Electronic photo frame-based home-based elderly health abstract credibility analysis method and system

Abstract

This invention discloses a method and system for credibility analysis of home-based elderly care health summaries based on electronic photo frames, relating to the field of multi-source data fusion technology. This invention retrieves health records and knowledge bases based on event objects to generate display requirement objects; generates health summary text and a set of interactive controls according to the display requirements, forming a display payload; generates a display queue based on device status and determines the content to be pushed; loads and renders a template to complete the display; collects touch and voice behaviors within the interactive window, generates a unified interaction record, and reports any non-response; extracts features from touch, voice, and physiological signals, calculates signal quality scores, and generates a comprehensive confirmation confidence level; summarizes the signal quality scores, calculates the global mean, compares it with a threshold, and determines whether to execute a correction process, significantly improving the security, reliability, and user experience of the home-based elderly care health summary system.

Description

Technical Field

[0001] This invention relates to the field of multi-source data fusion technology, specifically to a method and system for credibility analysis of home-based elderly care health summaries based on electronic photo frames. Background Technology

[0002] With the development of home-based elderly care and remote health management applications, systems based on smart terminals to sense user status and trigger service responses are gradually being applied. The handling of health events usually relies on terminal devices to display prompt information, and the user clicks to confirm or provides voice feedback as the basis for event handling, thereby driving subsequent notifications or service processes. Among elderly users, due to differences in cognitive abilities and insufficient operational stability, the confirmation results obtained through a single interaction method are highly uncertain, making it difficult for the system to accurately determine whether the user has truly completed the confirmation.

[0003] Although some solutions have introduced multi-source sensing information to assist in the judgment, various signals usually participate in the decision independently, lacking a unified quantitative fusion mechanism and a dynamic weight allocation strategy for signal quality differences, thus affecting the reliability of the confirmation results.

[0004] Therefore, how to perform multi-source information fusion and credible quantification of user confirmation behavior under weak interaction conditions, and improve the accuracy and stability of decision results, has become a technical problem that needs to be solved in this field. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for credibility analysis of home-based elderly care health summaries based on electronic photo frames, so as to solve the problems raised in the prior art.

[0006] To address the aforementioned technical problems, this invention provides the following technical solution: a method for credibility analysis of home-based elderly care health summaries based on electronic photo frames, the method comprising: Step S100: Receive event data from various input sources, preprocess the event data, and generate event objects; Step S200: Based on the event object, retrieve the individual health context from the health record database, combine it with the knowledge base to retrieve and match event handling suggestions, and generate a display requirement object; Step S300: Based on the display requirements, generate health summary text for elderly users, combine the preset control configuration rules in the display template to generate an interactive control set, and generate a display payload based on the health summary text and the interactive control set; Step S400: Obtain the set of display payloads to be displayed, calculate the comprehensive priority score of the display payloads, generate a display queue based on the comprehensive priority score, and determine the display payloads to be pushed in combination with the target photo frame device status; Step S500: Load the corresponding rendering template based on the push display payload, generate the final display interface and complete the display. Collect user touch and voice interaction behaviors within the preset interaction time window, identify the interaction intent and generate a unified interaction record; when no valid interaction is detected, generate a non-response mark and report it to the platform. Step S600: Extract features from the collected touch signals, voice signals and physiological signals, calculate the quality scores of each type of signal, and generate a comprehensive confirmation confidence score; Step S700: Summarize the quality scores of each signal, calculate the global signal quality mean, compare it with the preset threshold, and determine the correction process.

[0007] Furthermore, generating the event object in step S100 includes: The input sources include home device alarms, customer service work orders, follow-up plan triggers, and family member requests. The event data from each input source is preprocessed. The preprocessing of home device alarms includes obtaining device identifier, alarm type, trigger time, and alarm level information. The alarm type and alarm level are then validated for format validity. This validation includes determining whether the alarm type belongs to a preset set of event types and whether the alarm level belongs to a preset level range. The integrity of the field format and the validity of the values ​​are also validated. Data that passes the validation is normalized for the event type, while data that fails the validation is marked as abnormal and discarded. The preprocessing of the customer service work order includes obtaining the work order identifier, request description information, submission time, and source channel information; performing semantic parsing on the request description information; the semantic parsing includes word segmentation, keyword extraction, and intent recognition based on preset semantic rules; determining the corresponding event type and urgency level based on the extracted keywords and intent recognition, combined with preset event type mapping rules and level determination rules; when there are unparsable or parsing results that do not meet preset rules, the corresponding content is filled with a preset unknown identifier, and an input credibility identifier is generated based on the matching degree of the semantic parsing results. The preprocessing for triggering the follow-up plan includes obtaining the plan identifier, object identifier, plan trigger time, and follow-up type; generating the corresponding event type based on the mapping relationship between the follow-up type and the preset event type; and filling in the corresponding emergency level based on the preset level allocation rule to form event data. The preprocessing of the family's request includes recognizing and processing the voice or text information submitted by the family. The recognition and processing includes voice-to-text processing or text normalization processing to obtain the recognized text and intent tags. Based on the recognized text and intent tags, and in combination with preset event type matching rules, the corresponding event type and urgency level are determined. Based on the matching degree of the recognition process, a recognition confidence score is generated. When the recognition confidence score is lower than a preset threshold, the input confidence score is marked as low. The preprocessed input source data is uniformly standardized. The standardization process includes normalizing the event types and unifying semantic equivalence into standard event types; filling missing fields with default values ​​according to the input source type, and filling undetermined fields with preset unknown identifiers; and deduplicating and merging data with the same source and event type within a preset time window to generate event objects. Differentiated preprocessing processes are designed for different input sources such as home device alarms, customer service work orders, follow-up plans and family requests, which can effectively improve the accuracy and adaptability of data analysis and avoid information distortion or processing deviations caused by heterogeneous data sources. By using format validity verification, semantic parsing, and rule mapping mechanisms, the system can uniformly determine event types and urgency levels. It also introduces input credibility and identification confidence indicators, which helps improve the system's ability to identify low-quality or abnormal data, thereby enhancing overall data reliability and robustness.

[0008] Furthermore, the step S200 of generating the display requirement object includes: Using the elderly identifier in the event object as an index, retrieve the individual health information of the elderly identifier from the health record database. The individual health information includes historical health data, medication records, daily routine preferences, and contact information. When no individual health information is retrieved, the relevant fields are initialized to empty and the context status is set to incomplete information. Using event type and urgency as indexes, health knowledge entries matching the event type are retrieved from a preset health knowledge base. The process of retrieving health knowledge entries includes obtaining health knowledge using an exact matching method based on event type. When no exact match is found, a fuzzy matching method based on symptom tag similarity is used. The fuzzy matching includes calculating the similarity between the symptom tags corresponding to the event and the tag set of each health knowledge entry in the knowledge base. The similarity calculation is based on tag overlap, semantic similarity, or distance between vector representations. Based on the calculated similarity results, health knowledge entries with similarity greater than a preset threshold are selected as candidate health entries. Candidate health items are sorted from high to low similarity, and combined with the event type consistency rule, candidate health items that do not match the event type are eliminated to obtain the final matched health knowledge items. The retrieved individual health information is integrated with the corresponding health knowledge entries to generate contextual information containing an individual health summary, work and rest preferences, and health knowledge entry identifiers. Sensitive fields in the generated contextual information are anonymized according to preset permission levels. Based on the contextual information and event objects, display requirement objects are generated. By retrieving historical health data, medication records, lifestyle preferences, and contact information from the health record database, the system can comprehensively grasp individual differences and maintain the integrity and controllability of the processing flow even in the event of missing information, thereby improving the robustness of the system. Based on event type and urgency, health knowledge is retrieved, and a strategy combining exact matching and fuzzy matching is adopted to improve the knowledge coverage while ensuring matching accuracy. By introducing multi-dimensional similarity calculation methods such as tag overlap, semantic similarity and vector distance, the intelligence level and adaptability of knowledge matching in complex scenarios are effectively improved. By sorting and consistency screening candidate health knowledge items, the final matching results are ensured to be both relevant and semantically consistent with the event, significantly reducing the risk of mismatches and improving the quality of knowledge recommendation.

