Cognition disease care empathic training system based on three-view linkage VR

Abstract

The application discloses a cognitive disease care empathy training system based on three-view linkage VR, relates to the technical field of VR training, and improves the matching precision of caregivers and scenes and the adjustment accuracy of training parameters in the cognitive disease care empathy training process, can adapt to the view angle distribution requirements and AI behavior intensity generation characteristics under the conditions of different caregiver ability distribution characteristics and training scene difficulty, realizes efficient and accurate caregiver scene allocation scheme solving and personalized training parameter dynamic adjustment, significantly improves the ability improvement rate in the cognitive disease care empathy training process and the rationality of training resource configuration, effectively prevents the insufficient response of fixed training parameters to actual caregiver feature differences and scene state evolution, avoids poor training effect and resource waste caused by lagging ability improvement, and ensures that training optimization measures are highly consistent with actual caregiver training needs.

Description

Technical Field

[0001] This invention relates to the field of VR training technology, specifically to an empathy training system for dementia care based on three-view linkage VR. Background Technology

[0002] As a core technical component in building a high-quality dementia care service system, empathy training for dementia care requires precise matching between caregiver competency characteristics and training scenario difficulty, as well as dynamic coupling and adjustment between the training process and real-time performance, under conditions of large-scale caregiver training. Therefore, it is essential to construct an intelligent empathy training system based on two-way matching analysis between caregivers and scenarios and feedback correction of empathy cognition synergy during the training process.

[0003] In existing technologies, the design of empathy training programs for dementia care relies on human experience and fixed training course settings. This makes it difficult to accurately distinguish between insufficient empathy performance caused by caregivers' own ability deficiencies and adaptation problems caused by the difficulty setting of training scenarios. Although existing technologies have introduced basic scoring and testing devices and behavioral observation mechanisms, they do not consider the spatial coupling relationship between the multidimensional ability distribution of caregivers and the symptom types of training scenarios, and lack coupled modeling of the potential for empathy improvement and the decay of scenario adaptability. As a result, it is difficult to accurately assess the actual ability status, the trajectory of personalized training parameter adjustment, and future ability risks during the empathy training process for dementia care. When faced with the multi-factor coupling effect caused by the complex distribution of caregiver characteristics or uneven matching of scenario difficulty, problems such as poor training effect, delayed ability improvement, and failure to detect key training risks may occur.

[0004] Furthermore, existing technologies lack a similarity matching mechanism for the distribution characteristics of different caregivers' abilities and the differences in the difficulty state of training scenario groups during the training process. This makes it impossible to effectively achieve predictive adjustment of training parameters for the next scenario based on the current caregiver's performance data and the formulation of differentiated perspective allocation schemes. In addition, there are defects such as weak scenario difficulty compensation ability, low accuracy of AI virtual elderly behavior intensity distribution, lack of risk correlation analysis for ability improvement, insufficient empathy coordination deviation recovery strategy, and unreasonable determination of training adjustment priority. As a result, the timeliness and accuracy of caregiver training quality early warning cannot be effectively guaranteed. At the same time, there is a lack of adaptive control strategies for sequential training scenarios within the same training scenario sequence, and it is impossible to adaptively generate differentiated training compensation schemes for the next training scenario based on the current dynamic changes such as caregiver's empathy performance and ability improvement evolution. Summary of the Invention

[0005] The purpose of this invention is to provide an empathy training system for dementia care based on three-view linkage VR, which solves the problems existing in the background technology.

[0006] To solve the above-mentioned technical problems, the present invention adopts the following technical solution: The present invention provides a dementia care empathy training system based on three-view linkage VR, including: a baseline acquisition and competency profile module, used to collect the voice feature data, limb movement trajectory data and physiological signal data of caregivers in a standardized scenario testing environment, and perform joint feature extraction to generate a competency profile of caregivers.

[0007] The personalized training scenario generation module is used to obtain various stress scenarios from the local database, and based on the caregiver's competency profile, evaluate the matching degree between each stress scenario and the caregiver's competency profile, and construct a training scenario sequence for the caregiver accordingly.

[0008] The three-view linkage training module is used to load the first training stimulating scenario in the training scenario sequence of caregivers into the first-person experience perspective module, the second-person interaction perspective module, and the third-person observation perspective module. It collects real-time behavioral feedback data of caregivers in each perspective module and calculates the empathy and cognitive synergy of caregivers among the three perspective modules, thereby generating an adaptive three-view linkage training scenario for caregivers.

[0009] The comprehensive ability assessment module is used to collect caregivers' heart rate variability data, eye movement trajectory data, frequency data of verbal empathy keywords, and frequency data of nonverbal reassurance behaviors during the adaptive three-view linkage training scenario, and calculate a three-dimensional comprehensive ability assessment report for caregivers based on this data.

[0010] The training iteration optimization module is used to perform deviation analysis based on the three-dimensional comprehensive ability assessment report of caregivers, identify the lagging abilities of caregivers, and dynamically adjust the next training stimulation scenario for caregivers accordingly, thereby driving the next round of training iteration, and repeating the cycle until all training scenario sequences for caregivers are completed.

[0011] The beneficial effects of this invention are as follows: This invention improves the accuracy of matching caregivers and scenarios and the precision of training parameter adjustment during the empathy training process for dementia caregivers. It can adapt to the perspective allocation needs and AI behavior intensity generation characteristics under different caregiver ability distribution characteristics and training scenario difficulty states. Through bidirectional matching analysis and coupled computation, it accurately distinguishes the caregiver's own ability reserves and scenario adaptability decay factors, thereby achieving efficient and accurate solutions for caregiver scenario allocation and personalized dynamic adjustment of training parameters. This significantly improves the ability improvement rate and the rationality of training resource allocation during the empathy training process for dementia caregivers, and can be based on the ability gap value of each caregiver and the overall difficulty coefficient of the scenario. The system dynamically determines the risk coefficient for capability enhancement and generates behavioral intensity adjustment schemes and perspective allocation compensation schemes for the next training scenario through real-time feedback of empathy and cognitive synergy. This effectively prevents fixed training parameters from being insufficiently responsive to the differences in actual caregiver characteristics and the evolution of scenario states, avoids poor training results and resource waste caused by lagging capability enhancement, and achieves accurate calculation of training parameter compensation values ​​based on the dimension of defective capabilities and the formulation of differentiated scenario difficulty compensation schemes. This effectively improves the pertinence and implementation effect of the final training plan, accurately identifies high-risk training links that threaten the capability enhancement of caregivers and their parameter adjustment windows, and ensures that training optimization measures are highly consistent with the actual training needs of caregivers. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 This is a schematic diagram of the system flow of the present invention. Detailed Implementation

[0014] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0015] Reference Figure 1As shown, the present invention provides a dementia care empathy training system based on three-view linkage VR, including: a baseline acquisition and competency profile module, used to collect the voice feature data, limb movement trajectory data and physiological signal data of caregivers in a standardized scenario testing environment, and perform joint feature extraction to generate a competency profile of caregivers.

