A multimodal dementia risk reduction intervention system that promotes social engagement

Abstract

The application discloses a multi-modal dementia risk reduction intervention system for promoting social participation, and belongs to the technical field of dementia intervention, and specifically comprises: a user behavior collection module, which is used for collecting and correcting a basic behavior set of a user; a behavior symbol conversion module, which is used for converting the basic behavior set into a behavior code sequence composed of standardized behavior symbols; an intervention path generation module, which is used for generating the behavior code sequence into an intervention instruction set, extracting corresponding interactive materials from a multi-modal content library, and generating collaborative intervention content; a real-time behavior monitoring module, which is used for pushing the collaborative intervention content to the user and monitoring real-time behavior feedback of the user; and a behavior expectation comparison module, which is used for calculating a difference degree between the real-time behavior feedback and a preset behavior expectation, and triggering reconfiguration of the intervention instruction set when the difference degree is greater than a preset threshold value; and the application realizes the precision and dynamic of social participation promotion for elderly dementia patients.

Description

Technical Field

[0001] This invention relates to the field of dementia intervention technology, specifically to a multimodal dementia risk reduction intervention system that promotes social participation. Background Technology

[0002] With the accelerating aging of the global population, dementia prevention has become a public health priority. Epidemiological studies have confirmed that sustained social participation can effectively maintain cognitive reserve and reduce the risk of neurodegenerative diseases. The World Health Organization has clearly stated that promoting social connection and multi-dimensional activity participation among middle-aged and older adults is a core strategy for dementia risk intervention. Current mainstream medical opinion emphasizes the need to construct comprehensive intervention programs that integrate physical activity, social interaction, and cognitive stimulation to achieve multi-target synergistic activation of neural pathways.

[0003] In the existing technology, computer-assisted dementia risk intervention programs are mainly developing in two directions: one type adopts a standardized cognitive training paradigm, providing structured training through pre-set memory games or logic task software; the other type adopts a behavior monitoring trigger mechanism, pushing general suggestions when sensors detect specific physiological indicators.

[0004] However, standardized cognitive training tasks are disconnected from real-world social scenarios, making it difficult to reflect an individual's behavioral interaction patterns in daily life and unable to be adjusted in real time according to individual behavioral patterns. For example, forcing socially avoidant individuals to participate in group activities may trigger resistance. Furthermore, intervention dimensions such as physical activities, social interactions, and cognitive stimulation lack organic integration. For instance, simple VR social training is not combined with real-world mobile tasks, resulting in low user engagement. Moreover, the evaluation of intervention effectiveness relies heavily on periodic clinical scales and cannot dynamically reconstruct the intervention path based on behavioral feedback, leading to a lag in strategy adjustments.

[0005] These limitations make it difficult for existing systems to adapt to the spatiotemporal variability of individual behavioral patterns, and also prevent them from achieving synergistic optimization of multimodal intervention elements. Summary of the Invention

[0006] The purpose of this invention is to provide a multimodal dementia risk reduction intervention system that promotes social participation, and to solve the following technical problems:

[0007] Existing dementia intervention systems are ill-suited to the spatiotemporal variability of individual behavioral patterns.

[0008] The objective of this invention can be achieved through the following technical solutions:

[0009] A multimodal dementia risk reduction intervention system that promotes social participation includes:

[0010] The user behavior collection module is used to collect and correct the user's basic behavior set, including the user's displacement records in physical space, interaction frequency records in social scenarios, and multimedia content contact records.

[0011] The behavior symbol conversion module is used to convert a basic set of behaviors into a behavior code sequence composed of standardized behavior symbols through a preset standardized symbol library. Each behavior symbol corresponds to a set social participation behavior pattern.

[0012] The intervention path generation module is used to generate an intervention instruction set that includes a symbol substitution sequence and a symbol insertion sequence from the behavior encoding sequence. Based on the intervention instruction set, it extracts corresponding interactive materials from the multimodal content library to generate collaborative intervention content with spatial movement guidance, social task guidance, and media content guidance.

[0013] The real-time behavior monitoring module is used to push collaborative intervention content to users through portable terminals, while monitoring real-time behavioral feedback during the user's execution process.

[0014] The behavior expectation comparison module is used to calculate the difference between real-time behavior feedback and the behavior expectations preset in the intervention instruction set. When the difference is greater than the preset threshold, the intervention path generation module is triggered to reconfigure the intervention instruction set.

[0015] As a further aspect of the present invention: the process of modifying the user's basic behavior set in the user behavior profile is as follows:

[0016] The meteorological data interface is integrated into the displacement record, and the movement distance threshold corresponding to the distance symbol is automatically lowered when the rainfall intensity exceeds the rainfall intensity threshold.

[0017] Acoustic environment analysis is introduced into the interaction frequency recording to reduce the allocation priority of dialogue symbols when the background noise decibel exceeds the noise tolerance threshold.

[0018] Light sensor data is integrated into the multimedia touch content recording to block the generation request of text symbols when the ambient illuminance is lower than the reading illuminance standard;

[0019] Calculate the timestamp overlap rate of the above three types of records. If the timestamp overlap rate is higher than the overlap rate threshold, input the corrected behavior set into the behavior symbol conversion module; otherwise, verify the timestamps.