[0009] Furthermore, generating the demonstration payload in step S300 includes: Based on the event type, individual health summary, and matching health knowledge items in the display requirements, a health summary text for elderly users is generated. The process includes: determining the corresponding display template identifier based on the event type; reading the template structure corresponding to the display template identifier, whereby the template structure includes an event prompt information field, a suggested action information field, and an individual context field; filling the event information into the event prompt information field, filling the suggested content from the health knowledge items into the suggested action information field, and filling the individual health summary and lifestyle preference information into the individual context field according to preset field mapping rules; performing statement simplification and readability adjustments on the filled text content to suit the reading habits of elderly users; filtering and verifying sensitive information in the generated health summary text; and outputting health summary content that conforms to preset security rules. According to the preset control configuration rules in the display template, a set of interactive controls associated with the health summary content is generated. The process of generating the set of interactive controls includes parsing the control type configuration and trigger condition configuration in the display template; determining the control types to be loaded based on the urgency and type of the event; binding corresponding behavior processing logic to the confirmation, delayed processing, and manual call controls respectively. The confirmation control is used to record that the elderly user has read the current health summary; the delayed processing control is used to record the behavior of not responding temporarily, and generate a reminder task based on the event urgency level and preset time interval; the manual call control is used to trigger contact notification or nursing staff intervention process. Based on the health record fields, health knowledge items, rule version information, model version information, display template version information, and strategy version information referenced in this demonstration process, corresponding reference chain information is constructed. The health summary content, interactive control set, reference chain information, display validity period, and display language information are structured and encapsulated to generate the display payload. By accurately matching template structures and filling in information, the health summary text ensures that it covers individual health status, relevant health knowledge and event handling suggestions. The text readability is optimized according to the reading habits of elderly users, which can significantly improve the efficiency of information delivery, reduce the difficulty of understanding for elderly users, and ensure that they can obtain the health information they need in a timely manner. By appropriately setting control types and behavioral logic, such as confirmation, delayed processing, and manual call controls, the system can not only guide users to interact effectively, but also automatically trigger appropriate follow-up operations based on the urgency of the event. This can improve the sense of participation of elderly users, reduce erroneous operations, and ensure the timeliness of event response.

[0010] Furthermore, step S400 determines the push display payload, including: Obtain the set of display payloads to be displayed, and extract the event urgency level, display validity period expiration time, target picture frame device identifier, and object's work and rest preference information for each display payload; calculate the remaining valid duration of each display payload based on the current time, and obtain the online status information of the corresponding picture frame device; Based on the event urgency level, remaining valid time, photo frame online status, and daily routine preference information, priority scoring is performed on each display load. The priority scoring process includes assigning a numerical value to the event urgency level; the remaining valid time scoring process includes pre-setting several remaining valid time levels and assigning values; the photo frame online status scoring process includes assigning a value to the photo frame online status; and the scores of the above items are weighted and fused to generate a comprehensive priority score for each display load. All display loads participating in the scheduling are sorted based on their comprehensive priority scores to generate a display queue. The display queue is arranged from high to low priority scores. Based on the current time and the expected display duration of each display load, a corresponding planned delivery time is assigned to each display load in the queue using an additive calculation method. The display queue is subjected to deduplication and merging processing. The deduplication and merging processing includes determining whether there are multiple display payloads with the same object identifier and the same event type in the display queue. When there are multiple display payloads that meet the conditions, the display payload with the highest comprehensive priority score is retained, and the remaining display payloads are marked as redundant data and removed from the display queue. The display payloads are dispatched sequentially according to the display queue order. When the corresponding planned dispatch time is reached, the dispatch process is triggered. Before dispatch, the online status of the target frame device is confirmed. When the frame device is detected to be offline or unresponsive, the corresponding display payload is processed with a delay and retry. The delay and retry process includes re-adding the payload to the scheduling queue at a preset time interval and increasing the retry count. When the number of retries exceeds a preset limit and the corresponding event urgency level is high, a backup notification mechanism is triggered. The distributed display payload is serialized and pushed to the corresponding electronic photo frame device. During the push process, the actual distribution timestamp and message sequence number information are attached. After receiving the display payload, the photo frame device performs local rendering processing and returns display confirmation information to the platform. By extracting the event urgency level, remaining effective duration, online status of the target frame, and the user's work and rest preferences from the display payloads, a multi-dimensional scoring mechanism is used to comprehensively prioritize each display payload. This enables intelligent sorting and reasonable allocation of display resources. By assigning values ​​to different dimensions and weighting and integrating them, the system can prioritize the display of urgent, time-sensitive, and user-friendly health information when handling a large number of display tasks, thereby improving the timeliness and accuracy of information delivery. By deduplicating and merging the display queue, duplicate display tasks are effectively eliminated, system resource waste is reduced, and users are prevented from receiving duplicate or redundant health information, thereby improving user experience and display efficiency. Combined with the expected display duration of the display load, the delivery time of each health information can be scientifically arranged to achieve a continuous and reasonable display rhythm. The system ensures reliable delivery of display payloads even when the target frame device is offline or unresponsive by using online status confirmation and delayed retry mechanisms. For high-urgent events, a backup notification mechanism is triggered to ensure that critical health information reaches the user in a timely manner, thereby improving the system's robustness and event response capabilities.

[0011] Furthermore, step S500 includes: After receiving the display payload, the electronic photo frame device parses and processes it to extract the health summary content, interactive control set, display template identifier, display template version information, display validity period, display language information, and reference chain information; it also verifies the structural integrity of the display payload, including field integrity verification, field type validity verification, and display validity period validity verification; if the verification fails, it generates a payload anomaly record and returns an anomaly status information to the platform, terminating the current display process; Based on the display template identifier and version information carried in the display payload, the corresponding rendering template is retrieved from the local template cache. If the corresponding version of the rendering template does not exist locally, a template synchronization request is initiated to the platform to obtain the corresponding version of the rendering template and cache it locally. After the template is loaded, an integrity check is performed on the rendering template, which includes template file integrity check, version consistency check, and template structure legality check. If the integrity check fails, the previously available version of the rendering template saved locally is called to perform rollback processing, and a template exception log is recorded. Based on the validated rendering template, the layout parsing and interface parameter calculation processing of the display interface are performed. The layout parsing and interface parameter calculation processing includes determining the position, size and spacing information of each display area according to the screen resolution, display orientation and template layout parameters; the display interface is divided into a summary information area and an interactive control area, wherein the summary information area is located at the top of the screen and is used to display health summary content, and the interactive control area is located at the bottom of the screen and is used to display user interaction controls. The health summary content undergoes text formatting and readability adaptation processing, which includes determining font size, font type, character spacing, line spacing, and paragraph spacing parameters according to preset readability rules. The preset readability rules include using a font size no smaller than the minimum readable font size, using a high-contrast foreground and background color combination, and using a sans-serif font. Key prompt fields, suggested action fields, and time prompt fields in the summary content are displayed differently according to preset emphasis rules. When the health summary content exceeds the display range of a single screen, pagination is performed on the health summary content. The pagination process includes pagination based on sentence boundaries, semantic segment boundaries, or paragraph boundaries to avoid semantic information being truncated at the pagination position; generating page number identifiers and page turning prompts for each page after pagination; and loading touch page turning controls or sliding page turning interaction areas in the interface to support users to perform page turning operations. The system collects the current ambient light intensity information of the electronic photo frame device and performs adaptive screen brightness adjustment processing based on the ambient light intensity information. The adaptive screen brightness adjustment processing includes increasing the screen brightness to the corresponding high brightness level when the ambient light intensity is higher than a preset high brightness threshold; reducing the screen brightness to a preset eye-protection brightness range when the ambient light intensity is lower than a preset low brightness threshold; and linearly or segmentally adjusting the screen brightness according to a preset continuous mapping rule when the ambient light intensity is between the preset high brightness threshold and the preset low brightness threshold. After adjusting the screen brightness, the final display interface is generated based on the rendering template, layout calculation results, and text layout results, and then the final display interface is output to the electronic photo frame display screen for display. After the display is completed, the rendering completion time, the current display page identifier, the display template version information, and the device status information are recorded to generate a display execution record and return display confirmation information to the platform. Simultaneously, an interactive monitoring process is initiated, and user interaction behavior is collected within a preset interactive time window. The interactive monitoring process includes touch interaction acquisition and voice interaction acquisition. The touch interaction acquisition includes detecting the user's touch input on the display interface, obtaining touch start time, touch end time, touch position coordinates, touch trajectory, and touch duration information; determining whether an interactive control is hit based on the mapping relationship between the touch position and the interface control area, and recording the identifier of the hit control; performing abnormal touch filtering on inputs with touch duration below a preset lower limit, abnormal touch trajectory, or touch position exceeding the effective interactive area; and generating a corresponding touch intent label based on preset control intent mapping rules when a valid interactive control is hit. The voice interaction acquisition includes acquiring user voice input signals within a preset voice monitoring period; performing voice activity detection on the voice input signals to determine valid voice segments; performing voice recognition processing on the valid voice segments to generate corresponding recognized text; performing intent recognition processing based on the recognized text to obtain corresponding voice intent labels and intent confidence levels; simultaneously extracting speaker feature information from the voice input and comparing it with authorized user feature templates pre-stored on the device side to generate identity consistency results; when the intent confidence level is lower than a preset threshold, the corresponding voice interaction result is marked as an uncertain intent. The touch interaction results and voice interaction results are summarized according to a unified field structure to generate a unified interaction record. When multiple interaction records exist within the same preset interaction time window, they are associated and integrated in chronological order. When no valid touch interaction or valid voice interaction is detected within the preset interaction time window, the current user status is determined to be unresponsive, and a unresponsive identifier is generated. The unresponsive identifier, along with the display payload identifier, device identifier, and timestamp information, is reported to the platform. The platform then performs confirmation, delay processing, manual call, or re-reminder task generation based on the interaction results. By performing structural integrity checks on the displayed load, including field integrity, field type validity, and display validity period validity checks, abnormal data is intercepted and processed before display, preventing erroneous or invalid information from entering the display process. This improves the overall stability and data security of the system and enables the traceability and rapid location of problems. By introducing a local caching and dynamic synchronization mechanism based on display template identifiers and version information, efficient management of rendering templates is achieved. When a local template is missing or the version is mismatched, the latest template can be automatically synchronized from the platform, and if the verification fails, it can be rolled back to a historically available version, effectively ensuring the continuity and reliability of the display process and avoiding display interruptions caused by template abnormalities. By analyzing the layout and calculating the interface parameters, the interface is adaptively divided according to the device screen resolution and display orientation. The display interface is reasonably divided into a summary information area and an interactive control area, making the display structure clear and hierarchical, which helps to improve the reading experience and operation convenience of elderly users. By employing a semantic boundary-based pagination mechanism, semantic fragmentation at pagination points is avoided. Combined with page number indicators and page-turning prompts, this allows users to browse long texts more naturally and smoothly, enhancing the overall user experience. By collecting ambient light intensity information and performing adaptive screen brightness adjustment, dynamic optimization of display brightness is achieved. It automatically adjusts to a suitable brightness range under different lighting conditions, which not only improves display clarity but also takes into account eye protection needs, enhancing the comfort and user-friendliness of the device. It supports both touch and voice dual-mode interaction, and improves the accuracy and robustness of interaction recognition through abnormal touch filtering, intent recognition and confidence assessment mechanisms; it achieves identity consistency verification through speaker feature comparison, effectively improving interaction security; and it enables the system to fully and accurately understand user intent through unified interaction record structure and multi-interaction fusion processing within a time window.