[0016] In one specific embodiment, the voice feature data, limb movement trajectory data, and physiological signal data of caregivers are collected. The specific method is as follows: voice feature data of caregivers in standardized scenario tests are collected through the microphone built into the VR headset; limb movement trajectory data of caregivers are tracked through the spatial positioning sensor and controller of the VR headset; and physiological signal data are collected through a physiological monitoring device worn on the wrist of the caregiver.

[0017] In a specific embodiment of the present invention, joint feature extraction is performed to generate a capability profile of caregivers. The specific method is as follows: the voice feature data, limb movement trajectory data and physiological signal data of caregivers are aligned and integrated according to timestamps to form baseline data of each modality of caregivers.

[0018] In one specific embodiment, alignment and integration are performed to form baseline data for each modality of caregivers. The specific method is as follows: current timestamp alignment technology is relatively mature, and baseline data for each modality of caregivers can be formed using existing technology.

[0019] The multimodal feature fusion model is obtained from the local database. The baseline data of each modality of the caregiver is input into the multimodal feature fusion model. The temporal correlation features between the baseline data of each modality of the caregiver are analyzed and fused to output the joint feature vector of the caregiver.

[0020] It should be noted that the multimodal feature fusion model is a deep learning model that includes an encoding layer, a fusion layer, and an output layer. The baseline data of each modality of the caregiver is input into the encoding layer. The encoding layer extracts features from the speech feature data, limb movement trajectory data, and physiological signal data respectively to obtain the feature representation vector of each modality. The fusion layer calculates the temporal correlation weight between the feature representation vectors of each modality through a self-attention mechanism, identifies the mutual influence relationship between different modal data in the time series, and weights and sums the feature representation vectors of each modality according to the temporal correlation weight to obtain the fused multimodal feature vector. The output layer performs dimensionality transformation on the fused multimodal feature vector to output the joint feature vector of the caregiver.

[0021] The feature mapping matrix of each ability dimension is obtained from the local database. The joint feature vector of the caregiver is multiplied by the feature mapping matrix of each ability dimension to obtain the projection vector of the caregiver in each ability dimension. The magnitude of the caregiver in each projection vector is extracted to obtain the score of the caregiver in each ability dimension, and it is used as the ability profile of the caregiver.

[0022] It should be noted that the feature mapping matrix for each ability dimension is a mathematical matrix used to map the joint feature vector of caregivers to specific ability dimensions. The feature mapping matrix for the cognitive dimension contains mapping parameters for symptom recognition ability and risk prediction ability, the feature mapping matrix for the attitude dimension contains mapping parameters for emotional stability and empathy expression, and the feature mapping matrix for the skill dimension contains mapping parameters for operational standardization and intervention timing. The number of rows in each matrix is ​​equal to the number of dimensions of the joint feature vector, and the number of columns is equal to the number of sub-abilities of the corresponding ability dimension. The parameters in the matrix are obtained through training with a large amount of caregiver sample data.

[0023] In one specific embodiment, matrix multiplication is performed to obtain the caregiver's projection vectors for each ability dimension. The modulus of each projection vector is then extracted to obtain the caregiver's score for each ability dimension. Specifically, the caregiver's joint feature vector is multiplied by the feature mapping matrix of the cognitive dimension to obtain the caregiver's projection vector for the cognitive dimension. This projection vector contains two components: symptom recognition ability and risk prediction ability. The square root of the sum of squares of each component in the projection vector is calculated to obtain the modulus of the projection vector, which is then used as the caregiver's score for the cognitive dimension. Similarly, the joint feature vector is multiplied by the feature mapping matrices of the attitude and skill dimensions respectively to obtain the corresponding projection vectors, and the modulus is calculated to obtain the caregiver's score for the attitude and skill dimensions respectively.

[0024] The personalized training scenario generation module is used to obtain various stress scenarios from the local database, and based on the caregiver's competency profile, evaluate the matching degree between each stress scenario and the caregiver's competency profile, and construct a training scenario sequence for the caregiver accordingly.

[0025] It should be noted that agitated scenarios refer to agitated behavioral situations exhibited by elderly people with dementia during the course of the disease. These include verbal agitation scenarios, such as the elderly person continuously shouting or repeatedly asking questions; physical aggression scenarios, such as the elderly person pushing or hitting caregivers; wandering scenarios, such as the elderly person repeatedly walking back and forth along a fixed route; and care refusal scenarios, such as the elderly person refusing to take medication or bathing, or other daily care activities. Each agitated scenario corresponds to different symptom type labels, behavioral intensity levels, and intervention difficulty coefficients, which are used to construct training scenarios for caregivers.

[0026] In a specific embodiment of the present invention, the matching and fit between each adrenaline scenario and the caregiver's competency profile is evaluated. The specific method is as follows: the symptom type labels, behavior intensity levels, intervention difficulty coefficients, and competency requirement mapping tables corresponding to each symptom type label are obtained from the local database for each adrenaline scenario. The competency requirement coefficients for each adrenaline scenario in each competency dimension are extracted. Combined with the behavior intensity level and intervention difficulty coefficient of each adrenaline scenario, the adjusted competency requirement coefficients for each adrenaline scenario in each competency dimension are evaluated. The difference between each dimension is calculated based on the caregiver's score in each competency dimension to obtain the competency gap value of the caregiver in each competency dimension of each adrenaline scenario. The comprehensive competency gap value of the caregiver in each adrenaline scenario is calculated by summing them and is used as the initial matching degree between each adrenaline scenario and the caregiver's competency profile.