[0020] As a further aspect of the present invention: the standardized symbol library in the behavior symbol conversion module specifically includes:

[0021] The standardized symbol library includes a spatial displacement symbol set, a social interaction symbol set, and a media contact symbol set; the spatial displacement symbol set includes distance symbols representing different movement distance thresholds, dwell symbols representing different dwell time thresholds, and speed symbols representing different movement speed thresholds.

[0022] The set of social interaction symbols includes dialogue symbols representing face-to-face conversations, collaboration symbols representing group activities, and stranger symbols representing contact between strangers.

[0023] The media contact symbol set includes text symbols for reading text, auditory symbols for listening to audio, and visual symbols for viewing images.

[0024] As a further aspect of the present invention: in the behavior symbol conversion module, the process of converting the basic behavior set into a behavior encoding sequence composed of standardized behavior symbols is as follows:

[0025] Acquire the user's displacement record in the physical space. When the continuous movement distance exceeds the preset distance threshold, assign the corresponding distance symbol. When the dwell time at a specific location reaches the dwell time threshold, assign the dwell symbol. When the movement speed is lower than the speed threshold, assign the low speed symbol.

[0026] Event recognition is performed on the frequency records of user interactions in social scenarios. When a two-way language communication event is recognized, a dialogue symbol is assigned; when three or more people are detected to be engaged in activities together, a collaboration symbol is assigned; and when contact with an object whose identity information is not stored is captured, a stranger symbol is assigned.

[0027] The media type of the user's multimedia access content is determined. When the access content is written text, a text symbol is assigned; when it is pure audio information, an auditory symbol is assigned; and when it is dynamic image, a visual symbol is assigned.

[0028] All assigned symbols are arranged in chronological order to form a behavior encoding sequence with timestamps.

[0029] As a further aspect of the present invention: the process of generating the intervention instruction set in the intervention path generation module is as follows:

[0030] The system retrieves a symbol association rule base and a symbol priority weight table. The symbol association rule base stores spatial proximity, social complementarity, and media compatibility relationships between behavioral symbols. The symbol priority weight table records the quantitative value of the intensity of neural function stimulation for each type of symbol, which is divided into high weight, medium weight, and low weight.

[0031] Scan the behavior encoding sequence and mark low-frequency symbols, which are symbols that appear less frequently than the average number of symbols of the same type within a set time period; retrieve target symbols that are strongly associated with low-frequency symbols in the symbol association rule base, where the strong association is that the spatial proximity score, social complementarity score, and media compatibility score are all higher than the corresponding thresholds; generate a symbol insertion sequence containing the target symbols, with the insertion position determined based on the temporal gap density in the behavior encoding sequence;

[0032] High-risk symbol combinations in the behavior encoding sequence are identified. A high-risk symbol combination refers to a symbol segment in which the number of consecutively appearing symbols of the same type exceeds a type continuity threshold and the total priority weight is lower than a weight threshold. A symbol replacement sequence is generated, and a set proportion of the original symbols in the high-risk symbol combination are replaced with high-weight symbols. The replacement position is preferably selected from the symbols in the middle of the time sequence in the high-risk symbol combination.

[0033] As a further aspect of the present invention: the mechanism for identifying high-risk symbol combinations in the behavior encoding sequence is as follows:

[0034] Record the magnitude of changes in the user's EEG characteristics and the score of the memory recall test the next day after each execution of a symbol combination; when a new symbol combination meets the requirement that both the magnitude of changes in EEG characteristics and the score of the memory recall test are lower than the corresponding benchmark values, the symbol combination is included in the high-risk symbol combination library.

[0035] As a further aspect of the present invention: in the intervention path generation module, the generation process of the collaborative intervention content is as follows:

[0036] Spatial movement guidance consists of a geofence coordinate set and a movement path animation. The geofence coordinate set generates a ring-shaped activity area based on the movement distance threshold corresponding to the distance symbol in the symbol insertion sequence. The movement path animation draws a recommended route from the current location to the target geofence.

[0037] The social task guidance adopts a dual-track system of role dialogue scripts and collaborative task lists. The role dialogue scripts generate a virtual dialogue tree containing open-ended questions based on dialogue symbols, while the collaborative task lists are designed based on collaborative symbols to complete physical item transfer tasks that require multiple people to cooperate.

[0038] Media content guides the implementation of cross-media mapping strategies, generating illustrated news summaries when the target symbol is a text symbol, generating audio clips with subtitles when the target symbol is an auditory symbol, and generating interactive dynamic infographics when the target symbol is a visual symbol.

[0039] All collaborative intervention content is arranged in symbolic order of the intervention instruction set on the timeline. Spatial movement guidance and social task guidance have overlapping periods, while media content guidance runs through the entire intervention cycle.

[0040] As a further aspect of the present invention: in the real-time behavior monitoring module, the process of pushing collaborative intervention content to the user via a portable terminal is as follows:

[0041] Spatial movement guidance is based on augmented reality technology, which overlays geofence boundary lines and movement direction indicator arrows on the real-time image of the terminal camera. When the user enters the geofence area, a visual confirmation signal is triggered.

[0042] Social tasks are guided to implement dual verification. Dialogue tasks use microphones to collect voice responses and perform keyword matching verification, while collaboration tasks complete contact verification by scanning a QR code generated by the other user's terminal.

