AI-powered intelligent companion-style touch-screen emergency voice inspection system

Abstract

This invention relates to the field of narrowband Internet of Things (IoT) and discloses an AI-powered intelligent companion-style touch-based emergency voice inspection system. The system includes: establishing corresponding inspection record units based on communication status and interaction results; tracking changes between adjacent inspection record units and dividing the system into different levels of communication impact zones based on changes in network access status; adjusting the triggering method and message sending order of subsequent inspection tasks for each communication impact zone, and recording effective response results under different scheduling methods; calculating the reliability of response results within each communication impact zone based on the corresponding relationships, and identifying the differences between communication anomalies-induced disconnection and potential disability-induced disconnection; revising the inspection task scheduling strategy based on newly generated communication statuses, effective response results, and disability assessment candidate results in subsequent inspection cycles, and re-executing the communication impact zone division process. This invention has the advantage of improving the accuracy of potential disability identification.

Description

Technical Field

[0001] This invention relates to the field of narrowband Internet of Things (IoT), specifically to an AI-powered intelligent, companion-style touch-based emergency voice inspection system. Background Technology

[0002] With the accelerating aging of the population, an increasing number of people living alone are in a state of prolonged isolation without care. Existing care systems for people living alone typically rely on camera monitoring, wearable devices, or scheduled telephone follow-ups for safety monitoring. In practice, although individuals living alone may not experience obvious abnormal events such as falls or unconsciousness, they may remain in a static state for extended periods due to physical discomfort, temporary disability, drug reactions, hypoglycemia, sudden dizziness, or emotional disturbances. While they retain some awareness, they may struggle to promptly dial, press buttons for help, or make voice calls. Furthermore, the magnitude of their activity patterns and changes in vital signs may not reach traditional alarm thresholds, preventing existing systems from accurately identifying potential risks.

[0003] Existing systems typically classify no response as an abnormal state and trigger alarms, resulting in a high false alarm rate. However, simply extending the waiting time may cause missed opportunities for genuine rescue. In narrowband IoT communication environments, terminal devices are generally constrained by low-power operation requirements and limited communication resources, making it difficult to continuously conduct high-frequency voice interactions and upload complex data. This makes it even more difficult to distinguish between the actual disability status of individuals living alone, the state of environmental noise interference, and the normal unresponsive state.

[0004] Therefore, it is essential to design an AI-powered intelligent companion touch-screen emergency voice inspection system that improves the accuracy of identifying potential disability states. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides an AI-powered intelligent companion touch-screen emergency voice inspection system for single individuals, which has the advantage of improving the accuracy of identifying potential disability states and solves the problems mentioned in the background technology.

[0006] To achieve the aforementioned goal of improving the accuracy of identifying potential disability states, this invention provides the following technical solution: an AI-powered intelligent companion-style touch-based emergency voice inspection system for single individuals, comprising: Inspection and filing module: The NB-IoT terminal periodically reports touch interaction records, voice inspection results and network access status, and counts uplink latency, message retransmission times and response completion status in each inspection cycle. The module establishes corresponding inspection record units based on communication status and interaction results. Impact Classification Module: Tracks changes between adjacent inspection record units. When the response completion rate decreases and the number of message retransmissions continues to increase, it identifies uncertain segments in the corresponding inspection link and divides them into different levels of communication impact zones based on changes in network access status. The scheduling and control module adjusts the triggering method and message sending order of subsequent inspection tasks for each communication impact zone. It prioritizes issuing touch confirmation commands to high-level communication impact zones and maintains voice inspection mode for low-level communication impact zones. It records the effective response results formed under different scheduling methods and establishes the correspondence between communication impact zones and effective response results. Trustworthy discrimination module: Calculates the trustworthiness of response results in each communication impact area based on the correspondence, performs correlation analysis between low-trust response areas and continuously missing response areas, identifies the difference characteristics between the disconnection state caused by communication anomalies and the disconnection state caused by potential disability, and forms candidate results for disability determination; Strategy Correction Module: Corrects the inspection task scheduling strategy based on the new communication status, effective response results and candidate results of disability determination generated in subsequent inspection cycles, re-executes the communication influence zone division process, and outputs the potential disability status identification results corresponding to the current inspection cycle.