[0027] It should be noted that the symptom type labels are used to distinguish the types of agitation scenarios, including verbal agitation, physical aggression, wandering, and care-refusal. The behavior intensity level describes the severity of the elderly's agitation behavior and is divided into three levels: mild, moderate, and severe. The intervention difficulty coefficient is obtained by researchers based on the number of intervention steps required for the agitation scenario, the depth of empathy, and the level of stress control. The ability requirement mapping table records the ability requirement coefficients for each symptom type label in the dimensions of empathy ability, communication skills, and stress response. For example, the ability requirement coefficient for verbal agitation scenarios is relatively high in the communication skills dimension. This mapping table was obtained by researchers based on a large number of samples.

[0028] In one specific embodiment, the adjusted capability requirement coefficients for each adrenaline-fueled scenario across each capability dimension are evaluated. The specific method is as follows: the behavioral intensity level of each adrenaline-fueled scenario is converted into a numerical value, such as 1 for mild, 1.2 for moderate, and 1.4 for severe. The capability requirement coefficients for each adrenaline-fueled scenario are multiplied by the behavioral intensity level value to obtain the intensity-adjusted capability requirement coefficients. Then, the intensity-adjusted capability requirement coefficients are multiplied by the intervention difficulty coefficients for each adrenaline-fueled scenario to obtain the adjusted capability requirement coefficients for each adrenaline-fueled scenario across each capability dimension.

[0029] Based on the caregiver's scores in each competency dimension, each competency dimension with a score below a preset competency threshold is identified as the caregiver's competency gap dimension. Based on the competency gap value in each competency gap dimension of each stressful scenario, the initial matching degree between each stressful scenario and the caregiver's competency profile is adjusted, thereby obtaining the matching and fit degree between each stressful scenario and the caregiver's competency profile.

[0030] In one specific embodiment, adjustments are made to obtain the matching degree between each intense scenario and the caregiver's competency profile. The specific method is as follows: the competency gap values ​​of each intense scenario in each competency gap dimension are added together to obtain the total gap, and the total gap of each intense scenario is added to the initial matching degree of the intense scenario to obtain the matching degree between each intense scenario and the caregiver's competency profile.

[0031] In a specific embodiment of the present invention, the training scenario sequence for caregivers is constructed by: obtaining the target adaptation range from the local database, and filtering each intense scenario whose matching degree is in the target adaptation range based on the matching degree between each intense scenario and the caregiver's ability profile, thereby obtaining each training intense scenario for caregivers.

[0032] Based on the intervention difficulty coefficient of each training agitation scenario for caregivers, the scenarios are sorted in ascending order to obtain the sorted training agitation scenarios for caregivers. Adjacent training agitation scenarios are traversed and their corresponding symptom type labels are detected. If the symptom type labels of adjacent training agitation scenarios are the same, the subsequent training agitation scenario is swapped with the training agitation scenario with different symptom type labels in the subsequent sequence. If there are no interchangeable training agitation scenarios, they are kept in their original positions. This process yields the adjusted sorted training agitation scenarios for caregivers, which are then used as the training scenario sequence for caregivers.

[0033] The three-view linkage training module is used to load the first training stimulating scenario in the training scenario sequence of caregivers into the first-person experience perspective module, the second-person interaction perspective module, and the third-person observation perspective module. It collects real-time behavioral feedback data of caregivers in each perspective module and calculates the empathy and cognitive synergy of caregivers among the three perspective modules, thereby generating an adaptive three-view linkage training scenario for caregivers.

[0034] It should be noted that the first-person experience perspective module refers to caregivers directly facing an AI-powered virtual elderly person with dementia from a first-person perspective through VR headsets, experiencing the elderly person's agitated behavior and recognizing their emotions. The second-person interaction perspective module refers to caregivers directly interacting with the AI-powered virtual elderly person with dementia in the VR environment, intervening through voice reassurance and physical contact. The third-person observation perspective module refers to caregivers watching standardized care demonstration videos from an observer's perspective, learning the standardized procedures for handling agitated behavior by professional caregivers. The three perspective modules respectively train caregivers' empathy and perception abilities, interactive intervention abilities, and cognitive learning abilities.

[0035] In one specific embodiment, real-time behavioral feedback data of caregivers in each perspective module is collected. The specific method is as follows: In the first-person experience perspective module, the eye-tracking camera of the VR headset is used to collect the coordinates of the caregiver's gaze point and the duration of gaze lingering. The microphone is used to collect the caregiver's voice and identify emotion recognition words. In the second-person interactive perspective module, the voice data of the caregiver is collected through the voice recognition system and reassuring words and empathetic expressions are identified. The haptic feedback sensor of the VR controller is used to detect contact events between the caregiver's virtual hand and the AI ​​virtual elderly person with dementia. In the third-person observation perspective module, the eye-tracking camera is used to collect the gaze focus trajectory of the caregiver when watching the demonstration video.

[0036] In a specific embodiment of the present invention, the empathic cognitive coordination degree of caregivers in the three-viewpoint modules is calculated as follows: in the first-person experience perspective module, the coordinates of the caregiver's gaze point are collected, the proportion of the caregiver's gaze lingering on the facial expression area of ​​the AI ​​virtual cognitively impaired elderly person is counted as the emotional attention degree, the time difference between the caregiver's utterance of the emotional recognition words after the occurrence of agitated behavior is detected as the emotional recognition reaction time, and the caregiver's focus empathy index is calculated accordingly.

[0037] In one specific embodiment, the coordinates of the caregiver's gaze points are collected, and the percentage of time the caregiver's gaze lingers on the facial expression area of ​​the AI-simulated elderly person with dementia is counted as the emotional attention level. The time difference between the caregiver's utterance of the emotion-recognizing word after the occurrence of agitated behavior is detected as the emotion recognition reaction time, and the caregiver's attention empathy index is calculated accordingly. The specific method is as follows: the coordinates of the caregiver's gaze points are collected through the eye-tracking camera of the VR headset, the coordinate range of the facial expression area of ​​the AI-simulated elderly person with dementia is marked in the virtual scene, and it is determined whether the caregiver's gaze point coordinates fall within the range. In this area, the duration of caregivers' gaze lingering on the facial expression area is counted and divided by the total observation time to obtain the emotional attention score. The timestamps of agitated behaviors of AI-simulated elderly people with dementia are recorded. The caregivers' voices are collected through a microphone and converted into text by a speech recognition system. An emotion recognition vocabulary is obtained from a local database. The timestamps of caregivers uttering emotion recognition words are detected. The difference between the two timestamps is calculated to obtain the emotion recognition reaction time. The emotional attention score is multiplied by a preset weighting coefficient, and the normalized value of the emotion recognition reaction time is subtracted by multiplying it by the preset weighting coefficient to obtain the caregiver's focus empathy index.