[0043] Media content guidance is based on user gaze detection. When the front-facing camera detects that the user's gaze has deviated from the screen by more than the angle of deviation, the content playback is paused until the user's gaze refocuses on the screen and reaches the required duration of sustained gaze, after which playback resumes.

[0044] As a further aspect of the present invention: the specific process of difference calculation and reconfiguration in the behavior expectation comparison module is as follows:

[0045] Extract symbol response delay data and symbol execution completeness data from real-time behavior feedback. Symbol response delay data records the time difference from pushing collaborative intervention content to detecting the corresponding behavior symbol. Symbol execution completeness data records the feature matching degree between the actual generated symbol and the expected symbol.

[0046] When the symbol response delay exceeds the delay tolerance threshold, a delay compensation reconfiguration is triggered, and a transition symbol is added before the symbol insertion sequence. The transition symbol has a strong correlation with the target symbol and its priority weight is lower than that of the target symbol.

[0047] When the symbol execution completeness is lower than the completeness qualification threshold, complexity degradation reconfiguration is triggered, replacing high-weight symbols in the symbol replacement sequence with medium-weight symbols of the same type, while reducing the number of target symbols in the symbol insertion sequence;

[0048] The reconfigured intervention instruction set retains the total time length of the original instruction set, and the total length is kept unchanged by extending the execution period of a single symbol or adding parallel symbol execution channels.

[0049] The beneficial effects of this invention are:

[0050] (1) This invention establishes a digital archive of an individual’s daily behavior trajectory and implements behavior symbolization conversion, transforming three types of raw data—physical displacement, social contact, and media interaction—into a standardized sequence of behavioral symbols with social participation attributes, thereby realizing the digital mapping of complex behavioral patterns in real social scenarios and eliminating the semantic gap between standardized cognitive training tasks and real social activities.

[0051] (2) This invention relies on the dynamic intervention path generation unit to construct a symbol association rule base and priority weight table. Based on the identification of low-frequency symbols and high-risk symbol combinations, it generates symbol insertion sequences and symbol replacement sequences. Based on this, it extracts three-dimensional elements of spatial movement guidance, social task guidance and media content guidance from the multimodal content library. Through the time axis coordination of the geofence coordinate set and the spatial positioning of the movement path animation, the social drive of the role dialogue script and the collaborative task list, and the cognitive activation of the cross-media mapping strategy, it realizes the organic integration of physical activities, social interaction and cognitive stimulation in the real environment, and overcomes the participation defect of virtual training and real movement being disconnected.

[0052] (3) Based on real-time behavioral feedback data monitored by a portable terminal, this invention triggers the reconfiguration of the intervention instruction set by calculating the difference between symbol response delay and symbol execution completeness. Combined with an environmental parameter correction module, it dynamically adjusts the symbol allocation logic by integrating meteorological, acoustic, and optical data, forming a closed-loop response mechanism to replace the traditional periodic clinical scale assessment model. This ensures that the intervention strategy is dynamically optimized at the minute level according to individual behavioral patterns and environmental changes. Through environmentally adapted data processing, multi-dimensional collaborative intervention, and a closed-loop feedback mechanism, this invention achieves precision and dynamism in promoting social participation in Alzheimer's patients. Attached Figure Description

[0053] The invention will now be further described with reference to the accompanying drawings.

[0054] Figure 1 This is a schematic diagram of the modules of the present invention. Detailed Implementation

[0055] 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.

[0056] Please see Figure 1 As shown, this invention is a multimodal dementia risk reduction intervention system that promotes social participation, comprising:

[0057] The user behavior collection module uses sensors to acquire physical displacement, social interaction, and multimedia contact data. It combines environmental parameters such as meteorology, acoustics, and lighting to correct the data. For example, it lowers the movement distance threshold when it rains, reduces the priority of dialogue symbols when noise exceeds the standard, and blocks the generation of text symbols when there is insufficient light. It also ensures data consistency through timestamp overlap rate verification.

[0058] The behavior symbol conversion module is based on a standardized symbol library. It converts displacement records into spatial symbols according to distance, dwell time, and speed; social interactions into social symbols according to dialogue, collaboration, and stranger contact; and multimedia contacts into media symbols according to text, audio, and video. It generates a behavior coding sequence with timestamps in chronological order.

[0059] The intervention path generation module uses a symbol association rule base and priority weight table to mark low-frequency symbols and retrieve strongly associated target symbols to generate insertion sequences. It also identifies high-risk combinations of consecutive similar symbols exceeding a threshold and with insufficient total weight to generate replacement sequences. Based on this, materials are extracted from a multimodal content library to generate collaborative intervention content including geofence navigation, social task scripts, and cross-media content. Spatial movement and social task guidance are pushed synchronously during overlapping periods, with media content integrated throughout the entire process.

[0060] The real-time behavior monitoring module pushes intervention content via a portable terminal. Spatial movement guidance uses AR overlays of fence boundaries and directional arrows. Social tasks are verified through voice keyword matching and QR code scanning. Media content is paused based on gaze detection. The behavior expectation comparison module calculates the difference between response latency and execution completeness. When thresholds are exceeded, reconfiguration is triggered, such as inserting transition symbols for latency compensation and replacing weighted symbols in complexity downgrading. Dynamic optimization is achieved by adjusting the execution period to maintain the intervention duration.