[0007] Preferably, the process of statistically analyzing uplink latency, message retransmission count, and response completion status within each inspection cycle is as follows: NB-IoT terminal devices with touch interaction and voice inspection functions are deployed in the living environment of people living alone, and a communication connection is established with the remote inspection platform. According to the preset inspection cycle, the inspection task is sent to the terminal, triggering the terminal to perform touch confirmation interaction or voice question and answer interaction. Record the time when the inspection task is issued, the time when the terminal responds, and the time when the data is transmitted back, and calculate the uplink latency within the corresponding inspection cycle; Statistics were compiled on the number of message retransmissions, successful responses, and failed responses during the NB-IoT communication process for the same inspection task. Based on the completion status of touch confirmation, voice response, and data transmission, the response completion status of the corresponding inspection task is determined, and communication status data for the corresponding inspection cycle is generated.

[0008] Preferably, the process of establishing corresponding inspection record units based on communication status and interaction results is as follows: Call the communication status data to extract the uplink latency, message retransmission count and network access status corresponding to each inspection cycle; Extract touch interaction records, voice inspection results, and response completion status within the corresponding inspection cycle; The communication status parameters and interaction result parameters within the same inspection cycle are time-aligned to establish a unified data index, and corresponding inspection record units are generated according to the inspection cycle order.

[0009] Preferably, the process of identifying uncertain sections in the corresponding inspection link is as follows: Call the inspection record unit to continuously obtain the response completion status and message retransmission count from multiple adjacent inspection record units; Based on the changes in response completion rate and message retransmission frequency between adjacent inspection cycles, corresponding inspection change characteristics are formed. Filter out inspection record units where the response completion rate continues to decline and the number of message retransmissions continues to increase; The number of consecutive cycles and the intensity of change corresponding to the statistical inspection record unit are used to form abnormal change sections, and the inspection links corresponding to the abnormal change sections are marked as uncertain sections.

[0010] Preferably, the process of dividing communication impact zones into different levels based on changes in network access status is as follows: Extract the network access status information corresponding to each inspection record unit within the uncertain section; Statistics were collected on the number of times the NB-IoT terminal changed its dwell status, the number of cell handovers, the signal quality level, and the number of connection interruptions within the corresponding time period. The communication impact index is calculated based on the degree of fluctuation in network access status. The communication impact zone is divided into high-level, medium-level, and low-level communication impact zones based on the magnitude of the communication impact index.

[0011] Preferably, the process of recording the effective response results generated under different scheduling methods is as follows: For high-level communication impact areas, prioritize triggering touch confirmation inspection tasks and increase the priority of sending corresponding messages; For medium-level communication impact areas, a combination of touch confirmation and voice inspection is used. For areas affected by low-level communication, maintain the voice inspection mode and perform inspection tasks according to the original inspection cycle. The touch response results, voice response results, and response success rates generated under each scheduling method are statistically analyzed to form the corresponding effective response results.

[0012] Preferably, the process of establishing the correspondence between the communication influence area and the effective response results is as follows: Based on the effective response results, the response success rate, response latency, and response integrity rate in different levels of communication impact areas were statistically analyzed. Calculate the effective response contribution value corresponding to each communication impact area, and construct an impact mapping matrix based on the correlation between communication impact level and effective response contribution value; The influence mapping matrix is ​​used to establish the correspondence between the communication influence area and the effective response results, and to form the basic data for response credibility analysis.

[0013] Preferably, the process of performing correlation analysis between low-confidence response regions and continuously missing response regions is as follows: Call upon the basic data for response reliability analysis to extract the response success rate, response integrity rate, and response latency corresponding to each communication impact zone; Response credibility score is calculated based on response success rate, response integrity rate, and response latency. The communication impact area with a credibility score lower than the preset credibility threshold is marked as a low-credibility response area; Areas that do not receive effective response results within multiple consecutive inspection cycles are identified as continuously missing response areas. The degree of overlap and temporal correlation between low-confidence response regions and continuously missing response regions are analyzed to form the correlation results of disconnection features.