[0038] In the second-person interactive perspective module, the number of times the caregiver uses voice to comfort the elderly is identified, and the contact time between the caregiver and the AI ​​virtual elderly with dementia during each voice comfort is detected, and the caregiver's interaction empathy index is calculated accordingly.

[0039] In one specific embodiment, the number of times a caregiver uses verbal reassurance is identified, and the contact duration between the caregiver and the AI-generated virtual elderly person with dementia during each verbal reassurance is detected. Based on this, the caregiver's interaction empathy index is calculated. The specific method is as follows: Voice data from the caregiver is collected via microphone; the voice is converted to text using a speech recognition system; a reassurance vocabulary lexicon is retrieved from a local database, containing reassurance words, empathic expressions, and emotionally supportive statements; the frequency of each of these voice types in the caregiver's voice text is counted and summed to obtain the caregiver's voice empathy index. The number of reassurance events is determined by detecting contact events between the caregiver's virtual hand and the AI-powered virtual elderly person with dementia using the haptic feedback sensor of the VR controller. The start and end timestamps of each contact event are recorded, and the difference between the two timestamps is calculated to obtain the duration of each contact. The time intervals in which the caregiver performs voice reassurance are identified, and the contact durations within the voice reassurance time intervals are extracted and summed to obtain the total contact duration. The normalized value of the number of voice reassurance events is multiplied by a preset weighting coefficient, and then the normalized value of the total contact duration is multiplied by the preset weighting coefficient to obtain the caregiver's interaction empathy index.

[0040] The system retrieves standardized care demonstration videos corresponding to the current training stimulating scenarios from the local database and uses them as target videos. In the third-person observation perspective module, the target videos are played, and the caregiver's gaze focus trajectory is collected. This trajectory is then matched with the key operational steps in the target videos, and the accuracy rate of the caregiver's behavioral points in the target videos is calculated. This accuracy rate is then used as the caregiver's cognitive empathy index.

[0041] In one specific embodiment, the caregiver's gaze focus trajectory is collected and matched with key operation steps in the target video to calculate the accuracy rate of the caregiver's behavioral points recognition in the target video. The specific method is as follows: the eye-tracking camera of the VR headset collects the gaze point coordinate sequence of the caregiver's eyes when watching the target video, records the gaze point coordinates at each time point to form the caregiver's gaze focus trajectory, obtains the timestamps and spatial coordinate ranges of each key operation step in the target video from the local database, traverses the caregiver's gaze focus trajectory, and for each key operation step, extracts the caregiver's gaze point coordinates corresponding to the timestamp of the step, determines whether the gaze point coordinates fall within the spatial coordinate range of the step, and marks the step as correctly recognized if it does. The number of correctly recognized steps is counted and divided by the total number of steps in the target video to obtain the accuracy rate of the caregiver's behavioral points recognition in the target video.

[0042] Based on caregivers’ attentional empathy index, interactive empathy index, and cognitive empathy index, the variance among the three is calculated and used as the caregiver’s empathy-cognitive synergy among the three perspective modules.

[0043] In one specific embodiment, the variance among the three is calculated as follows: existing variance calculation methods are mature, and the variance among the three can be calculated using existing variance techniques.

[0044] In a specific embodiment of the present invention, an adaptive three-view linkage training scenario for caregivers is generated. The specific method is as follows: the basic behavioral intensity parameters of the AI ​​virtual dementia elderly are obtained from the local database, and the intensity is adjusted according to the empathic cognitive coordination degree of caregivers among the three-view modules, thereby obtaining the target behavioral intensity parameters of the AI ​​virtual dementia elderly.

[0045] In one specific embodiment, intensity adjustment is performed to obtain the target behavior intensity parameters of the AI ​​virtual dementia elderly person. The specific method is as follows: obtain the adjustment ratio corresponding to each empathy cognition coordination degree interval from the local database, thereby mapping the adjustment ratio of the caregiver, and multiply the basic behavior intensity parameters by the adjustment ratio to obtain the target behavior intensity parameters of the AI ​​virtual dementia elderly person.

[0046] Based on caregivers' attentional empathy index, interactive empathy index, and cognitive empathy index, the perspective module corresponding to the lowest value index is identified as the lagging perspective module. The difference between the empathy index of the lagging perspective module and the average of the three empathy indices is calculated and used as the caregiver's degree of lag.

[0047] The baseline coefficients of voice feedback and motion amplitude corresponding to the lagging perspective module are obtained from the local database. Based on the lag degree value of the caregiver, dynamic enhancement processing is performed on them to obtain the voice feedback enhancement coefficient and motion amplitude enhancement coefficient of the AI ​​virtual cognitively impaired elderly person.

[0048] In one specific embodiment, dynamic enhancement processing is performed to obtain the voice feedback enhancement coefficient and motion amplitude enhancement coefficient of the AI ​​virtual dementia elderly person. The specific method is as follows: the voice feedback baseline coefficient and motion amplitude baseline coefficient corresponding to the lagging perspective module are obtained from the local database. The baseline coefficient describes the default feedback intensity of the virtual elderly person to the caregiver's behavior in the perspective module. Based on the caregiver's lag degree value, the enhancement factor corresponding to each lag degree value interval is obtained from the local database. The larger the lag degree value, the larger the enhancement factor, thereby mapping the caregiver's enhancement factor. The voice feedback baseline coefficient is multiplied by the enhancement factor to obtain the voice feedback enhancement coefficient of the AI ​​virtual dementia elderly person. The motion amplitude baseline coefficient is multiplied by the enhancement factor to obtain the motion amplitude enhancement coefficient of the AI ​​virtual dementia elderly person.

[0049] The target behavior intensity parameter, voice feedback enhancement coefficient, and motion amplitude enhancement coefficient are applied to the behavioral and feedback performance of the AI ​​virtual elderly with dementia, and combined with the current training intensity scene for rendering, thereby generating an adaptive three-view linkage training scene for caregivers.