[0012] Furthermore, step S600 includes: Feature extraction is performed on each interaction signal to generate a signal quality score. The touch signal quality score includes the touch input signals of the electronic photo frame device collected within a preset interaction time window, including touch start time, end time, touch position and trajectory, and duration information. Based on the mapping relationship between the touch position and the interface control area, it is determined whether a valid control is hit, and abnormal touch inputs with touch duration below a preset threshold, abnormal trajectory, or position outside the valid area are excluded. For touch inputs that hit valid controls, a touch validity sub-score is calculated based on the control hit status and dwell time. The speech signal quality score includes: acquiring user speech input signals within a preset speech monitoring period; detecting speech activity in the speech signals and identifying valid speech segments; performing speech recognition on valid speech segments to generate recognized text and performing intent recognition to obtain speech intent labels and intent confidence; extracting speaker feature information and comparing it with pre-stored authorized user feature templates on the device to generate voiceprint consistency markers; and generating a speech interaction feature score based on a weighted sum of speech recognition confidence, intent clarity, and voiceprint matching. The respiratory signal quality score includes collecting real-time signals from respiratory-related sensors, extracting respiratory feature parameters, comparing real-time respiratory features with individual baseline respiratory features, and calculating respiratory sensor feature scores. The signal quality score of the interactive signal is normalized, and each sub-score after normalization is used as a feature weighting coefficient. The signal quality is then weighted and summed according to the corresponding feature weighting coefficients to obtain the comprehensive confirmation confidence level. By collecting key parameters of touch signals and judging their validity based on the preset effective area and control mapping relationship, abnormal touch inputs can be effectively filtered out, improving the accuracy and reliability of touch interaction. For valid touch inputs, a touch validity score is calculated based on the control hit situation and dwell time, which provides more accurate data support for subsequent interaction intent recognition. By generating a speech signal quality score, the speech input signal is comprehensively detected and analyzed. Through multi-level feature extraction such as speech activity detection, speech recognition, intent recognition and voiceprint matching, and weighted summation of factors such as speech recognition confidence, intent clarity and voiceprint consistency, the quality of speech interaction is effectively evaluated, ensuring that the system can accurately understand the user's intent, especially the accurate recognition of speech signals in complex environments.

[0013] Furthermore, step S700 includes: The signal quality scores corresponding to each interaction signal are summarized, the global signal quality mean is calculated, and the global signal quality mean is compared with a preset quality threshold. When the global signal quality mean is lower than the preset quality threshold, the current overall interaction signal is determined to be unusable, the final correction confidence is forcibly set to zero, and the upgrade review process is triggered. When the global signal quality meets the preset conditions, the consistency detection of touch interaction intent and voice interaction intent is performed, and conflict determination is performed in combination with the voiceprint consistency result; when a semantic conflict or voiceprint inconsistency is detected, an interaction conflict is determined, and a guided correction process is triggered to guide the user to confirm the true intent through another interaction; when any modal interaction is missing or the voice recognition result is in an uncertain state, the corresponding semantic conflict determination does not participate in the conflict determination. Based on the signal quality assessment results and conflict assessment results, the comprehensive confirmation confidence level is corrected to generate a final corrected confidence level. The final corrected confidence level is then compared with a preset confirmation threshold and a review threshold. Based on the comparison results, a tiered triggering process is executed: when the final corrected confidence level is greater than or equal to the confirmation threshold, a standard closed-loop process is triggered; when the final corrected confidence level is between the review threshold and the confirmation threshold, a secondary confirmation process is triggered; when the final corrected confidence level is lower than the review threshold or no effective response is received within a preset time, an escalated review process is triggered. When the global signal quality is lower than the preset quality threshold, the current overall interaction signal is directly determined to be unusable, and the final correction confidence is forcibly set to zero. At the same time, the upgrade review process is triggered to avoid low-quality or abnormal data from participating in decision-making from the source, effectively reducing the risk of misjudgment and significantly improving the robustness and security of system operation. By guiding users to perform secondary confirmation, the system effectively avoids accidental triggering and erroneous execution, enhancing the accuracy and security of human-computer interaction. It also provides fault tolerance for situations with missing modalities or uncertain speech recognition, avoiding misjudgments caused by incomplete information and improving the robustness of the system.

[0014] To better implement the above methods, a home-based elderly care health summary credibility analysis system based on electronic photo frames is proposed. The system includes an event object module, a display demand object module, a display payload module, a push display payload module, an identification response module, a comprehensive confirmation confidence module, and a correction process module. Event Object Module: Receives event data from various input sources, preprocesses the event data, and generates event objects; Display Requirement Object Module: Based on the event object, retrieve the individual health context from the health record database, combine it with the knowledge base to retrieve and match event handling suggestions, and generate the display requirement object; Display payload module: Based on the display requirements, it generates health summary text for elderly users, combines the preset control configuration rules in the display template to generate a set of interactive controls, and generates a display payload based on the health summary text and the set of interactive controls; Push Display Load Module: Obtain the set of display loads to be displayed, calculate the comprehensive priority score of the display loads, generate a display queue based on the comprehensive priority score, and determine the display loads to be pushed in combination with the target picture frame device status; Response identification module: Based on the push display payload, load the corresponding rendering template, generate the final display interface and complete the display. Collect user touch and voice interaction behaviors within a preset interaction time window, identify the interaction intent and generate a unified interaction record; when no valid interaction is detected, generate a non-response mark and report it to the platform. Comprehensive confirmation confidence module: Extracts features from the collected touch signals, voice signals and physiological signals, calculates the quality scores of each type of signal, and generates a comprehensive confirmation confidence score; Correction Process Module: Summarizes the quality scores of each signal, calculates the global average signal quality, compares it with a preset threshold, and determines the correction process.