[0061] In a preferred embodiment of the present invention, the process of modifying the user's basic behavior set in the user behavior profile is as follows:

[0062] For physical spatial displacement records, the system obtains rainfall intensity information in real time through a meteorological data interface. When the rainfall intensity exceeds a preset threshold, the system automatically adjusts the movement distance judgment standard corresponding to the distance symbol. For example, 500 meters is usually used as the distance threshold for moderate-intensity activities, but this threshold can be dynamically lowered to 200 meters during rainfall to avoid misjudging the user's activity level due to weather factors.

[0063] In the social interaction frequency recording stage, the system incorporates an acoustic environment analysis module. This module collects background noise decibel levels in real time via the terminal device's microphone. When the noise level exceeds the human tolerance threshold for comfortable communication, the system automatically reduces the priority of dialogue symbols. This mechanism ensures that brief and ineffective voice interactions in high-noise environments such as square dancing are not misjudged as valid social behavior, thus improving the validity of the data.

[0064] For recording multimedia content, the system integrates light sensor data from the terminal device to continuously monitor ambient illuminance levels. When the ambient illuminance is detected to be lower than the standard reading illuminance, it automatically blocks requests for generating text symbols to avoid recording text browsing in low-light environments as valid cognitive stimuli, while prioritizing the triggering of auditory or visual symbol generation logic.

[0065] After completing the environmental correction for the three types of data, the timestamp overlap rate of displacement, social interaction, and multimedia contact records is further calculated. By comparing the consistency of the timestamps of the three types of data, it is determined whether they belong to coordinated behaviors occurring at the same time. If the overlap rate is higher than a preset threshold, it indicates that the data has spatiotemporal correlation and can be input into the behavior symbol conversion module; if the overlap rate is insufficient, the timestamp verification process is initiated, and time deviations are corrected through mechanisms such as sensor clock synchronization calibration and data transmission delay compensation to ensure the accuracy of subsequent behavior analysis.

[0066] In another preferred embodiment of the present invention, the standardized symbol library in the behavior symbol conversion module specifically includes:

[0067] This system leverages a standardized symbol library to achieve semantic mapping of user behavior. The library employs a three-layer classification system to construct a behavioral semantic network. The spatial displacement symbol set of the standardized symbol library is classified based on the spatial attributes of physical activities. Distance symbols are divided into multiple granularities according to movement distance thresholds; for example, 500 meters of continuous movement is set as the basic distance symbol, and 1000 meters as the high-intensity distance symbol, thus indicating differences in user activity intensity. Dwelling symbols are determined by the duration of stay at a specific location; for example, staying more than 30 minutes at a community activity center is assigned a long-stay symbol, and less than 10 minutes is assigned a short-stay symbol, distinguishing between passive and active social dwelling. Speed ​​symbols include low, medium, and high speed levels; a low-speed symbol is assigned when the user's walking speed is below 1.5 km / h, and a high-speed symbol is assigned when it is above 5 km / h, achieving a refined description of movement status.

[0068] The social interaction symbol set focuses on the scene characteristics and relationship attributes of interpersonal interaction. The allocation of dialogue symbols must meet the condition of two-way language communication. The system uses speech recognition technology to detect turn transitions in the dialogue. When both parties speak at least twice in a single exchange, it is judged as a valid dialogue and a symbol is allocated. Collaboration symbols are for joint activities of three or more people, such as community painting groups or square dance teams. Video analysis technology is used to identify group collaborative behavior patterns. Symbol allocation is triggered when task division or synchronized actions are detected among members. Stranger symbols are used to mark contact behavior with objects whose identity information is not stored. The system uses facial recognition technology to compare with the local identity database. When the interaction object is not in the preset contact list, a stranger symbol is allocated regardless of the duration of the exchange, in order to quantify the social participation of non-acquaintances.

[0069] The media contact symbol set is divided according to the medium characteristics of the information carrier. Text symbols cover various written text contact behaviors, including paper books, electronic documents, and community bulletin board text. The system identifies pure text content through file type detection and content analysis technology. Auditory symbols correspond to pure audio information contact scenarios, such as radio programs, voice chat, and audiobooks. Multimedia content containing video is excluded through audio spectrum analysis. Visual symbols are for dynamic image carriers, including movies, short videos, and animations. They are determined by detecting frame sequence changes and visual element motion characteristics in the content. Static image contact is not included in this symbol category.

[0070] In a preferred embodiment, the process of converting the basic set of behaviors into a behavior encoding sequence composed of standardized behavior symbols in the behavior symbol conversion module is as follows:

[0071] The symbol assignment for displacement records adopts a threshold trigger mechanism. The system calculates the cumulative movement distance of the GPS trajectory in real time. When the continuous movement distance exceeds the 500-meter threshold, the corresponding distance symbol is automatically assigned. At the same time, a dwell time timer is started. If the dwell time at a certain geographic coordinate point exceeds 15 minutes, a dwell symbol is assigned. The speed determination is based on the displacement change per unit time. The instantaneous speed is calculated after smoothing the positioning data through the Kalman filter algorithm. If the speed is lower than the threshold, a low speed symbol is assigned.