[0050] In one specific embodiment, rendering is performed to generate an adaptive three-view linkage training scene for caregivers. The specific method is as follows: the scene description information of the current training agitation scene is obtained from the local database. This information includes the spatial layout of the virtual environment, the initial position and initial state of the virtual elderly person. The target behavior intensity parameter is applied to the behavior generation module of the AI ​​virtual dementia elderly person to adjust the intensity level of the overall agitation behavior of the virtual elderly person. The voice feedback enhancement coefficient is applied to the voice feedback module of the virtual elderly person to adjust the feedback intensity of the virtual elderly person's voice reassurance to caregivers. The motion amplitude enhancement coefficient is applied to the motion feedback module of the virtual elderly person to adjust the feedback amplitude of the virtual elderly person's physical contact with caregivers. The above parameters and scene description information are integrated through the virtual scene rendering engine to generate a complete three-dimensional virtual scene, thereby generating an adaptive three-view linkage training scene for caregivers.

[0051] The comprehensive ability assessment module is used to collect caregivers' heart rate variability data, eye movement trajectory data, frequency data of verbal empathy keywords, and frequency data of nonverbal reassurance behaviors during the adaptive three-view linkage training scenario, and calculate a three-dimensional comprehensive ability assessment report for caregivers based on this data.

[0052] In one specific embodiment, heart rate variability data, eye movement tracking data, frequency data of verbal empathy keywords, and frequency data of nonverbal reassurance behaviors of caregivers are collected. The specific method is as follows: heart rate variability data is collected through a physiological monitoring device worn on the caregiver's wrist; eye movement tracking data is collected through an eye-tracking camera built into a VR headset; verbal empathy keyword frequency data is collected through a microphone to capture the caregiver's speech, which is converted into text using a speech recognition system; an empathy keyword lexicon is obtained from a local database; and the total frequency of caregivers using comprehension-related words, comfort-related words, and support-related words is counted; nonverbal reassurance behavior frequency data is collected by a VR controller to detect the number of times caregivers perform reassurance actions such as patting the shoulder, shaking hands, and patting the back.

[0053] In a specific embodiment of the present invention, the method for calculating the three-dimensional comprehensive ability assessment report of caregivers is as follows: based on the caregiver's eye movement trajectory data, extract the caregiver's gaze hotspot distribution map, detect the caregiver's gaze reaction time difference before each agitated behavior, and calculate the caregiver's current cognitive dimension score accordingly.

[0054] In one specific embodiment, a gaze hotspot distribution map of caregivers is extracted, the gaze reaction time difference before each agitated behavior is detected, and the current cognitive dimension score of the caregivers is calculated accordingly. The specific method is as follows: based on the caregiver's eye movement trajectory data, the spatial distribution frequency of gaze point coordinates when the caregiver observes an AI-generated virtual elderly person with dementia is statistically analyzed. The virtual elderly person's body area is divided into a grid, the number of gaze points in each grid is calculated, a gaze hotspot distribution map is generated, and typical gaze patterns corresponding to each dementia symptom are obtained from a local database. This pattern describes the typical gaze hotspot of a professional caregiver when recognizing the symptom. The point distribution method calculates the similarity between the caregiver's gaze hotspot distribution map and various typical gaze patterns. It counts the number of symptoms with similarity exceeding the matching threshold and divides it by the total number of symptoms actually exhibited by the AI ​​virtual cognitive impairment elderly to obtain the key symptom recognition accuracy. It records the timestamps of the agitated behavior precursors of the AI ​​virtual cognitive impairment elderly and detects the timestamps of the caregiver's gaze shifting to the abnormal behavior area. It calculates the difference between the two timestamps to obtain the gaze reaction time difference. It multiplies the key symptom recognition accuracy by a preset weight coefficient and subtracts the normalized value of the gaze reaction time difference multiplied by the preset weight coefficient to obtain the caregiver's current cognitive dimension score.

[0055] Baseline heart rate variability (HRV) indices of caregivers in a standardized scenario testing environment are obtained from a local database. The rate of change between the caregiver's HRV data and the baseline HRV indices is calculated and used as the caregiver's emotional stability coefficient.

[0056] In one specific embodiment, the rate of change between the caregiver's heart rate variability data and the baseline heart rate variability index is calculated. The specific method is as follows: based on the caregiver's heart rate variability data, the ratio of the caregiver's heart rate variability time-domain index and frequency-domain index in the current training stress scenario is extracted. The current time-domain index is calculated by subtracting the baseline time-domain index and then dividing by the baseline time-domain index to obtain the time-domain index change rate. Similarly, the frequency-domain index change rate is obtained. The time-domain index change rate and the frequency-domain index change rate are weighted and averaged to obtain the rate of change between the heart rate variability data and the baseline heart rate variability index.

[0057] The empathy keyword lexicon is obtained from the local database. Based on the frequency data of caregivers' verbal empathy keywords, the frequency of caregivers' empathic language expressions is statistically analyzed. Combined with the caregivers' emotional stability coefficient, the current attitude dimension score of caregivers is calculated.

[0058] In one specific embodiment, the frequency of empathic language expressions by caregivers is statistically analyzed, and combined with the caregiver's emotional stability coefficient, the current attitude dimension score of the caregiver is calculated. The specific method is as follows: the caregiver's voice data in the current training intense scenario is collected through a microphone, the voice is converted into text using a speech recognition system, an empathic keyword lexicon is obtained from a local database, which includes comprehension-related words, comfort-related words, and support-related words, the total number of times empathic keywords appear in the caregiver's voice text is counted, divided by the total dialogue time between the caregiver and the AI ​​virtual elderly person with dementia, to obtain the frequency of empathic language expressions by the caregiver, the frequency of empathic language expressions is multiplied by a preset weight coefficient, and then added to the emotional stability coefficient multiplied by the preset weight coefficient to obtain the current attitude dimension score of the caregiver.

[0059] A standard soothing action sequence library is obtained from the local database. The frequency data of caregivers' nonverbal soothing behaviors are matched with the standard soothing action sequence library to obtain the standardization of caregivers' soothing techniques. The optimal intervention time window corresponding to each stage of agitated behavior development is obtained from the local database. The frequency data of caregivers' nonverbal soothing behaviors are compared to obtain the accuracy rate of caregivers' intervention timing. The current skill dimension score of caregivers is calculated by combining the results.