[0015] Furthermore, the presentation payload module includes a health summary content unit and a unit for generating presentation payloads; Health Summary Content Unit: Based on the event type, individual health summary, and matching health knowledge items in the display needs object, a health summary text for elderly users is generated. The process of generating the health summary text includes: determining the corresponding display template identifier according to the event type; reading the template structure corresponding to the display template identifier, the template structure including an event prompt information field, a suggested action information field, and an individual context field; filling the event information into the event prompt information field according to preset field mapping rules, filling the suggested content from the health knowledge items into the suggested action information field, and filling the individual health summary and lifestyle preference information into the individual context field; performing statement simplification and readability adjustment on the filled text content to adapt to the reading habits of elderly users; performing sensitive information filtering and verification on the generated health summary text; and outputting health summary content that conforms to preset security rules. The display payload generation unit generates a set of interactive controls associated with the health summary content based on the preset control configuration rules in the display template. This generation process includes parsing the control type configuration and trigger condition configuration in the display template; determining the types of controls to be loaded based on the urgency and type of the event; and binding corresponding behavior processing logic to the confirmation, delayed processing, and manual call controls. The confirmation control records that the elderly user has read the current health summary; the delayed processing control records the behavior of not responding temporarily and generates a reminder task based on the event urgency level and a preset time interval; and the manual call control triggers contact notification or nursing staff intervention. Based on the health record fields, health knowledge items, rule version information, model version information, display template version information, and strategy version information referenced in this demonstration process, corresponding reference chain information is constructed. The health summary content, interactive control set, reference chain information, display validity period, and display language information are then structured and encapsulated to generate the display payload.

[0016] Compared with existing technologies, the beneficial effects of this invention are: by constructing a multi-source data fusion, individual health context awareness, knowledge-driven generation, multimodal interaction signal fusion, and dynamic credibility assessment mechanism, it realizes a closed-loop control of the entire process from data collection, information generation, display and interaction to result confirmation. This can significantly improve the accuracy, personalization, interaction reliability, and system security of health information services in home-based elderly care scenarios, and overcome the problems of isolated data, single interaction, and insufficient result credibility in existing technologies. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the structure of the home-based elderly care health summary credibility analysis system based on electronic photo frames according to the present invention. Detailed Implementation

[0018] 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.

[0019] Please see Figure 1 This invention provides a technical solution: a method for credibility analysis of home-based elderly care health summaries based on electronic photo frames, the method comprising: Step S100: Receive event data from various input sources, preprocess the event data, and generate event objects; The event object generated in step S100 includes: The input sources include home device alarms, customer service work orders, follow-up plan triggers, and family member requests. The event data from each input source is preprocessed. The preprocessing of home device alarms includes obtaining device identifier, alarm type, trigger time, and alarm level information. The alarm type and alarm level are then validated for format validity. This validation includes determining whether the alarm type belongs to a preset set of event types and whether the alarm level belongs to a preset level range. The integrity of the field format and the validity of the values ​​are also validated. Data that passes the validation is normalized for the event type, while data that fails the validation is marked as abnormal and discarded. The preprocessing of the customer service work order includes obtaining the work order identifier, request description information, submission time, and source channel information; performing semantic parsing on the request description information; the semantic parsing includes word segmentation, keyword extraction, and intent recognition based on preset semantic rules; determining the corresponding event type and urgency level based on the extracted keywords and intent recognition, combined with preset event type mapping rules and level determination rules; when there are unparsable or parsing results that do not meet preset rules, the corresponding content is filled with a preset unknown identifier, and an input credibility identifier is generated based on the matching degree of the semantic parsing results. The preprocessing for triggering the follow-up plan includes obtaining the plan identifier, object identifier, plan trigger time, and follow-up type; generating the corresponding event type based on the mapping relationship between the follow-up type and the preset event type; and filling in the corresponding emergency level based on the preset level allocation rule to form event data. The preprocessing of the family's request includes recognizing and processing the voice or text information submitted by the family. The recognition and processing includes voice-to-text processing or text normalization processing to obtain the recognized text and intent tags. Based on the recognized text and intent tags, and in combination with preset event type matching rules, the corresponding event type and urgency level are determined. Based on the matching degree of the recognition process, a recognition confidence score is generated. When the recognition confidence score is lower than a preset threshold, the input confidence score is marked as low. The preprocessed input source data is uniformly standardized. The standardization process includes normalizing the event types and unifying semantic equivalence into standard event types; filling missing fields with default values ​​according to the input source type, and filling undetermined fields with preset unknown identifiers; and deduplicating and merging data with the same source and event type within a preset time window to generate event objects. For example, the system receives event data from multiple input sources within a preset time window from 08:30 to 09:00 on a certain day. For instance, a home device generates an alarm data, with the device identifier "DEV-HT-001", the alarm type being "abnormal heart rate", the trigger time being "2026-03-30 08:32:15", and the alarm level being "high". The system performs a format validity check on this data, confirming that "abnormal heart rate" belongs to the preset event type set, the "high" level is within the preset level range, and the fields are complete and the values ​​are valid. Therefore, the system passes the validation and normalizes the alarm type into the standard event type "abnormal vital signs", forming standardized event data.

[0020] Step S200: Based on the event object, retrieve the individual health context from the health record database, combine it with the knowledge base to retrieve and match event handling suggestions, and generate a display requirement object; The step S200 of generating the display requirement object includes: Using the elderly identifier in the event object as an index, retrieve the individual health information of the elderly identifier from the health record database. The individual health information includes historical health data, medication records, daily routine preferences, and contact information. When no individual health information is retrieved, the relevant fields are initialized to empty and the context status is set to incomplete information. Using event type and urgency as indexes, health knowledge entries matching the event type are retrieved from a preset health knowledge base. The process of retrieving health knowledge entries includes obtaining health knowledge using an exact matching method based on event type. When no exact match is found, a fuzzy matching method based on symptom tag similarity is used. The fuzzy matching includes calculating the similarity between the symptom tags corresponding to the event and the tag set of each health knowledge entry in the knowledge base. The similarity calculation is based on tag overlap, semantic similarity, or distance between vector representations. Based on the calculated similarity results, health knowledge entries with similarity greater than a preset threshold are selected as candidate health entries. Candidate health items are sorted from high to low similarity, and combined with the event type consistency rule, candidate health items that do not match the event type are eliminated to obtain the final matched health knowledge items. The retrieved individual health information is integrated with the corresponding health knowledge entries to generate contextual information containing an individual health summary, work and rest preferences, and health knowledge entry identifiers. Sensitive fields in the generated contextual information are anonymized according to preset permission levels. Based on the contextual information and event objects, display requirement objects are generated. For example, assuming the elderly person identified in the event object is "USER-001", the event type is "abnormal vital signs", and the urgency level is "high", the system first uses this elderly person's identifier as an index to retrieve the corresponding individual health information from the health record database. The search results show that the elderly person's historical health data includes "hypertension (diagnosis time: 2022-05-12), coronary heart disease (diagnosis time: 2023-08-03)", the average heart rate in the past week is "92 beats / minute", and the blood pressure fluctuation range is "140-160 / 90-100". The patient's blood pressure was measured in mmHg; medication records included "nifedipine sustained-release tablets (30mg once daily), aspirin enteric-coated tablets (100mg once daily)"; sleep preferences were "wake up at 06:30, go to sleep at 21:30, and take a 30-minute nap at 12:30"; contact information included "children's contact person: Zhang (phone: 13800000000), community doctor: Dr. Li (phone: 13900000001)". Since all the above fields were successfully retrieved, the context status was marked as "information complete". The system uses the event type "abnormal vital signs" and the urgency level "high" as indexes to retrieve matching health knowledge entries from a pre-set health knowledge base. First, it uses exact matching, matching the knowledge entry "K-HL-001," whose content is "Abnormal heart rate accompanied by dizziness may indicate arrhythmia or insufficient myocardial blood supply; it is recommended to rest immediately and monitor heart rate changes." If no exact match is found in another scenario, the system will calculate similarity based on symptom tags (such as "rapid heart rate," "dizziness," "palpitations") and the tag set of the entries in the knowledge base. For example, through... The combined similarity calculated from tag overlap (0.67), semantic similarity (0.81), and vector distance was 0.79, which is higher than the preset threshold of 0.75. This allowed candidate knowledge entries such as "K-HL-003" and "K-HL-007" to be selected and sorted from highest to lowest similarity. Further, by combining the event type consistency rule, entries that did not match "abnormal vital signs" (such as entries related to "psychological and emotional intervention") were removed. Finally, knowledge entries "K-HL-001" and "K-HL-003" were determined as the matching results. The system integrates the retrieved individual health information with the aforementioned health knowledge entries to generate contextual information. For example, the individual health summary is "This elderly person has a history of hypertension and coronary heart disease, and recently has a rapid heart rate and large fluctuations in blood pressure, indicating a cardiovascular risk." It also retains the lifestyle preferences "early to bed and early to rise, with a habit of taking a midday nap" and associates them with the health knowledge entry identifiers "K-HL-001" and "K-HL-003". During the generation process, the system desensitizes sensitive fields according to the preset permission level (e.g., currently customer service seat permission). For example, the contact phone number "13800000000" is desensitized to "1380000", and "13900000001" is desensitized to "1390001".