[0072] Event recognition of social interaction records integrates audio energy detection and visual posture analysis. Two-way language communication events must simultaneously satisfy the alternating changes in voice signal energy and facial orientation interaction features. For group activities of three or more people, the number of human key points in the same scene is counted through human posture estimation technology. When the threshold is exceeded, collaborative symbol allocation is triggered. Stranger contact recognition combines facial recognition confidence and interaction distance sensor data. When a non-acquaintance enters within a 1.5-meter social distance, a stranger symbol is assigned.

[0073] The determination of the media type of multimedia contact recording adopts a multi-level detection strategy. First, the file extension is used to initially distinguish between text, audio and video formats. Then, the content is analyzed in depth. Text files must contain recognizable sequences of text characters. Audio files must exclude video tracks and the spectrum must conform to speech or music characteristics. The dynamic images corresponding to visual symbols must meet the requirement of more than 12 frames per second and pixel displacement changes.

[0074] Each symbol represents a behavior related to participation in social activities. For example, the distance symbol corresponds to going out and engaging with the outside world; the dialogue symbol corresponds to interpersonal communication and emotional interaction; and the collaboration symbol corresponds to group cooperation and the establishment of social relationships.

[0075] All assigned symbols are sorted by the timestamp of the event, accurate to the second level, forming a behavioral coding sequence with a time dimension, such as a temporal structure of "[T1: distance symbol 500 meters], [T2: dialogue symbol], [T3: visual symbol short video]", providing standardized behavioral semantic input for subsequent intervention path generation. This symbol conversion mechanism maps raw behavioral data into semantic units with social participation attributes through multi-dimensional feature detection and thresholding, achieving an abstract upgrade from physical signals to behavioral patterns.

[0076] In another preferred embodiment of the present invention, the process of generating the intervention instruction set in the intervention path generation module is as follows:

[0077] The intervention path generation module constructs a decision core through a symbol association rule base and a priority weight table to achieve intelligent generation of intervention instruction sets. The symbol association rule base uses a triplet structure to store the associations between behavioral symbols, including spatial proximity, social complementarity, and media compatibility. Spatial proximity describes the co-occurrence probability of the corresponding behavior in physical space; for example, a co-occurrence probability exceeding 70% between "500 meters away from the symbol" and "dialogue symbol" in a community park scene indicates strong spatial proximity. Social complementarity quantifies the synergistic effect of different social symbols on social participation; for example, the combination of "collaboration symbol" and "stranger symbol" can increase social network expansion efficiency by 40%. Media compatibility assesses the complementarity of different media symbols in information transmission; for example, the combination of "text symbol" and "auditory symbol" can increase information retention rate by 30%. The symbol priority weight table, based on neuroscience research, quantifies the stimulation intensity of symbols on brain regions such as the prefrontal cortex and hippocampus into high, medium, and low levels. "Collaboration symbols" are given high weight due to their involvement in multi-brain region co-activation, while "staying symbols" are given low weight due to their lower stimulation intensity.

[0078] The module first scans the behavior encoding sequence for low-frequency symbols. It calculates the frequency of each symbol within a set period, such as 7 days, using a sliding time window, and marks symbols with a frequency lower than 60% of the average frequency of the same type as low-frequency symbols. For example, in a social interaction symbol set, if a "strange symbol" appears less than twice a week (less than the average of 5 times for that type), an association retrieval mechanism is triggered. In the symbol association rule base, strongly associated target symbols must simultaneously meet three conditions: a spatial proximity score of at least 0.7 (calculated based on the co-occurrence probability of both symbols in historical data); a social complementarity score of at least 0.6 (derived from an evaluation of social network expansion effects); and a media compatibility score of at least 0.5 (determined through information transmission efficiency testing). Target symbols retrieved, such as "collaboration symbols," are included in the symbol insertion sequence. The insertion position is determined using a temporal gap density algorithm: identifying periods of more than 15 consecutive minutes without symbol records in the behavior sequence, new symbols are preferentially inserted in high-density gap areas, such as the typical free time period of 2-4 pm on weekdays, to avoid disrupting the user's established circadian rhythm.

[0079] To identify high-risk symbol combinations, the module employs a dual threshold detection mechanism: first, it determines whether there are consecutive occurrences of the same type of symbols, such as three or more consecutive "staying symbols" in a spatial displacement symbol set; second, it calculates the total priority weight of the symbol segment. If the total weight is below the weight threshold (e.g., the consecutive occurrence of low-weight symbols causing the total weight to be less than 1.5), it is marked as a high-risk combination. When generating the symbol replacement sequence, a "center-priority" principle is followed: for high-risk segments containing five consecutive "staying symbols," the third symbol (i.e., the high-weight "collaboration symbol" in the middle of the time sequence) is replaced first, achieving an overall weight increase for the symbol segment with minimal intervention cost. The replacement ratio is dynamically adjusted based on the symbol segment length: 20%-30% is replaced for 5-7 symbol segments, and 30%-40% for 8 or more symbol segments, ensuring that the intervention intensity matches the risk level.