[0014] Preferably, the process for forming candidate results for disability assessment is as follows: Call the loss of contact feature association results to extract communication status change features, response behavior change features and inspection history records in the corresponding area; Determine whether there is continuous signal attenuation, frequent reconnection, or network access interruption based on the characteristics of communication status changes. Based on the characteristics of changes in response behavior, determine whether there are missing touch responses, missing voice responses, or a continuous increase in response latency; When communication anomalies are significant and historical behavior patterns remain stable, the state of being out of contact is determined to be caused by communication anomalies. When the communication status remains normal but the response behavior is continuously missing, it is determined to be a disconnection state caused by potential disability, and a corresponding disability determination candidate result is generated.

[0015] Preferably, the process of outputting the potential disability status identification result corresponding to the current inspection cycle is as follows: Acquire new communication status data, effective response results and candidate results for disability determination in subsequent inspection cycles, and statistically analyze the trend of response reliability change and the frequency of disability determination in each communication impact area. Adjust the frequency of inspection tasks, the priority of touch confirmation, and the interval of voice inspection in the corresponding areas according to the changing trends. The communication impact zone delineation process was re-executed using the revised inspection task scheduling strategy, and the communication impact zone level was updated. By combining the updated communication impact zone level, response credibility score, and candidate results for disability determination, a comprehensive analysis is performed to output the identification results of potential disability status corresponding to the current inspection cycle, and the identification results are sent to the monitoring terminal.

[0016] Compared with existing technologies, this invention provides an AI-powered intelligent companion-style touch-based emergency voice inspection system for single-person living, which has the following beneficial effects: This invention integrates and analyzes NB-IoT communication status, touch interaction behavior, and voice inspection results to construct a correlation mechanism between inspection recording units, communication impact zones, and candidate failure judgment results. This enables the differentiation between communication anomalies and actual user disconnection, avoiding false alarms caused by network fluctuations, signal attenuation, or terminal retransmissions. By dynamically identifying uncertain segments in the inspection link and adaptively adjusting the triggering strategies for touch confirmation and voice inspection for different levels of communication impact zones, it improves the completion rate and response acquisition capability of inspection tasks in complex communication environments. Furthermore, by maintaining the reliability of effective response results… Continuous evaluation and correlation analysis between low-reliability response areas and continuously missing response areas can promptly detect abnormal signs when users experience limited mobility, abnormal consciousness, or risk of disability, improving the accuracy of potential disability identification and early warning capabilities. Based on the newly generated communication status, response results, and judgment results in subsequent inspection cycles, the inspection scheduling strategy is continuously revised to form a self-learning optimization mechanism, enabling the inspection method to adapt to different user behavior habits and changes in different network environments. This improves the reliability, stability, and intelligence level of emergency inspections in scenarios involving companionship for people living alone, reduces the cost of manual intervention, and enhances the safety protection capabilities of people living alone. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the method of 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] Example 1: Please refer to Figure 1 The AI-powered intelligent companion-style touch-based emergency voice inspection system for single-person living, as described in this embodiment of the invention, includes: Inspection and filing module: The NB-IoT terminal periodically reports touch interaction records, voice inspection results and network access status, and counts uplink latency, message retransmission times and response completion status in each inspection cycle. Based on the communication status and interaction results, the module establishes corresponding inspection record units.