[0021] Step S300: Based on the display requirements, generate health summary text for elderly users, combine the preset control configuration rules in the display template to generate an interactive control set, and generate a display payload based on the health summary text and the interactive control set; The generation of the demonstration payload in step S300 includes: Based on the event type, individual health summary, and matching health knowledge items in the display requirements, a health summary text for elderly users is generated. The process includes: determining the corresponding display template identifier based on the event type; reading the template structure corresponding to the display template identifier, whereby the template structure includes an event prompt information field, a suggested action information field, and an individual context field; filling the event information into the event prompt information field, filling the suggested content from the health knowledge items into the suggested action information field, and filling the individual health summary and lifestyle preference information into the individual context field according to preset field mapping rules; performing statement simplification and readability adjustments on the filled text content to suit the reading habits of elderly users; filtering and verifying sensitive information in the generated health summary text; and outputting health summary content that conforms to preset security rules. According to the preset control configuration rules in the display template, a set of interactive controls associated with the health summary content is generated. The process of generating the set of interactive controls includes parsing the control type configuration and trigger condition configuration in the display template; determining the control types to be loaded based on the urgency and type of the event; binding corresponding behavior processing logic to the confirmation, delayed processing, and manual call controls respectively. The confirmation control is used to record that the elderly user has read the current health summary; the delayed processing control is used to record the behavior of not responding temporarily, and generate a reminder task based on the event urgency level and preset time interval; the manual call control is used to trigger contact notification or nursing staff intervention process. Based on the health record fields, health knowledge items, rule version information, model version information, display template version information, and strategy version information referenced in this demonstration process, corresponding reference chain information is constructed. The health summary content, interactive control set, reference chain information, display validity period, and display language information are then structured and encapsulated to generate the display payload.

[0022] Step S400: Obtain the set of display payloads to be displayed, calculate the comprehensive priority score of the display payloads, generate a display queue based on the comprehensive priority score, and determine the display payloads to be pushed in combination with the target photo frame device status; The presentation payload to be pushed in step S400 includes: Obtain the set of display payloads to be displayed, and extract the event urgency level, display validity period expiration time, target picture frame device identifier, and object's work and rest preference information for each display payload; calculate the remaining valid duration of each display payload based on the current time, and obtain the online status information of the corresponding picture frame device; Based on the event urgency level, remaining valid time, photo frame online status, and daily routine preference information, priority scoring is performed on each display load. The priority scoring process includes assigning a numerical value to the event urgency level; the remaining valid time scoring process includes pre-setting several remaining valid time levels and assigning values; the photo frame online status scoring process includes assigning a value to the photo frame online status; and the scores of the above items are weighted and fused to generate a comprehensive priority score for each display load. All display loads participating in the scheduling are sorted based on their comprehensive priority scores to generate a display queue. The display queue is arranged from high to low priority scores. Based on the current time and the expected display duration of each display load, a corresponding planned delivery time is assigned to each display load in the queue using an additive calculation method. The display queue is subjected to deduplication and merging processing. The deduplication and merging processing includes determining whether there are multiple display payloads with the same object identifier and the same event type in the display queue. When there are multiple display payloads that meet the conditions, the display payload with the highest comprehensive priority score is retained, and the remaining display payloads are marked as redundant data and removed from the display queue. The display payloads are dispatched sequentially according to the display queue order. When the corresponding planned dispatch time is reached, the dispatch process is triggered. Before dispatch, the online status of the target frame device is confirmed. When the frame device is detected to be offline or unresponsive, the corresponding display payload is processed with a delay and retry. The delay and retry process includes re-adding the payload to the scheduling queue at a preset time interval and increasing the retry count. When the number of retries exceeds a preset limit and the corresponding event urgency level is high, a backup notification mechanism is triggered. The distributed display payload is serialized and pushed to the corresponding electronic photo frame device. The actual distribution timestamp and message sequence number are attached during the push process. After receiving the display payload, the photo frame device performs local rendering and returns display confirmation information to the platform.

[0023] Step S500: Load the corresponding rendering template based on the push display payload, generate the final display interface and complete the display. Collect user touch and voice interaction behaviors within the preset interaction time window, identify the interaction intent and generate a unified interaction record; when no valid interaction is detected, generate a non-response mark and report it to the platform. Step S500 includes: After receiving the display payload, the electronic photo frame device parses and processes it to extract the health summary content, interactive control set, display template identifier, display template version information, display validity period, display language information, and reference chain information; it also verifies the structural integrity of the display payload, including field integrity verification, field type validity verification, and display validity period validity verification; if the verification fails, it generates a payload anomaly record and returns an anomaly status information to the platform, terminating the current display process; Based on the display template identifier and version information carried in the display payload, the corresponding rendering template is retrieved from the local template cache. If the corresponding version of the rendering template does not exist locally, a template synchronization request is initiated to the platform to obtain the corresponding version of the rendering template and cache it locally. After the template is loaded, an integrity check is performed on the rendering template, which includes template file integrity check, version consistency check, and template structure legality check. If the integrity check fails, the previously available version of the rendering template saved locally is called to perform rollback processing, and a template exception log is recorded. Based on the validated rendering template, the layout parsing and interface parameter calculation processing of the display interface are performed. The layout parsing and interface parameter calculation processing includes determining the position, size and spacing information of each display area according to the screen resolution, display orientation and template layout parameters; the display interface is divided into a summary information area and an interactive control area, wherein the summary information area is located at the top of the screen and is used to display health summary content, and the interactive control area is located at the bottom of the screen and is used to display user interaction controls. The health summary content undergoes text formatting and readability adaptation processing, which includes determining font size, font type, character spacing, line spacing, and paragraph spacing parameters according to preset readability rules. The preset readability rules include using a font size no smaller than the minimum readable font size, using a high-contrast foreground and background color combination, and using a sans-serif font. Key prompt fields, suggested action fields, and time prompt fields in the summary content are displayed differently according to preset emphasis rules. When the health summary content exceeds the display range of a single screen, pagination is performed on the health summary content. The pagination process includes pagination based on sentence boundaries, semantic segment boundaries, or paragraph boundaries to avoid semantic information being truncated at the pagination position; generating page number identifiers and page turning prompts for each page after pagination; and loading touch page turning controls or sliding page turning interaction areas in the interface to support users to perform page turning operations. The system collects the current ambient light intensity information of the electronic photo frame device and performs adaptive screen brightness adjustment processing based on the ambient light intensity information. The adaptive screen brightness adjustment processing includes increasing the screen brightness to the corresponding high brightness level when the ambient light intensity is higher than a preset high brightness threshold; reducing the screen brightness to a preset eye-protection brightness range when the ambient light intensity is lower than a preset low brightness threshold; and linearly or segmentally adjusting the screen brightness according to a preset continuous mapping rule when the ambient light intensity is between the preset high brightness threshold and the preset low brightness threshold. After adjusting the screen brightness, the final display interface is generated based on the rendering template, layout calculation results, and text layout results, and then the final display interface is output to the electronic photo frame display screen for display. After the display is completed, the rendering completion time, the current display page identifier, the display template version information, and the device status information are recorded to generate a display execution record and return display confirmation information to the platform. Simultaneously, an interactive monitoring process is initiated, and user interaction behavior is collected within a preset interactive time window. The interactive monitoring process includes touch interaction acquisition and voice interaction acquisition. The touch interaction acquisition includes detecting the user's touch input on the display interface, obtaining touch start time, touch end time, touch position coordinates, touch trajectory, and touch duration information; determining whether an interactive control is hit based on the mapping relationship between the touch position and the interface control area, and recording the identifier of the hit control; performing abnormal touch filtering on inputs with touch duration below a preset lower limit, abnormal touch trajectory, or touch position exceeding the effective interactive area; and generating a corresponding touch intent label based on preset control intent mapping rules when a valid interactive control is hit. The voice interaction acquisition includes acquiring user voice input signals within a preset voice monitoring period; performing voice activity detection on the voice input signals to determine valid voice segments; performing voice recognition processing on the valid voice segments to generate corresponding recognized text; performing intent recognition processing based on the recognized text to obtain corresponding voice intent labels and intent confidence levels; simultaneously extracting speaker feature information from the voice input and comparing it with authorized user feature templates pre-stored on the device side to generate identity consistency results; when the intent confidence level is lower than a preset threshold, the corresponding voice interaction result is marked as an uncertain intent. The touch interaction results and voice interaction results are summarized according to a unified field structure to generate a unified interaction record. When multiple interaction records exist within the same preset interaction time window, they are associated and integrated in chronological order. When no valid touch interaction or valid voice interaction is detected within the preset interaction time window, the current user status is determined to be non-responsive, and a non-responsive identifier is generated. The non-responsive identifier, along with the display payload identifier, device identifier, and timestamp information, is reported to the platform. The platform then performs confirmation, delay processing, manual call, or re-reminder task generation based on the interaction results.