[0080] In a preferred embodiment, the mechanism for identifying high-risk symbol combinations in the behavior encoding sequence is as follows:

[0081] The amplitude changes of alpha and beta waves after a user executes a symbol combination are recorded using an EEG acquisition device such as a dry electrode EEG headband. Simultaneously, a mobile application is used to conduct a memory recall test for the following day, including episodic memory tasks such as recalling the clothing characteristics of a social contact from the previous day, and semantic memory tasks such as retelling key information from text encountered the previous day. When the amplitude of the EEG characteristics of a new symbol combination is less than 30% of the historical average and the memory test score is lower than the benchmark (e.g., a score below 6 out of 10), the combination is automatically added to the high-risk symbol combination library, triggering a rule base update process. For example, if consecutive "text symbol" combinations lead to a 20% decrease in hippocampal activation and a text content recall accuracy rate below 50%, they will be marked as high-risk, and the weight threshold for subsequent similar combinations will be automatically lowered by 15% to improve detection sensitivity.

[0082] In another preferred embodiment of the present invention, the process of generating the collaborative intervention content in the intervention path generation module is as follows:

[0083] The spatial movement guidance system constructs physical activity scenarios based on geofencing technology. The system generates a circular activity area centered on the user's current location on an electronic map based on the movement distance threshold corresponding to the distance symbols in the symbol insertion sequence. For example, a circular area with a radius of 250 meters is generated for a movement threshold of 500 meters. The movement path animation module uses real-time positioning as a starting point, combined with street view data and pedestrian path distribution, to draw a recommended route to the target geofence. The route prioritizes passing through high-frequency social locations such as community parks and convenience stores, transforming abstract distance symbols into actionable physical movement commands.

[0084] The social task guidance adopts a dual-track design to adapt to the needs of different social scenarios. The role-based dialogue script module generates a virtual dialogue tree containing open-ended questions based on dialogue symbols, such as "What changes have you noticed in the community's recent greenery?" Users can interact via voice or text input, and the dialogue tree depth is dynamically adjusted based on users' historical social data. The collaborative task list module designs physical item delivery tasks for collaborative symbols, such as "Give the brochure from the community activity center to the volunteer wearing the blue shirt." The task includes the target object's characteristics, a description of the item to be delivered, and the completion location. Users need to cooperate with others to complete this task, and the system detects the collaborative contact status through near-field communication or Bluetooth signals.

[0085] Media content guidance implements a cross-modal mapping strategy to achieve multi-sensory information delivery. When the target symbol is a text symbol, the system generates a news summary containing 3-5 images, with the word count of a single article controlled within 200 words; when it is an auditory symbol, it generates a 5-10 minute audio clip with subtitles, and the font size of the subtitles is dynamically adjusted according to the ambient light; when it is a visual symbol, it generates an interactive dynamic infographic, and users can click on hotspot areas to view detailed information, such as clicking on a venue icon in a community activity map to display opening hours, etc.

[0086] In terms of timeline arrangement, all collaborative intervention content is arranged according to the symbolic order of the intervention instruction set. Spatial movement guidance and social task guidance overlap in geographically fenced areas. For example, when a user arrives at a community square, collaborative tasks within the square are automatically triggered, achieving scenario coupling between physical movement and social interaction. Media content guidance runs throughout the entire intervention cycle, pushing auditory content while walking and displaying visual infographics during social breaks to create continuous cognitive stimulation.

[0087] In a preferred embodiment, the process of pushing collaborative intervention content to the user via a portable terminal in the real-time behavior monitoring module is as follows:

[0088] Spatial movement guidance is based on augmented reality technology. A semi-transparent geofence boundary line and an arrow-shaped movement indicator are superimposed on the real-time image of the terminal camera. When the user enters the fenced area, a flashing green light effect is triggered at the edge of the screen as a visual confirmation signal. The duration of the light effect is related to the preset duration of the stay symbol.

[0089] Social tasks employ a dual verification mechanism to ensure the effectiveness of interactions. For dialogue tasks, voice responses are captured via microphone, and keywords are extracted using natural language processing. A match of 60% or higher with the pre-set keyword library of the dialogue script is considered a valid interaction. For collaborative tasks, users are required to scan a dynamic QR code generated by the other party's device. This QR code contains the task ID and the identity of the collaborating partner. Upon successful scanning, the system uses GPS location data to verify whether the interaction was completed within a designated area.

[0090] Media content guidance integrates eye-tracking technology to optimize information reception efficiency. The terminal's front-facing camera monitors the direction of the user's gaze using an iris detection algorithm. If the gaze deviates from the screen by more than 15 degrees for more than 3 seconds, the content playback is automatically paused, and a semi-transparent prompt box is displayed on the screen. When the user's gaze refocuses on the screen and remains focused for more than 2 seconds, the content resumes playback from the pause point, improving the effective reach of media content.

[0091] In another preferred embodiment of the present invention, the specific process of difference calculation and reconfiguration in the behavior expectation comparison module is as follows:

[0092] First, two types of key data are extracted from real-time behavioral feedback: symbol response delay data records the time difference between the push of collaborative intervention content and the detection of the corresponding behavioral symbol. For example, after pushing a community activity task corresponding to a "collaboration symbol", the time interval between the user's actual participation in the activity is recorded. Symbol execution completeness data evaluates the similarity between the actual generated symbol and the expected symbol through a feature matching algorithm. For example, for a reading task expected to be a "text symbol", the system will analyze whether the actual content encountered contains text features and whether the reading time meets the standard.