[0060] In one specific embodiment, the standardization of the caregiver's soothing techniques is obtained by: acquiring a standard soothing action sequence library from a local database. This library contains standard action sequences for various soothing actions such as patting the shoulder, shaking hands, patting the back, and helping someone up. Each standard action sequence describes the standard trajectory and speed curve of the action. Based on the frequency data of the caregiver's nonverbal soothing behaviors, the trajectory data and speed data of each soothing action actually performed by the caregiver are extracted. The actual soothing actions performed by the caregiver are matched with the standard action sequences of the corresponding actions in the standard soothing action sequence library using a dynamic time warping algorithm. The similarity score between the two sequences is calculated. The similarity scores of all soothing actions are averaged to obtain the standardization of the caregiver's soothing techniques.

[0061] In one specific embodiment, the frequency data of caregivers' nonverbal reassurance behaviors are compared to obtain the accuracy rate of intervention timing. The caregiver's current skill dimension score is then calculated. Specifically, based on the frequency data of caregivers' nonverbal reassurance behaviors, timestamps of each reassurance behavior are extracted. These timestamps are compared with the optimal intervention time window for the current agitated behavior development stage of the AI-simulated dementia elderly person. If the timestamp falls within the optimal intervention time window, the intervention timing is considered correct. The number of correct intervention times is counted and divided by the total number of reassurance behaviors performed by the caregiver to obtain the accuracy rate of intervention timing. The standardization of the reassurance technique is multiplied by a preset weighting coefficient, and this is added to the accuracy rate of intervention timing multiplied by the preset weighting coefficient to obtain the caregiver's current skill dimension score.

[0062] The caregiver's current cognitive dimension score, current attitude dimension score, and current skill dimension score are combined to obtain a three-dimensional comprehensive ability assessment report of the caregiver.

[0063] In a specific embodiment of the present invention, deviation analysis is performed to identify the lagging ability of caregivers to improve. The specific method is as follows: based on the caregiver's ability profile, the initial score value of the caregiver in each ability dimension is extracted, and combined with the caregiver's current cognitive dimension score value, current attitude dimension score value and current skill dimension score value, the score improvement value of the caregiver in each ability dimension is calculated.

[0064] In one specific embodiment, the improvement value of caregivers' scores in each competency dimension is calculated by subtracting the initial score of the cognitive dimension from the current cognitive dimension score to obtain the improvement value of the cognitive dimension score; subtracting the initial score of the attitude dimension from the current attitude dimension score to obtain the improvement value of the attitude dimension score; and subtracting the initial score of the skill dimension from the current skill dimension score to obtain the improvement value of the skill dimension score.

[0065] Based on the competency gap values ​​of caregivers in each competency dimension of each stressful scenario, the competency gap values ​​of caregivers in each competency dimension of the current training stressful scenario are extracted. Combined with the improvement values ​​of caregivers' scores in each competency dimension, the competency dimensions of each deficiency of caregivers are screened and regarded as the competencies of caregivers that need improvement.

[0066] In one specific embodiment, the method for screening the deficient competency dimensions of caregivers is as follows: based on the competency gap values ​​of caregivers in each competency dimension in the current training stress scenario, gap thresholds are obtained from the local database. The competency gap value of caregivers in the cognitive dimension is compared with the gap threshold. If the competency gap value is greater than the gap threshold, the cognitive dimension is determined to be a deficient competency dimension. Similarly, the competency gap values ​​of caregivers in the attitude and skill dimensions are compared with the gap thresholds respectively to identify each competency dimension with a competency gap value greater than the gap threshold. Combined with the score improvement value of caregivers in each competency dimension, if the score improvement value of a certain competency dimension is less than the improvement threshold, and the competency gap value of that competency dimension is greater than the gap threshold, then that competency dimension is marked as a deficient competency dimension, thereby screening out the various deficient competency dimensions of caregivers.

[0067] The training iteration optimization module is used to perform deviation analysis based on the three-dimensional comprehensive ability assessment report of caregivers, identify the lagging abilities of caregivers, and dynamically adjust the next training stimulation scenario for caregivers accordingly, thereby driving the next round of training iteration, and repeating the cycle until all training scenario sequences for caregivers are completed.

[0068] In a specific embodiment of the present invention, the next training intensity scenario for caregivers is dynamically adjusted to drive the next round of training iterations. The specific method is as follows: based on the intervention difficulty coefficient of each intensity scenario, the difference in intervention difficulty coefficient between the current training intensity scenario and the next training intensity scenario for caregivers is calculated. Combined with the ability gap values ​​of caregivers in each defect ability dimension, the target behavior intensity parameters, voice feedback enhancement coefficient, and movement amplitude enhancement coefficient of the AI ​​virtual cognitively impaired elderly person are updated. Then, the three-view linkage training module is re-executed to drive the caregiver to train in the next training intensity scenario.

[0069] In a specific embodiment of the present invention, the target behavior intensity parameters, voice feedback enhancement coefficient, and movement amplitude enhancement coefficient of the AI ​​virtual cognitively impaired elderly are updated. The specific method is as follows: based on the ability gap values ​​of caregivers in each defect ability dimension, the proportion of the ability gap value of each defect ability dimension of caregivers to the sum of the ability gap values ​​of all defect ability dimensions is calculated, thereby obtaining the weighted comprehensive gap value of caregivers.

[0070] In one specific embodiment, the proportion of the capacity gap value of each deficiency capacity dimension of the caregiver to the sum of the capacity gap values ​​of all deficiency capacity dimensions is calculated to obtain the weighted composite capacity gap value of the caregiver. The specific method is as follows: the capacity gap values ​​corresponding to each deficiency capacity dimension of the caregiver are added together to obtain the sum of the capacity gap values ​​of all deficiency capacity dimensions. For each deficiency capacity dimension, the capacity gap value of that dimension is divided by the sum of the capacity gap values ​​to obtain the capacity gap value proportion of that deficiency capacity dimension. The importance weight coefficient of each capacity dimension is obtained from the local database. The capacity gap value proportion of each deficiency capacity dimension is multiplied by the importance weight coefficient of that dimension to obtain the weighted gap value of that dimension. The weighted gap values ​​of all deficiency capacity dimensions are added together to obtain the weighted composite capacity gap value of the caregiver.

[0071] The target adjustment coefficient of caregivers is assessed based on the difference in intervention difficulty coefficient between the current training stress scenario and the next training stress scenario, combined with the caregiver's weighted comprehensive gap value.