[0020] The process of statistically analyzing uplink latency, message retransmission count, and response completion status within each inspection cycle in the inspection and documentation module is as follows: NB-IoT terminal devices with touch interaction and voice inspection functions are deployed in the living environment of people living alone, and a communication connection is established with the remote inspection platform. NB-IoT terminal devices are installed in the living room, bedroom or other areas where people living alone have high daily activity frequency. The terminal has built-in touch buttons, voice broadcast unit, microphone acquisition unit and NB-IoT communication module. The terminal accesses the cellular mobile network through the operator's narrowband IoT base station, completes device registration with the remote inspection platform, obtains the corresponding device identifier and communication authentication parameters, and establishes a periodic heartbeat reporting mechanism. It maintains a low power standby state during non-inspection periods and enters a communication active state during inspection periods to ensure that subsequent inspection commands can be sent to the corresponding terminal through the NB-IoT network. According to the preset inspection cycle, the inspection platform issues inspection tasks to the terminal, triggering the terminal to perform touch confirmation interaction or voice question and answer interaction; the remote inspection platform generates inspection tasks according to the preset inspection plan and sends inspection trigger instructions to the terminal through the NB-IoT downlink. After receiving the inspection trigger instructions, the terminal exits the low power mode and executes the corresponding inspection process. When the touch confirmation method is used, the terminal broadcasts the prompt information and waits for the user to press the confirmation button. When the voice question and answer method is used, the terminal automatically plays the inspection voice content and collects the user's response voice. During the inspection process, the inspection task number, task type and task execution status are recorded simultaneously. The system records the time of inspection task issuance, terminal response, and data transmission, and calculates the uplink latency within the corresponding inspection cycle. When the remote inspection platform sends an inspection command, it records the task issuance timestamp. When the terminal detects that the user has completed touch confirmation or voice response, it records the terminal response timestamp. After completing data encapsulation, the terminal transmits the inspection results back to the platform via the NB-IoT network. After receiving the transmitted data, the platform records the data arrival timestamp. The system calculates the uplink transmission latency based on the time difference between the terminal response timestamp and the data arrival timestamp, and calculates the user response latency based on the time difference between the task issuance timestamp and the terminal response timestamp, thus forming the latency characteristic data for the corresponding inspection cycle. The system tracks the number of message retransmissions, successful responses, and failed responses during NB-IoT communication for the same inspection task. After sending inspection result data, the terminal waits for confirmation from the platform. If no confirmation is received within a preset time window, the terminal automatically retransmits the message. The terminal records each data transmission and retransmission process and writes the retransmission count to the communication log. The platform tracks the number of successful responses based on the number of complete data packets received and the number of failed responses based on timeouts, data packet verification failures, or task interruptions. By summarizing and analyzing the number of message retransmissions, successful responses, and failed responses, the system reflects the wireless communication quality and network stability corresponding to the current inspection cycle. Based on the completion status of touch confirmation, voice response, and data transmission, the response completion status of the corresponding inspection task is determined, and communication status data for the corresponding inspection cycle is generated. When the terminal receives a valid touch confirmation or valid voice response within a specified time and the corresponding inspection result is completely transmitted back to the remote inspection platform, the inspection task is marked as a fully responsive state. When only part of the interaction process is completed or the transmitted data has missing fields, the inspection task is marked as a partially responsive state. When no valid interaction result is obtained or the data is not successfully transmitted back within a specified time, the inspection task is marked as a non-responsive state. Communication status data is generated by combining the uplink latency, message retransmission count, number of successful responses, and number of failed responses within the inspection cycle.

[0021] The process of establishing corresponding inspection record units based on communication status and interaction results in the inspection and filing module is as follows: The system retrieves communication status data to extract uplink latency, message retransmission count, and network access status for each inspection cycle. It also reads communication log information for each inspection cycle from the communication status data to extract the uplink latency from the time the terminal responds until the platform receives the data, as well as the number of message retransmissions generated by the NB-IoT terminal during the same inspection task. This allows the system to obtain network access status information for the terminal within the inspection cycle, including normal attachment status, weak coverage attachment status, frequent reconnection status, and access failure status. To eliminate the impact of instantaneous network fluctuations, a sliding statistical method is used to calculate the uplink latency change rate and message retransmission change rate for multiple consecutive inspection cycles, thereby forming communication status characteristics that reflect the trend of communication quality changes. Extract touch interaction records, voice inspection results, and response completion status within the corresponding inspection cycle; read touch confirmation records and voice inspection records from the inspection result data returned by the terminal. Touch interaction records include touch trigger time, number of touches, and touch completion indicator. Voice inspection results include voice playback status, voice response status, and voice recognition validity indicator. Read response completion status data generated by the inspection filing module, and standardize the encoding of complete response, partial response, and no response status to form interaction result features corresponding to the communication status feature set. The communication status parameters and interaction result parameters within the same inspection cycle are time-aligned to establish a unified data index. Corresponding inspection record units are generated according to the inspection cycle order. The communication status feature set and the interaction result feature set are matched and associated using the inspection task number and inspection cycle number as the association primary key. For data items with time deviations, time window calibration is performed according to the task issuance timestamp to make the communication status parameters and interaction result parameters correspond to the same inspection cycle. A unified data index structure is established, and uplink latency, message retransmission count, network access status, touch interaction records, voice inspection results and response completion status are encapsulated into inspection record units and formed according to the order of the inspection cycle. Impact Classification Module: Tracks changes between adjacent inspection record units. When the response completion rate decreases and the number of message retransmissions continues to increase, it identifies uncertain segments in the corresponding inspection link and divides them into different levels of communication impact zones based on changes in network access status.