[0093] When the symbol response delay exceeds a preset delay tolerance threshold, the system triggers a delay compensation reconfiguration mechanism. This mechanism adds a transition symbol before the symbol insertion sequence. The selection of the transition symbol follows two principles: it must have a strong correlation with the target symbol to ensure the continuity of behavioral guidance; and its priority weight must be lower than the target symbol to avoid overstimulating the user. For example, when a user experiences a delay in responding to a social task corresponding to an "unfamiliar symbol," the system will insert a simple greeting task corresponding to a "dialogue symbol," such as "asking the community security guard about today's weather," before the original task. This type of transition task belongs to the same set of social interaction symbols as the target task, but has lower cognitive complexity, helping the user gradually enter the state.

[0094] If the symbol execution completeness falls below the acceptable threshold, the system initiates a complexity downgrade and reconfiguration. Specific measures include replacing high-weight symbols in the symbol replacement sequence with medium-weight symbols of the same type, while reducing the number of target symbols in the symbol insertion sequence. For example, if the original plan was to replace consecutively appearing "staying symbols" with group activities corresponding to "collaboration symbols," but the user's execution completeness is insufficient, the system will downgrade the group activity to a two-person collaborative task, such as "organizing the community bookshelf with a neighbor," and reduce the insertion frequency of similar tasks. This adjustment maintains consistency in the direction of intervention while reducing execution difficulty, aligning with the user's current ability level.

[0095] The reconfigured intervention instruction set strictly retains the total time length of the original instruction set, achieving time balance through two strategies. When the execution time of a single symbol is extended, the system intelligently compresses the execution time of other symbols. For example, the reading task of "text symbols" is extended from 15 minutes to 20 minutes, while the viewing task of subsequent "visual symbols" is shortened from 10 minutes to 5 minutes. If parallel symbol execution channels are added, the system will overlay low-complexity tasks within the same time window. For example, when the user listens to the broadcast of "auditory symbols," the simple hand movement task corresponding to the "staying symbol" is executed simultaneously, maximizing time utilization.

[0096] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A multi-modal dementia risk reduction intervention system to promote social engagement, characterized in that, The application relates to a social intervention system and method. The system comprises: a user behavior collection module for collecting and correcting a basic behavior set of a user, including displacement records of the user in a physical space, interaction frequency records in a social scene and multimedia contact content records; a behavior symbol conversion module for converting the basic behavior set into a behavior code sequence composed of standardized behavior symbols through a preset standardized symbol library, each behavior symbol corresponding to a set social participation behavior mode; an intervention path generation module for generating the behavior code sequence into an intervention instruction set containing a symbol replacement sequence and a symbol insertion sequence, extracting corresponding interaction materials from a multi-modal content library according to the intervention instruction set, and generating collaborative intervention content with space movement guidance, social task guidance and media content guidance; a real-time behavior monitoring module for pushing the collaborative intervention content to the user through a portable terminal and monitoring real-time behavior feedback of the user in an execution process; a behavior expectation comparison module for calculating a difference degree between the real-time behavior feedback and a behavior expectation preset in the intervention instruction set, and triggering the intervention path generation module to reconfigure the intervention instruction set when the difference degree is greater than a preset threshold value; In the behavior symbol conversion module, the standardized symbol library specifically comprises: The standardized symbol library contains a space displacement symbol set, a social interaction symbol set and a media contact symbol set; the space displacement symbol set contains distance symbols representing different moving distance thresholds, stay symbols representing different stay time length thresholds and speed symbols representing different moving speed thresholds; the social interaction symbol set contains dialogue symbols representing face-to-face conversation, cooperation symbols representing group activities and stranger symbols representing stranger contact; the media contact symbol set contains text symbols representing text reading, hearing symbols representing audio listening and vision symbols representing image watching; In the intervention path generation module, the generation process of the intervention instruction set is as follows: a symbol association rule library and a symbol priority weight table are acquired, the symbol association rule library stores spatial proximity relationships, social complementary relationships and media compatible relationships between behavior symbols; the symbol priority weight table records quantized values of each type of symbol on neural function stimulation intensity, and is divided into high weight, medium weight and low weight; low-frequency symbols in the behavior code sequence are scanned and marked, the low-frequency symbols refer to symbols with an appearance frequency lower than an average frequency of the same type of symbols in a set time period; target symbols with strong association with the low-frequency symbols are searched in the symbol association rule library, the strong association refers to that spatial proximity relationship scores, social complementary relationship scores and media compatible relationship scores are all higher than corresponding threshold values; a symbol insertion sequence containing the target symbols is generated, and an insertion position is determined according to a time gap density in the behavior code sequence; high-risk symbol combinations in the behavior code sequence are identified, the high-risk symbol combinations refer to symbol segments with a same type of symbol number exceeding a type continuous threshold value and a priority weight total value lower than a weight threshold value; a symbol replacement sequence is generated, high weight symbols are used to replace original symbols in a set proportion in the high-risk symbol combinations, and replacement positions are preferentially selected from symbol positions in a time sequence in the high-risk symbol combinations.