[0072] In one specific embodiment, the target adjustment coefficient of caregivers is evaluated as follows: Based on the difference in intervention difficulty coefficient between the current training stimuli scenario and the next training stimuli scenario, a difficulty difference mapping table is obtained from a local database. This mapping table records the basic adjustment coefficients corresponding to different intervention difficulty coefficient differences, thereby mapping the basic adjustment coefficient of the corresponding caregiver. Based on the weighted composite gap value of caregivers, a gap correction coefficient table is obtained from a local database. This table records the correction coefficients corresponding to different weighted composite gap values, thereby mapping the correction coefficient of the corresponding caregiver. The basic adjustment coefficient and the correction coefficient are multiplied to obtain the target adjustment coefficient of the caregiver.

[0073] The target behavior intensity parameter, voice feedback enhancement coefficient, and motion amplitude enhancement coefficient of the AI-simulated elderly person with dementia are multiplied with the target adjustment coefficient to obtain the updated target behavior intensity parameter, updated voice feedback enhancement coefficient, and updated motion amplitude enhancement coefficient simultaneously.

[0074] All formulas in this manual are dimensionless and calculated numerically. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0075] The above content is merely an example and illustration of the concept of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the concept of the invention or exceed the scope defined by the present invention, and all such modifications and additions should fall within the protection scope of the present invention.

Claims

1. A dementia care empathy training system based on three-view linkage VR, characterized in that, include: The baseline acquisition and competency profile module is used to collect voice feature data, limb movement trajectory data and physiological signal data of caregivers in a standardized scenario testing environment, and perform joint feature extraction to generate competency profiles of caregivers. The personalized training scenario generation module is used to obtain various stress scenarios from the local database, and based on the caregiver's competency profile, evaluate the matching degree between each stress scenario and the caregiver's competency profile, and construct a training scenario sequence for the caregiver accordingly. The three-view linkage training module is used to load the first training stimulating scenario in the training scenario sequence of caregivers into the first-person experience perspective module, the second-person interaction perspective module, and the third-person observation perspective module. It collects real-time behavioral feedback data of caregivers in each perspective module and calculates the empathy and cognitive synergy of caregivers among the three perspective modules, thereby generating an adaptive three-view linkage training scenario for caregivers. The comprehensive ability assessment module is used to collect caregivers' heart rate variability data, eye movement trajectory data, frequency data of verbal empathy keywords, and frequency data of nonverbal reassurance behaviors during the adaptive three-view linkage training scenario, and calculate a three-dimensional comprehensive ability assessment report for caregivers based on this data. The training iteration optimization module is used to perform deviation analysis based on the three-dimensional comprehensive ability assessment report of caregivers, identify the lagging abilities of caregivers, and dynamically adjust the next training stimulation scenario for caregivers accordingly, thereby driving the next round of training iteration, and repeating the cycle until all training scenario sequences for caregivers are completed.

2. The dementia care empathy training system based on three-view linkage VR as described in claim 1, characterized in that, The specific method for performing joint feature extraction to generate a competency profile of caregivers is as follows: The voice feature data, limb movement trajectory data and physiological signal data of caregivers are aligned and integrated according to timestamps to form baseline data of each modality of caregivers; The multimodal feature fusion model is obtained from the local database. The baseline data of each modality of the caregiver is input into the multimodal feature fusion model. The temporal correlation features between the baseline data of each modality of the caregiver are analyzed and fused to output the joint feature vector of the caregiver. The feature mapping matrix of each ability dimension is obtained from the local database. The joint feature vector of the caregiver is multiplied by the feature mapping matrix of each ability dimension to obtain the projection vector of the caregiver in each ability dimension. The magnitude of the caregiver in each projection vector is extracted to obtain the score of the caregiver in each ability dimension, and it is used as the ability profile of the caregiver.

3. The dementia care empathy training system based on three-view linkage VR as described in claim 2, characterized in that, The specific method for assessing the matching and fit between each stressful scenario and the caregiver's competency profile is as follows: The local database retrieves symptom type labels, behavior intensity levels, intervention difficulty coefficients, and capability requirement mapping tables corresponding to each symptom type label for each agitation scenario. This allows for the extraction of capability requirement coefficients for each agitation scenario across each capability dimension. Combined with the behavior intensity level and intervention difficulty coefficient of each agitation scenario, the adjusted capability requirement coefficients for each agitation scenario across each capability dimension are assessed. Based on the caregiver's scores in each capability dimension, a dimensional difference calculation is performed to obtain the capability gap value of the caregiver in each capability dimension of each agitation scenario. These values ​​are then summed to calculate the comprehensive capability gap value of the caregiver in each agitation scenario, which is used as the initial matching degree between each agitation scenario and the caregiver's capability profile. Based on the caregiver's scores in each competency dimension, each competency dimension with a score below a preset competency threshold is identified as the caregiver's competency gap dimension. Based on the competency gap value in each competency gap dimension of each stressful scenario, the initial matching degree between each stressful scenario and the caregiver's competency profile is adjusted, thereby obtaining the matching and fit degree between each stressful scenario and the caregiver's competency profile.

4. The dementia care empathy training system based on three-view linkage VR as described in claim 1, characterized in that, The specific method for constructing the training scenario sequence for caregivers is as follows: The target fit range is obtained from the local database. Based on the matching fit between each stress scenario and the caregiver's ability profile, each stress scenario with a matching fit within the target fit range is selected, thereby obtaining each training stress scenario for the caregiver. Based on the intervention difficulty coefficient of each training agitation scenario for caregivers, the scenarios are sorted in ascending order to obtain the sorted training agitation scenarios for caregivers. Adjacent training agitation scenarios are traversed and their corresponding symptom type labels are detected. If the symptom type labels of adjacent training agitation scenarios are the same, the subsequent training agitation scenario is swapped with the training agitation scenario with different symptom type labels in the subsequent sequence. If there are no interchangeable training agitation scenarios, they are kept in their original positions. This process yields the adjusted sorted training agitation scenarios for caregivers, which are then used as the training scenario sequence for caregivers.