[0024] Step S600: Extract features from the collected touch signals, voice signals and physiological signals, calculate the quality scores of each type of signal, and generate a comprehensive confirmation confidence score; Step S600 includes: Feature extraction is performed on each interaction signal to generate a signal quality score. The touch signal quality score includes the touch input signals of the electronic photo frame device collected within a preset interaction time window, including touch start time, end time, touch position and trajectory, and duration information. Based on the mapping relationship between the touch position and the interface control area, it is determined whether a valid control is hit, and abnormal touch inputs with touch duration below a preset threshold, abnormal trajectory, or position outside the valid area are excluded. For touch inputs that hit valid controls, a touch validity sub-score is calculated based on the control hit status and dwell time. The speech signal quality score includes: acquiring user speech input signals within a preset speech monitoring period; detecting speech activity in the speech signals and identifying valid speech segments; performing speech recognition on valid speech segments to generate recognized text and performing intent recognition to obtain speech intent labels and intent confidence; extracting speaker feature information and comparing it with pre-stored authorized user feature templates on the device to generate voiceprint consistency markers; and generating a speech interaction feature score based on a weighted sum of speech recognition confidence, intent clarity, and voiceprint matching. The respiratory signal quality score includes collecting real-time signals from respiratory-related sensors, extracting respiratory feature parameters, comparing real-time respiratory features with individual baseline respiratory features, and calculating respiratory sensor feature scores. The signal quality scores of the interactive signals are normalized, and each sub-score after normalization is used as a feature weighting coefficient. The signal quality scores are then weighted and summed according to the corresponding feature weighting coefficients to obtain the overall confirmation confidence level.

[0025] Step S700: Summarize the quality scores of each signal, calculate the global signal quality mean, compare it with the preset threshold, and determine the correction process; Step S700 includes: The signal quality scores corresponding to each interaction signal are summarized, the global signal quality mean is calculated, and the global signal quality mean is compared with a preset quality threshold. When the global signal quality mean is lower than the preset quality threshold, the current overall interaction signal is determined to be unusable, the final correction confidence is forcibly set to zero, and the upgrade review process is triggered. When the global signal quality meets the preset conditions, the consistency detection of touch interaction intent and voice interaction intent is performed, and conflict determination is performed in combination with the voiceprint consistency result; when a semantic conflict or voiceprint inconsistency is detected, an interaction conflict is determined, and a guided correction process is triggered to guide the user to confirm the true intent through another interaction; when any modal interaction is missing or the voice recognition result is in an uncertain state, the corresponding semantic conflict determination does not participate in the conflict determination. Based on the signal quality judgment results and conflict judgment results, the comprehensive confirmation confidence level is corrected to generate a final corrected confidence level. The final corrected confidence level is compared with a preset confirmation threshold and a review threshold. According to the comparison results, a tiered triggering process is executed: when the final corrected confidence level is greater than or equal to the confirmation threshold, a standard closed-loop process is triggered; when the final corrected confidence level is between the review threshold and the confirmation threshold, a secondary confirmation process is triggered; when the final corrected confidence level is lower than the review threshold or no effective response is obtained within a preset time, an escalation review process is triggered.

[0026] To better implement the above methods, a home-based elderly care health summary credibility analysis system based on electronic photo frames is proposed. The system includes an event object module, a display demand object module, a display payload module, a push display payload module, an identification response module, a comprehensive confirmation confidence module, and a correction process module. Event Object Module: Receives event data from various input sources, preprocesses the event data, and generates event objects; Display Requirement Object Module: Based on the event object, retrieve the individual health context from the health record database, combine it with the knowledge base to retrieve and match event handling suggestions, and generate the display requirement object; Display payload module: Based on the display requirements, it generates health summary text for elderly users, combines the preset control configuration rules in the display template to generate a set of interactive controls, and generates a display payload based on the health summary text and the set of interactive controls; The presentation payload module includes a health summary content unit and a presentation payload generation unit; Health Summary Content Unit: Based on the event type, individual health summary, and matching health knowledge items in the display needs object, a health summary text for elderly users is generated. The process of generating the health summary text includes: determining the corresponding display template identifier according to the event type; reading the template structure corresponding to the display template identifier, the template structure including an event prompt information field, a suggested action information field, and an individual context field; filling the event information into the event prompt information field according to preset field mapping rules, filling the suggested content from the health knowledge items into the suggested action information field, and filling the individual health summary and lifestyle preference information into the individual context field; performing statement simplification and readability adjustment on the filled text content to adapt to the reading habits of elderly users; performing sensitive information filtering and verification on the generated health summary text; and outputting health summary content that conforms to preset security rules. The display payload generation unit generates a set of interactive controls associated with the health summary content based on the preset control configuration rules in the display template. This generation process includes parsing the control type configuration and trigger condition configuration in the display template; determining the types of controls to be loaded based on the urgency and type of the event; and binding corresponding behavior processing logic to the confirmation, delayed processing, and manual call controls. The confirmation control records that the elderly user has read the current health summary; the delayed processing control records the behavior of not responding temporarily and generates a reminder task based on the event urgency level and a preset time interval; and the manual call control triggers contact notification or nursing staff intervention. Based on the health record fields, health knowledge items, rule version information, model version information, display template version information, and strategy version information referenced in this demonstration process, corresponding reference chain information is constructed. The health summary content, interactive control set, reference chain information, display validity period, and display language information are then structured and encapsulated to generate the display payload.

[0027] Push Display Load Module: Obtain the set of display loads to be displayed, calculate the comprehensive priority score of the display loads, generate a display queue based on the comprehensive priority score, and determine the display loads to be pushed in combination with the target picture frame device status; Response identification module: Based on the push display payload, load the corresponding rendering template, generate the final display interface and complete the display. Collect user touch and voice interaction behaviors within a preset interaction time window, identify the interaction intent and generate a unified interaction record; when no valid interaction is detected, generate a non-response mark and report it to the platform. Comprehensive confirmation confidence module: Extracts features from the collected touch signals, voice signals and physiological signals, calculates the quality scores of each type of signal, and generates a comprehensive confirmation confidence score; Correction Process Module: Summarizes the quality scores of each signal, calculates the global average signal quality, compares it with a preset threshold, and determines the correction process.

[0028] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A method for credibility analysis of home-based elderly care health summaries based on electronic photo frames, characterized in that, The methods include: Step S100: Receive event data from various input sources, preprocess the event data, and generate event objects; Step S200: Based on the event object, retrieve the individual health context from the health record database, combine it with the knowledge base to retrieve and match event handling suggestions, and generate a display requirement object; Step S300: Based on the display requirements, generate health summary text for elderly users, combine the preset control configuration rules in the display template to generate an interactive control set, and generate a display payload based on the health summary text and the interactive control set; Step S400: Obtain the set of display payloads to be displayed, calculate the comprehensive priority score of the display payloads, generate a display queue based on the comprehensive priority score, and determine the display payloads to be pushed in combination with the target photo frame device status; Step S500: Load the corresponding rendering template based on the push display payload, generate the final display interface and complete the display. Collect user touch and voice interaction behaviors within the preset interaction time window, identify the interaction intent and generate a unified interaction record; when no valid interaction is detected, generate a non-response mark and report it to the platform. Step S600: Extract features from the collected touch signals, voice signals and physiological signals, calculate the quality scores of each type of signal, and generate a comprehensive confirmation confidence score; Step S700: Summarize the quality scores of each signal, calculate the global signal quality mean, compare it with the preset threshold, and determine the correction process.

2. The method for credibility analysis of home-based elderly care health summaries based on electronic photo frames according to claim 1, characterized in that, The step S100 of generating an event object includes the following steps: The input sources include home device alarms, customer service work orders, follow-up plan triggers, and family member requests. The event data from each input source is preprocessed. The preprocessing of home device alarms includes obtaining device identifier, alarm type, trigger time, and alarm level information. The alarm type and alarm level are then validated for format validity. This validation includes determining whether the alarm type belongs to a preset set of event types and whether the alarm level belongs to a preset level range. The integrity of the field format and the validity of the values ​​are also validated. Data that passes the validation is normalized for the event type, while data that fails the validation is marked as abnormal and discarded. The preprocessing of the customer service work order includes obtaining the work order identifier, request description information, submission time, and source channel information; performing semantic parsing on the request description information; the semantic parsing includes word segmentation, keyword extraction, and intent recognition based on preset semantic rules; determining the corresponding event type and urgency level based on the extracted keywords and intent recognition, combined with preset event type mapping rules and level determination rules; when there are unparsable or parsing results that do not meet preset rules, the corresponding content is filled with a preset unknown identifier, and an input credibility identifier is generated based on the matching degree of the semantic parsing results. The preprocessing for triggering the follow-up plan includes obtaining the plan identifier, object identifier, plan trigger time, and follow-up type; generating the corresponding event type based on the mapping relationship between the follow-up type and the preset event type; and filling in the corresponding emergency level based on the preset level allocation rule to form event data. The preprocessing of the family's request includes recognizing and processing the voice or text information submitted by the family. The recognition and processing includes voice-to-text processing or text normalization processing to obtain the recognized text and intent tags. Based on the recognized text and intent tags, and in combination with preset event type matching rules, the corresponding event type and urgency level are determined. Based on the matching degree of the recognition process, a recognition confidence score is generated. When the recognition confidence score is lower than a preset threshold, the input confidence score is marked as low. The preprocessed input source data is subjected to unified standardization processing. The standardization processing includes normalizing the event types and unifying semantic equivalence into standard event types; filling missing fields with default values ​​according to the input source type, and filling undetermined fields with preset unknown identifiers; and deduplicating and merging data with the same source and event type within a preset time window to generate event objects.