2. A multi-modal dementia risk reduction intervention system that facilitates social engagement according to claim 1, wherein, The process of modifying the basic behavior set of the user in the user behavior collection module is: In the displacement record, the meteorological data interface is fused, and the moving distance threshold corresponding to the distance symbol is automatically adjusted to be lower when the rainfall intensity exceeds the rainfall intensity threshold; In the interaction frequency record, the acoustic environment analysis is introduced, and the allocation priority of the dialogue symbol is reduced when the background noise decibel exceeds the noise tolerance threshold; In the multimedia contact content record, the light sensor data is integrated, and the generation request of the text symbol is shielded when the ambient illuminance is lower than the reading illuminance standard; The timestamp overlap rate of the above three records is calculated, and if the timestamp overlap rate is higher than the overlap rate threshold, the modified behavior set is input into the behavior symbol conversion module, otherwise the timestamp is checked.

3. The multi-modal dementia risk reduction intervention system that facilitates social engagement of claim 1, wherein, In the behavior symbol conversion module, the process of converting the basic behavior set into a behavior code sequence composed of standardized behavior symbols is: Obtain the displacement record of the user in the physical space, assign the corresponding distance symbol when detecting that the continuous moving distance exceeds the preset distance threshold, assign the stay symbol when the stay time at a specific location reaches the stay time threshold, and assign the low-speed symbol when the moving speed is lower than the speed threshold; Perform event recognition on the interaction frequency record in the user's social scene, assign the dialogue symbol when recognizing the two-way language exchange event, assign the collaboration symbol when detecting that three or more people are participating in the activity together, and assign the stranger symbol when capturing the contact with an object whose identity information is not stored; Determine the medium type of the multimedia contact content record of the user, assign the text symbol when the contact content is written text, assign the auditory symbol when the contact content is pure audio information, and assign the visual symbol when the contact content is dynamic image; Arrange all the assigned symbols in chronological order to form a behavior code sequence with timestamp markers.

4. The multi-modal dementia risk reduction intervention system that facilitates social engagement of claim 1, wherein, The mechanism for identifying high-risk symbol combinations in the behavior code sequence is: Record the amplitude of the user's brain wave feature change after each execution of the symbol combination and the score of the memory backtracking test the next day; when a new symbol combination meets the criteria that both the amplitude of the brain wave feature change and the score of the memory backtracking test are lower than the corresponding baseline values, the symbol combination is included in the high-risk symbol combination library.

5. The multi-modal dementia risk reduction intervention system that facilitates social engagement of claim 1, wherein, In the intervention path generation module, the generation process of the collaborative intervention content is: The space movement guide is composed of a geographic fence coordinate set and a movement path animation, and the geographic fence coordinate set generates a ring-shaped activity area according to the moving distance threshold corresponding to the distance symbol in the symbol insertion sequence, and the movement path animation draws a recommended route from the current location to the target geographic fence; The social task guide adopts a dual-track system of role dialogue scripts and collaboration task lists, the role dialogue script generates a virtual dialogue tree containing open-ended questions based on the dialogue symbol, and the collaboration task list designs an entity item delivery task that requires multiple people to cooperate to complete; The media content guide implements a cross-media mapping strategy, generates a news digest with a picture when the target symbol is a text symbol, generates a captioned audio segment when the target symbol is an auditory symbol, and generates an interactive dynamic information graph when the target symbol is a visual symbol; All the collaborative intervention contents are arranged in the time axis according to the symbolic order of the intervention instruction set, the space movement guide and the social task guide have a period of position overlap, and the media content guide runs through the entire intervention period.

6. A multi-modal dementia risk reduction intervention system that facilitates social engagement according to claim 5, wherein, In the real-time behavior monitoring module, the process of pushing the collaborative intervention content to the user through the portable terminal is as follows: The space movement guide is based on augmented reality technology, and the geographic fence boundary line and the movement direction indication arrow are superimposed in the real-time picture of the terminal camera, and a visual confirmation signal is triggered when the user enters the geographic fence area; The social task guide implements double verification, the dialogue task is verified by keyword matching through voice reply collected by the microphone, and the cooperation task is verified by scanning the two-dimensional code generated by the terminal of the other user; The media content guide is based on user gaze detection, and the content playback is paused when the user's gaze deviates from the screen by more than the gaze deviation angle detected by the front camera, and the playback is resumed after the user's gaze refocuses on the screen and the continuous gaze duration is reached.

7. The multi-modal dementia risk reduction intervention system that facilitates social engagement of claim 1, wherein, In the behavior expectation comparison module, the specific process of difference calculation and reconfiguration is as follows: Extract the symbolic response delay data and the symbolic execution completeness data in the real-time behavior feedback, the symbolic response delay data records the time difference from pushing the collaborative intervention content to detecting the corresponding behavior symbol, and the symbolic execution completeness data records the feature matching degree between the actual generated symbol and the expected symbol; When the symbolic response delay exceeds the delay tolerance threshold, trigger delay compensation reconfiguration, add a transition symbol before the symbolic insertion sequence, the transition symbol has a strong association with the target symbol and a lower priority weight than the target symbol; When the symbolic execution completeness is lower than the completeness qualified threshold, trigger complexity degradation reconfiguration, replace the high-weight symbol in the symbol replacement sequence with a medium-weight symbol of the same type, and reduce the number of target symbols in the symbolic insertion sequence; The reconfigured intervention instruction set retains the total length of the original instruction set, and the total length remains unchanged by prolonging the execution period of a single symbol or increasing the parallel symbol execution channel.

Images