posture template library to determine the action recognition range. Based on the action recognition range, valid body sensory inputs are selected to establish a usable posture dataset. Based on the available posture dataset, motion trajectory analysis is performed to identify the action rhythm distribution. Emotion-related features are extracted from the action rhythm distribution to generate responsive action sequences. For composite postures in the body feature data, limb partitioning is performed to establish collaborative response groups. A preliminary mapping table is established based on the responsive action sequence and the coordinated response group to adapt lighting and shadow effects. The body feature data is converted into lighting and shadow configuration parameters. The preliminary mapping table is then rendered and verified using the lighting and shadow configuration parameters to generate an adaptation evaluation result. Based on the adaptation evaluation result, a scene-driven sequence for generating lighting and shadow projection is initiated. The interaction frequency and dwell time are analyzed using the scene-driven sequence to identify interaction deviation nodes. Based on the interaction deviation nodes, dynamic adjustment of lighting effects is triggered to generate experience optimization strategies. The parallelism of the scene-driven sequence and the experience optimization strategy is detected and the synchronous rendering stage is identified. The synchronous rendering stage and the responsive action sequence are cross-configured to generate an effect reuse mode. Based on the effect reuse mode, the collaborative response group is decomposed into light and shadow and outputs spatial interaction control commands.

2. The method according to claim 1, characterized in that... The step of extracting emotion-related features from the action rhythm distribution to generate a responsive action sequence includes: Extract the amplitude of movement and the rate of change of speed from the movement rhythm distribution; Emotional labels are generated by mapping the amplitude of the movement with the rate of change of speed. The emotion tags are used to rank the distribution of action rhythms by emotional response urgency, establishing a priority response queue. The priority response queue is used to organize a sequence of responsive actions.

3. The method according to claim 1, characterized in that... The step of establishing a collaborative response group by partitioning the composite postures in the body feature data into limbs includes: Identify multi-limb coordinated movements from the composite posture and extract a set of coordinated limbs; Motion correlation analysis was performed on the linked limb set to identify the dominant and subordinate limbs; A delay-compensated trigger chain is established based on the response timing difference between the dominant limb and the subordinate limb; The delayed compensation trigger chain organizes limb response units to form a coordinated response group.

4. The method according to claim 1, characterized in that... The step of establishing a preliminary mapping table based on the responsive action sequence and the coordinated response group to adapt lighting and shadow effects includes: Extract action intensity gradient features from the responsive action sequence to establish an intensity-luminance candidate association; Spatial coverage analysis is performed on the coordinated response group to identify areas of abnormal projection coverage; Based on the projection coverage anomaly region, conflict resolution is performed on the intensity-luminance candidate correlation to generate a conflict-free mapping scheme; A preliminary mapping table is constructed based on the conflict-free mapping scheme.

5. The method according to claim 1, characterized in that... The step of generating a compatibility evaluation result by rendering and verifying the preliminary mapping table using the lighting and shadow configuration parameters includes: Based on the lighting and shadow configuration parameters, the initial mapping table is instantiated to construct a rendering test frame; The frame rate and rendering load of the rendered test frames are measured to generate a set of performance metrics. The rendering pass rate is obtained by performing a deviation analysis between the set of performance indicators and a preset performance benchmark. The adaptability assessment results are determined based on the rendering compliance rate.

6. The method according to claim 1, characterized in that... The step of using the scene-driven sequence to analyze interaction frequency and dwell time to identify interaction deviation nodes includes: A frequency distribution curve is generated by counting the number of interaction triggers per unit time from the scene-driven sequence; Perform regional dwell time analysis on the frequency distribution curve to obtain a behavioral heatmap; The attention drift segments identified from the behavior heatmap are marked as candidate deviation areas. Immersion de-locking is performed on the candidate deviation region to determine the interactive deviation node.

7. The method according to claim 1, characterized in that... The method of cross-configuring the synchronous rendering process with the responsive action sequence to generate an effect reuse mode includes: Extract the resource usage characteristics of each rendering task in the synchronous rendering process; Align the resource occupancy characteristics with the trigger timing of the responsive action sequence to identify resource idle windows; The rendering task elastic preloading configuration is generated using the aforementioned resource idle window to create a reused scheduling table; Establish an effect reuse mode based on the reuse schedule table.

8. The method according to claim 2, characterized in that... The step of generating emotion labels by mapping emotion tendencies based on the amplitude of the movement and the rate of change of speed includes: An amplitude level sequence is established by performing energy attenuation gradient grading on the amplitude of the action; The rate of change of velocity is subjected to change inertial feature extraction to generate a rate feature identifier; The amplitude level sequence and the rate feature identifier are fused using an emotion transfer path to construct an emotion feature combination; Emotional tags are generated by querying the emotion mapping rules through the combination of the aforementioned emotional features.

9. The method according to claim 6, characterized in that... The step of determining the interactive deviation node by immersion de-locking the candidate deviation region includes: The observation time window is determined based on the candidate deviation region; Based on the observation time window, the number of interactive recovery attempts and the recovery depth are statistically analyzed to generate recovery quality indicators; When the recovery quality index is lower than a preset recovery threshold, the state is determined to be an immersion detachment state. The start time of the immersion-disengagement state is locked as the interaction deviation node.

10. An immersive light and shadow art space interaction system based on motion perception recognition, characterized in that... ,include: The signal acquisition module is used to acquire user body sensation signals, perform signal analysis to generate body posture feature data, compare the body posture feature data with a posture template library to determine the action recognition range, and filter valid body sensation inputs based on the action recognition range to establish a usable posture dataset; The motion analysis module is used to analyze motion trajectories and identify motion rhythm distribution based on the available posture dataset, extract emotion-related features from the motion rhythm distribution to generate responsive motion sequences, and decompose composite postures in the body feature data into limb partitions to establish coordinated response groups. The lighting and shadow mapping module is used to establish a preliminary mapping table based on the responsive action sequence and the coordinated response group to adapt lighting and shadow effects, convert the body feature data into lighting and shadow configuration parameters, perform rendering verification on the preliminary mapping table using the lighting and shadow configuration parameters to generate an adaptation evaluation result, and initiate lighting and shadow projection to generate a scene-driven sequence based on the adaptation evaluation result; The experience optimization module is used to analyze the interaction frequency and dwell time using the scene-driven sequence to identify interaction deviation nodes, and to dynamically adjust the lighting effects based on the interaction deviation nodes to generate experience optimization strategies. The output control module is used to detect and identify the parallelism of the scene-driven sequence and the experience optimization strategy, identify the synchronous rendering stage, cross-configure the synchronous rendering stage and the responsive action sequence to generate an effect reuse mode, and output spatial interaction control commands to the collaborative response group by decomposing light and shadow according to the effect reuse mode.

Images