5. The dementia care empathy training system based on three-view linkage VR as described in claim 1, characterized in that, The specific method for calculating the degree of empathic cognitive synergy among caregivers in the three-perspective modules is as follows: In the first-person perspective module, the coordinates of the caregiver's gaze point are collected, the percentage of time the caregiver's gaze lingers on the facial expression area of ​​the AI ​​virtual elderly person with dementia is counted as the emotional attention level, the time difference between the caregiver's agitated behavior and the expression of the emotional recognition words is detected as the emotional recognition reaction time, and the caregiver's focus empathy index is calculated accordingly. In the second-person interactive perspective module, the number of times the caregiver uses voice to comfort the elderly is identified, and the contact time between the caregiver and the AI ​​virtual elderly with dementia during each voice comfort is detected, and the caregiver's interaction empathy index is calculated accordingly. The standardized care demonstration video corresponding to the current training intense scenario is obtained from the local database and used as the target video. In the third-person observation perspective module, the target video is played, and the gaze focus trajectory of the caregiver is collected. It is matched with the key operation steps in the target video, and the accuracy of the caregiver's behavioral points in the target video is calculated and used as the caregiver's cognitive empathy index. Based on caregivers’ attentional empathy index, interactive empathy index, and cognitive empathy index, the variance among the three is calculated and used as the caregiver’s empathy-cognitive synergy among the three perspective modules.

6. The dementia care empathy training system based on three-view linkage VR as described in claim 5, characterized in that, The specific method for generating the adaptive three-view linkage training scenario for caregivers is as follows: The basic behavioral intensity parameters of the AI ​​virtual elderly with dementia are obtained from the local database, and the intensity is adjusted according to the empathic cognitive coordination of caregivers among the three perspective modules, so as to obtain the target behavioral intensity parameters of the AI ​​virtual elderly with dementia. Based on caregivers’ attention empathy index, interaction empathy index and cognitive empathy index, the perspective module corresponding to the lowest value index is identified as the lagging perspective module. The difference between the empathy index of the lagging perspective module and the average of the three empathy indices is calculated and used as the caregiver’s degree of lag. The baseline coefficients of voice feedback and motion amplitude corresponding to the lagging perspective module are obtained from the local database. Based on the lag value of the caregiver, dynamic enhancement processing is performed to obtain the voice feedback enhancement coefficient and motion amplitude enhancement coefficient of the AI ​​virtual cognitively impaired elderly person. The target behavior intensity parameter, voice feedback enhancement coefficient, and motion amplitude enhancement coefficient are applied to the behavioral and feedback performance of the AI ​​virtual elderly with dementia, and combined with the current training intensity scene for rendering, thereby generating an adaptive three-view linkage training scene for caregivers.

7. The dementia care empathy training system based on three-view linkage VR as described in claim 2, characterized in that, The specific method for calculating the three-dimensional comprehensive ability assessment report of caregivers is as follows: Based on the eye movement trajectory data of caregivers, the distribution map of caregivers' gaze hotspots is extracted, the gaze reaction time difference of caregivers before each agitated behavior is detected, and the current cognitive dimension score of caregivers is calculated accordingly. Baseline heart rate variability (HRV) indices of caregivers in a standardized scenario testing environment are obtained from a local database. The rate of change between the caregivers' HRV data and the baseline HRV indices is calculated and used as the caregivers' emotional stability coefficient. The empathy keyword lexicon is obtained from the local database. Based on the frequency data of caregivers' verbal empathy keywords, the frequency of caregivers' empathic language expressions is statistically analyzed. Combined with the caregivers' emotional stability coefficient, the current attitude dimension score of caregivers is calculated. A standard soothing action sequence library is obtained from the local database. The frequency data of caregivers' nonverbal soothing behaviors are matched with the standard soothing action sequence library to obtain the standardization of caregivers' soothing techniques. The optimal intervention time window corresponding to each stage of agitated behavior development is obtained from the local database. The frequency data of caregivers' nonverbal soothing behaviors are compared to obtain the accuracy rate of caregivers' intervention timing. The current skill dimension score of caregivers is calculated in a comprehensive manner. The caregiver's current cognitive dimension score, current attitude dimension score, and current skill dimension score are combined to obtain a three-dimensional comprehensive ability assessment report of the caregiver.

8. The dementia care empathy training system based on three-view linkage VR as described in claim 7, characterized in that, The specific method for conducting deviation analysis and identifying the lagging ability of caregivers to improve is as follows: Based on the caregiver's competency profile, the initial scores of the caregiver in each competency dimension are extracted. Then, combined with the caregiver's current cognitive dimension score, current attitude dimension score, and current skill dimension score, the improvement value of the caregiver's score in each competency dimension is calculated. Based on the competency gap values ​​of caregivers in each competency dimension of each stressful scenario, the competency gap values ​​of caregivers in each competency dimension of the current training stressful scenario are extracted. Combined with the improvement values ​​of caregivers' scores in each competency dimension, the competency dimensions of each deficiency of caregivers are screened and regarded as the competencies of caregivers that need improvement.

9. The dementia care empathy training system based on three-view linkage VR as described in claim 8, characterized in that, The specific method for dynamically adjusting the next training stimuli scenario for caregivers to drive the next round of training iterations is as follows: Based on the intervention difficulty coefficient of each intense scenario, the difference in intervention difficulty coefficient between the current training intense scenario and the next training intense scenario is calculated. Combined with the ability gap values ​​of the caregiver in each defective ability dimension, the target behavior intensity parameters, voice feedback enhancement coefficient and movement amplitude enhancement coefficient of the AI ​​virtual cognitively impaired elderly are updated. Then, the three-view linkage training module is re-executed to drive the caregiver to train in the next training intense scenario.

10. The dementia care empathy training system based on three-view linkage VR according to claim 9, characterized in that, The specific method for updating the target behavior intensity parameters, voice feedback enhancement coefficient, and movement amplitude enhancement coefficient of the AI ​​virtual dementia elderly person is as follows: Based on the competency gap values ​​of caregivers in each competency deficit dimension, the proportion of the competency gap value of each competency deficit dimension to the total competency gap value of all competency deficit dimensions is calculated, thereby obtaining the weighted composite competency gap value of caregivers. Based on the difference in intervention difficulty coefficient between the current training agitation scenario and the next training agitation scenario for caregivers, and combined with the caregiver's weighted comprehensive gap value, the target adjustment coefficient of caregivers is assessed. The target behavior intensity parameter, voice feedback enhancement coefficient, and motion amplitude enhancement coefficient of the AI-simulated elderly person with dementia are multiplied with the target adjustment coefficient to obtain the updated target behavior intensity parameter, updated voice feedback enhancement coefficient, and updated motion amplitude enhancement coefficient simultaneously.

Images