3. The method for credibility analysis of home-based elderly care health summaries based on electronic photo frames according to claim 1, characterized in that, The step S200 of generating the display requirement object includes the following steps: Using the elderly identifier in the event object as an index, retrieve the individual health information of the elderly identifier from the health record database. The individual health information includes historical health data, medication records, daily routine preferences, and contact information. When no individual health information is retrieved, the relevant fields are initialized to empty and the context status is set to incomplete information. Using event type and urgency as indexes, health knowledge entries matching the event type are retrieved from a preset health knowledge base. The process of retrieving health knowledge entries includes obtaining health knowledge using an exact matching method based on event type. When no exact match is found, a fuzzy matching method based on symptom tag similarity is used. The fuzzy matching includes calculating the similarity between the symptom tags corresponding to the event and the tag set of each health knowledge entry in the knowledge base. The similarity calculation is based on tag overlap, semantic similarity, or distance between vector representations. Based on the calculated similarity results, health knowledge entries with similarity greater than a preset threshold are selected as candidate health entries. Candidate health items are sorted from high to low similarity, and combined with the event type consistency rule, candidate health items that do not match the event type are eliminated to obtain the final matched health knowledge items. The retrieved individual health information is integrated with the corresponding health knowledge entries to generate contextual information that includes an individual health summary, lifestyle preferences, and health knowledge entry identifiers. Sensitive fields in the generated contextual information are anonymized according to preset permission levels. Based on the contextual information and event objects, display requirements objects are generated.

4. The method for credibility analysis of home-based elderly care health summaries based on electronic photo frames according to claim 1, characterized in that, The generation of the demonstration payload in step S300 includes the following steps: Based on the event type, individual health summary, and matching health knowledge items in the display requirements, a health summary text for elderly users is generated. The process includes: determining the corresponding display template identifier based on the event type; reading the template structure corresponding to the display template identifier, whereby the template structure includes an event prompt information field, a suggested action information field, and an individual context field; filling the event information into the event prompt information field, filling the suggested content from the health knowledge items into the suggested action information field, and filling the individual health summary and lifestyle preference information into the individual context field according to preset field mapping rules; performing statement simplification and readability adjustments on the filled text content to suit the reading habits of elderly users; filtering and verifying sensitive information in the generated health summary text; and outputting health summary content that conforms to preset security rules. According to the preset control configuration rules in the display template, a set of interactive controls associated with the health summary content is generated. The process of generating the set of interactive controls includes parsing the control type configuration and trigger condition configuration in the display template; determining the control types to be loaded based on the urgency and type of the event; binding corresponding behavior processing logic to the confirmation, delayed processing, and manual call controls respectively. The confirmation control is used to record that the elderly user has read the current health summary; the delayed processing control is used to record the behavior of not responding temporarily, and generate a reminder task based on the event urgency level and preset time interval; the manual call control is used to trigger contact notification or nursing staff intervention process. Based on the health record fields, health knowledge items, rule version information, model version information, display template version information, and strategy version information referenced in this demonstration process, corresponding reference chain information is constructed. The health summary content, interactive control set, reference chain information, display validity period, and display language information are then structured and encapsulated to generate the display payload.

5. The method for credibility analysis of home-based elderly care health summaries based on electronic photo frames according to claim 1, characterized in that, Determining the pushable display payload in step S400 includes the following steps: Obtain the set of display payloads to be displayed, and extract the event urgency level, display validity period expiration time, target picture frame device identifier, and object's work and rest preference information for each display payload; calculate the remaining valid duration of each display payload based on the current time, and obtain the online status information of the corresponding picture frame device; Based on the event urgency level, remaining valid time, photo frame online status, and daily routine preference information, priority scoring is performed on each display load. The priority scoring process includes assigning a numerical value to the event urgency level; the remaining valid time scoring process includes pre-setting several remaining valid time levels and assigning values; the photo frame online status scoring process includes assigning a value to the photo frame online status; and the scores of the above items are weighted and fused to generate a comprehensive priority score for each display load. All display loads participating in the scheduling are sorted based on their comprehensive priority scores to generate a display queue. The display queue is arranged from high to low priority scores. Based on the current time and the expected display duration of each display load, a corresponding planned delivery time is assigned to each display load in the queue using an additive calculation method. The display queue is subjected to deduplication and merging processing. The deduplication and merging processing includes determining whether there are multiple display payloads with the same object identifier and the same event type in the display queue. When there are multiple display payloads that meet the conditions, the display payload with the highest comprehensive priority score is retained, and the remaining display payloads are marked as redundant data and removed from the display queue. The display payloads are dispatched sequentially according to the display queue order. When the corresponding planned dispatch time is reached, the dispatch process is triggered. Before dispatch, the online status of the target frame device is confirmed. When the frame device is detected to be offline or unresponsive, the corresponding display payload is processed with a delay and retry. The delay and retry process includes re-adding the payload to the scheduling queue at a preset time interval and increasing the retry count. When the number of retries exceeds a preset limit and the corresponding event urgency level is high, a backup notification mechanism is triggered. The distributed display payload is serialized and pushed to the corresponding electronic photo frame device. The actual distribution timestamp and message sequence number are attached during the push process. After receiving the display payload, the photo frame device performs local rendering and returns display confirmation information to the platform.

6. The method for credibility analysis of home-based elderly care health summaries based on electronic photo frames according to claim 1, characterized in that, Step S500 includes the following steps: After receiving the display payload, the electronic photo frame device parses and processes it to extract the health summary content, interactive control set, display template identifier, display template version information, display validity period, display language information, and reference chain information; it also verifies the structural integrity of the display payload, including field integrity verification, field type validity verification, and display validity period validity verification; if the verification fails, it generates a payload anomaly record and returns an anomaly status information to the platform, terminating the current display process; Based on the display template identifier and version information carried in the display payload, the corresponding rendering template is retrieved from the local template cache. If the corresponding version of the rendering template does not exist locally, a template synchronization request is initiated to the platform to obtain the corresponding version of the rendering template and cache it locally. After the template is loaded, an integrity check is performed on the rendering template, which includes template file integrity check, version consistency check, and template structure legality check. If the integrity check fails, the previously available version of the rendering template saved locally is called to perform rollback processing, and a template exception log is recorded. Based on the validated rendering template, the layout parsing and interface parameter calculation processing of the display interface are performed. The layout parsing and interface parameter calculation processing includes determining the position, size and spacing information of each display area according to the screen resolution, display orientation and template layout parameters; the display interface is divided into a summary information area and an interactive control area, wherein the summary information area is located at the top of the screen and is used to display health summary content, and the interactive control area is located at the bottom of the screen and is used to display user interaction controls. The health summary content undergoes text formatting and readability adaptation processing, which includes determining font size, font type, character spacing, line spacing, and paragraph spacing parameters according to preset readability rules. The preset readability rules include using a font size no smaller than the minimum readable font size, using a high-contrast foreground and background color combination, and using a sans-serif font. Key prompt fields, suggested action fields, and time prompt fields in the summary content are displayed differently according to preset emphasis rules. When the health summary content exceeds the display range of a single screen, pagination is performed on the health summary content. The pagination process includes pagination based on sentence boundaries, semantic segment boundaries, or paragraph boundaries to avoid semantic information being truncated at the pagination position; generating page number identifiers and page turning prompts for each page after pagination; and loading touch page turning controls or sliding page turning interaction areas in the interface to support users to perform page turning operations. The system collects the current ambient light intensity information of the electronic photo frame device and performs adaptive screen brightness adjustment processing based on the ambient light intensity information. The adaptive screen brightness adjustment processing includes increasing the screen brightness to the corresponding high brightness level when the ambient light intensity is higher than a preset high brightness threshold; reducing the screen brightness to a preset eye-protection brightness range when the ambient light intensity is lower than a preset low brightness threshold; and linearly or segmentally adjusting the screen brightness according to a preset continuous mapping rule when the ambient light intensity is between the preset high brightness threshold and the preset low brightness threshold. After adjusting the screen brightness, the final display interface is generated based on the rendering template, layout calculation results, and text layout results, and then the final display interface is output to the electronic photo frame display screen for display. After the display is completed, the rendering completion time, the current display page identifier, the display template version information, and the device status information are recorded to generate a display execution record and return display confirmation information to the platform. Simultaneously, an interactive monitoring process is initiated, and user interaction behavior is collected within a preset interactive time window. The interactive monitoring process includes touch interaction acquisition and voice interaction acquisition. The touch interaction acquisition includes detecting the user's touch input on the display interface, obtaining touch start time, touch end time, touch position coordinates, touch trajectory, and touch duration information; determining whether an interactive control is hit based on the mapping relationship between the touch position and the interface control area, and recording the identifier of the hit control; performing abnormal touch filtering on inputs with touch duration below a preset lower limit, abnormal touch trajectory, or touch position exceeding the effective interactive area; and generating a corresponding touch intent label based on preset control intent mapping rules when a valid interactive control is hit. The voice interaction acquisition includes acquiring user voice input signals within a preset voice monitoring period; performing voice activity detection on the voice input signals to determine valid voice segments; performing voice recognition processing on the valid voice segments to generate corresponding recognized text; performing intent recognition processing based on the recognized text to obtain corresponding voice intent labels and intent confidence levels; simultaneously extracting speaker feature information from the voice input and comparing it with authorized user feature templates pre-stored on the device side to generate identity consistency results; when the intent confidence level is lower than a preset threshold, the corresponding voice interaction result is marked as an uncertain intent. The touch interaction results and voice interaction results are summarized according to a unified field structure to generate a unified interaction record. When multiple interaction records exist within the same preset interaction time window, they are associated and integrated in chronological order. When no valid touch interaction or valid voice interaction is detected within the preset interaction time window, the current user status is determined to be non-responsive, and a non-responsive identifier is generated. The non-responsive identifier, along with the display payload identifier, device identifier, and timestamp information, is reported to the platform. The platform then performs confirmation, delay processing, manual call, or re-reminder task generation based on the interaction results.