[0022] The process of identifying uncertain sections in the corresponding inspection link in the impact segmentation module is as follows: The system invokes inspection record units to continuously acquire response completion status and message retransmission counts from multiple adjacent inspection record units. Multiple consecutive inspection record units are read from the inspection records in chronological order, and the response completion status and message retransmission count for each unit are extracted. Response completion status is quantified using three levels: complete response, partial response, and no response, for example, assigned values ​​of 1, 0.5, and 0 respectively. The message retransmission count is directly read from the actual retransmission count in the NB-IoT communication log. To avoid the influence of occasional fluctuations in single inspection results on the analysis results, a continuous inspection cycle of 3 to 10 cycles is set as an observation window. Data from each inspection record unit within the window is centrally analyzed to form a corresponding continuous inspection sequence. Based on the changes in response completion rate and message retransmission count between adjacent inspection cycles, corresponding inspection change characteristics are formed. Using adjacent inspection record units as the analysis objects, the change values ​​of response completion rate and message retransmission count are calculated. The change value of response completion rate is used to reflect the degree of change in the user's ability to complete inspection interactions, and the change value of message retransmission count is used to reflect the degree of change in the quality of the NB-IoT wireless link. The cumulative direction and slope of the change values ​​within multiple consecutive inspection cycles are statistically analyzed. When the response completion rate shows a downward trend, the downward slope is recorded. When the message retransmission count shows an upward trend, the upward slope is recorded. The downward slope, the upward slope, and the duration of the cycle are combined to form the inspection change characteristics to represent the overall change status of the inspection link in the current stage. The system filters out inspection record units that show a continuous decrease in response completion rate and a continuous increase in message retransmission count. It then searches for inspection record units that simultaneously meet the conditions of a continuous decrease in response completion rate and a continuous increase in message retransmission count among the inspection change characteristics. A continuous decrease can be defined as the response completion rate consistently being lower than the previous period across multiple adjacent inspection cycles, and a continuous increase can be defined as the message retransmission count consistently being higher than the previous period. When the number of inspection record units that continuously meet the conditions reaches a preset threshold, the corresponding inspection record units are extracted to form an abnormal change set. To avoid misjudgment due to a single network fluctuation, the system verifies whether the network attachment status changes synchronously within the corresponding period, retaining only inspection record units that simultaneously exhibit interactive response attenuation characteristics and communication quality deterioration characteristics. The number of consecutive cycles and the intensity of change corresponding to the inspection record units are statistically analyzed to form abnormal change segments. The inspection links corresponding to the abnormal change segments are marked as uncertain segments. The inspection record units in the abnormal change set are connected in chronological order, and the number of inspection cycles in which continuous abnormal states exist is counted as the number of consecutive cycles. The intensity of change is calculated based on the cumulative decrease in response completion rate, the cumulative increase in message retransmission times, and the degree of deterioration in network access status. When the number of consecutive cycles exceeds a preset cycle threshold and the intensity of change reaches a preset intensity threshold, the corresponding time range is defined as an abnormal change segment. Since there are two possible reasons in this segment, namely the decline in user interaction capability and the deterioration of NB-IoT communication status, it is temporarily impossible to directly distinguish whether it belongs to communication abnormality or potential failure status. Therefore, the inspection links corresponding to this abnormal change segment are marked as uncertain segments.

[0023] The process of dividing communication impact zones into different levels based on changes in network access status in the impact classification module is as follows: Extract network access status information corresponding to each inspection record unit within the uncertain segment; read the network operation log of the corresponding inspection record unit from the inspection link marked as the uncertain segment, and extract the network access status information of the NB-IoT terminal in each inspection cycle. The network access status information includes terminal attachment status, RRC connection status, PSM mode exit status, eDRX wake-up status, and base station access result. Sort the network access status by time according to the inspection cycle and establish an access status sequence corresponding to the uncertain segment to reflect the continuous access process of the terminal during abnormal changes. The system counts the number of times the NB-IoT terminal changes its camping state, the number of cell handovers, the signal quality level, and the number of connection interruptions within the corresponding time period. It performs a traversal analysis of the access state sequence, counts the number of times the terminal hands over between the serving cell and neighboring cells, and records the time node corresponding to each handover. It also counts the number of times the terminal transitions between connected state, idle state, PSM state, and eDRX state to form the number of camping state changes. It reads the wireless signal measurement parameters reported by the NB-IoT module, including the reference signal received power (RSRP), the reference signal received quality (RSRQ), and the signal-to-noise ratio (SNR), and divides the corresponding signal quality levels according to preset intervals. For cases of access failure, reconstruction failure, or continuous timeout without response, it records the corresponding number of connection interruptions, thereby forming communication fluctuation characteristics that characterize the stability of the wireless link. The communication impact index is calculated based on the fluctuation of network access status. The number of times the camping status changes, the number of cell handovers, the signal quality level, and the number of connection interruptions are normalized to ensure that each parameter is within a unified dimension. Different parameters are assigned corresponding weights according to their impact on communication stability. The number of connection interruptions and the signal quality level can be assigned higher weights, while the number of cell handovers and the number of times the camping status changes can be assigned lower weights. The communication impact index is calculated by weighted accumulation so that it can comprehensively reflect the network fluctuation of NB-IoT terminals in the current uncertain segment. When the terminal frequently experiences connection interruptions, continuous decline in signal quality, or frequent network status changes, the corresponding communication impact index increases synchronously. The communication impact zone is divided into high-level, medium-level, and low-level communication impact zones based on the magnitude of the communication impact index. The communication impact index is compared with a pre-established threshold for level classification. When the communication impact index exceeds the high-level threshold, the corresponding uncertain section is determined to be a high-level communication impact zone, indicating that the current inspection results may be severely affected by network anomalies. When the communication impact index is within the medium-level threshold range, it is determined to be a medium-level communication impact zone, indicating that the communication link has significant fluctuations but still has a certain data transmission capability. When the communication impact index is below the medium-level threshold, it is determined to be a low-level communication impact zone, indicating that the communication environment is basically stable and the inspection results have high reliability.

[0024] The scheduling and control module adjusts the triggering method and message sending order of subsequent inspection tasks for each communication impact zone. It prioritizes issuing touch confirmation commands to high-level communication impact zones and maintains voice inspection mode for low-level communication impact zones. It records the effective response results formed under different scheduling methods and establishes the correspondence between communication impact zones and effective response results.

[0025] The process of recording valid response results under different scheduling modes in the scheduling control module is as follows: For high-level communication impact areas, touch confirmation inspection tasks are triggered first, and the corresponding message sending priority is increased. When the area corresponding to the inspection link is determined to be a high-level communication impact area, the remote inspection platform suspends the inspection method based on voice Q&A and prioritizes the issuance of touch confirmation inspection tasks. After receiving the inspection task, the terminal guides the person living alone to perform button confirmation, touch panel confirmation, or designated area touch confirmation operation through local audio and light prompts. On the NB-IoT network side, the sending priority of the corresponding inspection data packet is increased, the data queuing time is shortened, and the repeated transmission resource priority allocation mechanism is adopted to improve the transmission success rate of key inspection messages. Since the amount of touch confirmation data is much lower than the amount of voice data, the probability of inspection task completion can be increased in weak coverage environment and high retransmission environment, thereby obtaining more reliable response results. For medium-level communication impact areas, a combination of touch confirmation and voice inspection is used. When the communication impact level is in the medium range, the remote inspection platform organizes inspection tasks according to a preset alternation strategy. For example, touch confirmation is performed in the current inspection cycle, and voice inspection is performed in the next inspection cycle, or touch confirmation is performed first and then voice inspection is performed in the same inspection cycle. By alternating different types of interaction methods, the system can simultaneously obtain low-data-volume touch feedback information and high-data-volume voice feedback information. If touch confirmation is successful but voice response fails, it can be preliminarily determined that the user has the ability to interact but there may be a problem with voice data transmission in the communication link. If both fail, the frequency of subsequent inspections is further increased to enhance the ability to track abnormal states. For low-level communication impact areas, the voice inspection mode is maintained, and inspection tasks are performed according to the original inspection cycle. When the corresponding area is determined to be a low-level communication impact area, it indicates that the NB-IoT network connection status is stable, the number of message retransmissions is small, and the communication impact index is low. Therefore, the original voice inspection mode is maintained. The remote inspection platform sends voice inspection tasks to the terminal according to the established inspection plan cycle. The terminal automatically plays the inspection voice content and collects the user's response voice. After extracting the response result through the voice recognition module, it is sent back to the platform. In this mode, there is no need to add an additional touch confirmation process, thereby reducing the user's operational burden and reducing the terminal's power consumption and network resource occupation, so as to achieve long-term continuous inspection under normal conditions. The system statistically analyzes touch response results, voice response results, and response success rates under various scheduling methods to form corresponding effective response results. The inspection platform separately analyzes inspection execution results under different scheduling methods. Touch response results include the number of successful touch confirmations, the number of failed touch confirmations, and the average response time. Voice response results include the number of successful voice responses, the number of valid voice recognitions, and the voice inspection completion rate. Simultaneously, the response success rate under the corresponding scheduling method is calculated. The response success rate can be obtained by comparing the number of successfully completed inspection tasks with the total number of inspection tasks. The touch response results, voice response results, response success rates, and corresponding communication impact zone levels are correlated to form effective response results. This not only reflects the execution effect of different scheduling strategies in the actual communication environment but also provides a basis for subsequent response reliability calculations and potential disability status identification.