7. The method for credibility analysis of home-based elderly care health summaries based on electronic photo frames according to claim 1, characterized in that, Step S600 includes the following steps: Feature extraction is performed on each interaction signal to generate a signal quality score. The touch signal quality score includes the touch input signals of the electronic photo frame device collected within a preset interaction time window, including touch start time, end time, touch position and trajectory, and duration information. Based on the mapping relationship between the touch position and the interface control area, it is determined whether a valid control is hit, and abnormal touch inputs with touch duration below a preset threshold, abnormal trajectory, or position outside the valid area are excluded. For touch inputs that hit valid controls, a touch validity sub-score is calculated based on the control hit status and dwell time. The speech signal quality score includes: acquiring user speech input signals within a preset speech monitoring period; detecting speech activity in the speech signals and identifying valid speech segments; performing speech recognition on valid speech segments to generate recognized text and performing intent recognition to obtain speech intent labels and intent confidence; extracting speaker feature information and comparing it with pre-stored authorized user feature templates on the device to generate voiceprint consistency markers; and generating a speech interaction feature score based on a weighted sum of speech recognition confidence, intent clarity, and voiceprint matching. The respiratory signal quality score includes collecting real-time signals from respiratory-related sensors, extracting respiratory feature parameters, comparing real-time respiratory features with individual baseline respiratory features, and calculating respiratory sensor feature scores. The signal quality scores of the interactive signals are normalized, and each sub-score after normalization is used as a feature weighting coefficient. The signal quality scores are then weighted and summed according to the corresponding feature weighting coefficients to obtain the overall confirmation confidence level.

8. The method for credibility analysis of home-based elderly care health summaries based on electronic photo frames according to claim 1, characterized in that, Step S700 includes the following steps: The signal quality scores corresponding to each interaction signal are summarized, the global signal quality mean is calculated, and the global signal quality mean is compared with a preset quality threshold. When the global signal quality mean is lower than the preset quality threshold, the current overall interaction signal is determined to be unusable, the final correction confidence is forcibly set to zero, and the upgrade review process is triggered. When the global signal quality meets the preset conditions, the consistency detection of touch interaction intent and voice interaction intent is performed, and conflict determination is performed in combination with the voiceprint consistency result; when a semantic conflict or voiceprint inconsistency is detected, an interaction conflict is determined, and a guided correction process is triggered to guide the user to confirm the true intent through another interaction; when any modal interaction is missing or the voice recognition result is in an uncertain state, the corresponding semantic conflict determination does not participate in the conflict determination. Based on the signal quality judgment results and conflict judgment results, the comprehensive confirmation confidence level is corrected to generate a final corrected confidence level. The final corrected confidence level is compared with a preset confirmation threshold and a review threshold. According to the comparison results, a tiered triggering process is executed: when the final corrected confidence level is greater than or equal to the confirmation threshold, a standard closed-loop process is triggered; when the final corrected confidence level is between the review threshold and the confirmation threshold, a secondary confirmation process is triggered; when the final corrected confidence level is lower than the review threshold or no effective response is obtained within a preset time, an escalation review process is triggered.

9. A system for analyzing the credibility of home-based elderly care health summaries based on electronic photo frames, used to implement the method for analyzing the credibility of home-based elderly care health summaries based on electronic photo frames as described in any one of claims 1-8, characterized in that, The system includes an event object module, a display requirement object module, a display payload module, a push display payload module, an identification response module, a comprehensive confirmation confidence module, and a correction process module. The event object module receives event data from various input sources, preprocesses the event data, and generates event objects. The module for displaying demand objects: retrieves individual health context from the health record database based on event objects, and combines it with knowledge base retrieval to match event handling suggestions to generate display demand objects; The display payload module generates a health summary text for elderly users based on the display requirements, generates an interactive control set based on the preset control configuration rules in the display template, and generates a display payload based on the health summary text and the interactive control set. The push display payload module: obtains a set of display payloads to be displayed, calculates the comprehensive priority score of the display payloads, generates a display queue based on the comprehensive priority score, and determines the push display payloads in combination with the target photo frame device status; The identification and response module: loads the corresponding rendering template based on the pushed display payload, generates the final display interface and completes the display; collects user touch and voice interaction behaviors within a preset interaction time window, identifies the interaction intent and generates a unified interaction record; when no valid interaction is detected, it generates a non-response flag and reports it to the platform. The comprehensive confirmation confidence module extracts features from the collected touch signals, voice signals, and physiological signals, calculates the quality scores of each type of signal, and generates a comprehensive confirmation confidence score. The correction process module summarizes the quality scores of each signal, calculates the global average signal quality, compares it with a preset threshold, and determines the correction process.

10. The home-based elderly care health summary credibility analysis system based on electronic photo frames according to claim 9, characterized in that, The display payload module includes a health summary content unit and a display payload generation unit; The health summary content unit: Based on the event type, individual health summary and matching health knowledge items in the display demand object, a health summary text for elderly users is generated. The process of generating health summary text includes determining the corresponding display template identifier according to the event type. Read the template structure corresponding to the display template identifier. The template structure includes an event prompt information field, a suggested action information field, and an individual context field. According to the preset field mapping rules, the event information is filled into the event prompt information field, the suggested content in the health knowledge items is filled into the suggested action information field, and the individual health summary and work and rest preference information is filled into the individual context field. The system performs statement simplification and readability adjustments on the filled text content to suit the reading habits of elderly users, performs sensitive information filtering and verification on the generated health summary text, and outputs health summary content that conforms to preset security rules. The display payload generation unit generates a set of interactive controls associated with the health summary content according to the preset control configuration rules in the display template. The process of generating the interactive control set includes parsing the control type configuration and trigger condition configuration in the display template; determining the control types to be loaded according to the urgency and type of the event; and binding corresponding behavior processing logic to the confirmation, delayed processing, and manual call controls respectively. The confirmation control is used to record that the elderly user has read the current health summary; the delayed processing control is used to record the behavior of not responding temporarily, and generates a reminder task based on the event urgency level and preset time interval. The manual call control is used to trigger contact notifications or nursing staff intervention processes; Based on the health record fields, health knowledge items, rule version information, model version information, display template version information, and strategy version information referenced in this demonstration process, corresponding reference chain information is constructed. The health summary content, interactive control set, reference chain information, display validity period, and display language information are then structured and encapsulated to generate the display payload.

Images