[0026] The process of establishing the correspondence between the communication impact zone and the effective response result in the scheduling and control module is as follows: Based on the effective response results, the response success rate, response latency, and response integrity rate are statistically analyzed within different levels of communication impact zones. Inspection result data corresponding to high-level, medium-level, and low-level communication impact zones are extracted from the effective response result dataset and categorized for statistical analysis. The response success rate is obtained by the ratio of the number of successfully completed inspection tasks to the total number of inspection tasks. The response latency is obtained by the time difference between the time the inspection task is issued and the time the terminal completes a valid response. The response integrity rate is obtained by the ratio of the number of validly returned interactive data items to the total number of data items that should be returned. Statistical result sets are established according to the communication impact zone level to allow the system to intuitively reflect the actual execution effect of inspection tasks under different communication environments. For example, in high-level communication impact zones, there may be a lower response success rate and a longer response latency, while in low-level communication impact zones, there is usually a higher response success rate and a shorter response latency. The effective response contribution value corresponding to each communication impact zone is calculated, and an impact mapping matrix is ​​constructed based on the correlation between the communication impact level and the effective response contribution value. The effective response quality in different communication impact zones is comprehensively evaluated using response success rate, response latency, and response integrity rate as evaluation indicators. The higher the response success rate and response integrity rate, the higher the corresponding contribution level, and the greater the response latency, the lower the corresponding contribution level. The difference in the dimensions between different indicators is eliminated by normalization, and weighted fusion is performed according to preset weights to obtain the effective response contribution value corresponding to each communication impact zone. The communication impact level is used as the matrix row, and the effective response contribution value range is used as the matrix column. The statistical results formed during the historical inspection process are filled into the corresponding positions of the matrix to establish an impact mapping matrix that reflects the relationship between changes in the communication environment and changes in response effect, which can represent the formation pattern of effective response results under different communication conditions. An influence mapping matrix is ​​used to establish a correspondence between communication influence areas and effective response results, forming basic data for response credibility analysis. When a new effective response result is generated, the influence mapping matrix is ​​called to retrieve the historical contribution distribution under the corresponding communication influence level. The matching degree between the current effective response result and historical statistical characteristics is calculated. When a response result is consistent with most historical results under the corresponding communication influence level, it is determined to have a high degree of credibility. When the response result deviates significantly from the historical distribution pattern under the corresponding communication influence level, its credibility is reduced. The communication influence area level, effective response contribution value, response success rate, response delay, response integrity rate, and matching degree are correlated to form basic data for response credibility analysis. This basic data is used to subsequently distinguish between response loss caused by communication anomalies and response loss caused by potential disability status, thereby improving the accuracy of identifying potential disability status of people living alone.

[0027] Trustworthy discrimination module: Calculates the trustworthiness of response results in each communication impact area based on the correspondence, performs correlation analysis between low-trust response areas and continuously missing response areas, identifies the difference characteristics between the disconnection state caused by communication anomalies and the disconnection state caused by potential disability, and forms candidate results for disability determination.

[0028] The process of correlation analysis between low-confidence response regions and continuously missing response regions in the confidence discrimination module is as follows: The system retrieves basic data for response reliability analysis and extracts the response success rate, response integrity rate, and response latency corresponding to each communication impact zone. It then reads the statistical results corresponding to each communication impact zone from the basic data and categorizes them according to the level of the communication impact zone. The response success rate reflects the completion status of the inspection task, the response integrity rate reflects the completeness of the returned data, and the response latency reflects the time consumption between task triggering and the establishment of a valid response. For cases where multiple inspection cycle records exist within the same communication impact zone, a response feature sequence is established in chronological order, and the changes of each indicator in consecutive inspection cycles are calculated, thus forming the basic response features used for reliability analysis. The response reliability score is calculated based on the response success rate, response integrity rate, and response latency. The response success rate, response integrity rate, and response latency are standardized to ensure that each indicator is within a unified evaluation scale. They are then weighted according to preset weights, with the response success rate and response integrity rate assigned positive weights and the response latency assigned negative weights. This yields the response reliability score for the corresponding inspection cycle. When the response success rate, response integrity rate, and response latency are high, the response reliability score increases accordingly. Conversely, when the response success rate decreases, the response integrity rate decreases, or the response latency increases significantly, the response reliability score decreases accordingly. By continuously calculating the response reliability score for each inspection cycle, a sequence of response reliability changes is formed. Communication impact areas with credibility scores below the preset credibility threshold are marked as low-credibility response areas. The response credibility score is compared with the preset credibility threshold. When the response credibility score is below the credibility threshold, it indicates that the current inspection results are affected by factors such as communication fluctuations, data loss, or abnormal responses, and it is difficult to directly reflect the true status of people living alone. Therefore, the corresponding communication impact area is marked as a low-credibility response area. The start inspection cycle, end inspection cycle, and duration of the low-credibility response area are recorded, and the corresponding low-credibility response area is established. Areas that have not received a valid response within multiple consecutive inspection cycles are identified as continuous missing response areas. Inspection record units that have not received a valid response are retrieved from the inspection record link. Valid response results include valid touch confirmation results and valid voice response results. When no valid response results are received within multiple consecutive inspection cycles, the corresponding inspection record units are connected in chronological order to form continuous missing response areas. For each continuous missing response area, the start time of the missing response, the end time of the missing response, the number of duration cycles, and the corresponding communication impact zone level are recorded to reflect the duration and development process of the disconnection state. The overlap and temporal correlation between low-confidence response regions and continuously missing response regions are analyzed to form a disconnection feature association result. The low-confidence response regions and continuously missing response regions are mapped onto a unified time axis, and the overlap ratio and starting time difference between the two in the time range are calculated. When the continuously missing response regions are mainly distributed within the low-confidence response regions and the two have a high degree of temporal overlap, it is determined that the disconnection state and communication anomaly are strongly correlated. When the continuously missing response regions appear during a period of relatively stable communication or the duration significantly exceeds the coverage of the low-confidence response regions, it is determined that the disconnection state and communication anomaly are less correlated. The disconnection feature association result is established by combining the duration, overlap ratio and time offset of each region, which is used to distinguish between response missing caused by communication anomaly and response missing caused by potential disability state, thereby improving the accuracy of potential disability state identification.

[0029] The process of generating candidate results for disability determination in the trustworthy discrimination module is as follows: The system retrieves the loss-of-connection feature association results and extracts communication status change features, response behavior change features, and inspection history records within the corresponding area. It then reads the association analysis data corresponding to the target loss-of-connection area from the loss-of-connection feature association results and extracts communication status change features, response behavior change features, and inspection history records. Communication status change features include changes in uplink latency, message retransmission count, network access status, and communication impact index. Response behavior change features include changes in touch confirmation success rate, voice response completion rate, response latency, and the number of consecutive missing responses. Inspection history records include the response habits, normal response time distribution, common interaction methods, and historical communication quality levels of the corresponding single-person in the historical inspection cycle. This establishes a correspondence between the current loss-of-connection area and historical inspection behaviors. Based on the characteristics of communication status changes, determine whether there is continuous signal attenuation, frequent reconnection, or network access interruption; continuously analyze the characteristics of communication status changes in the corresponding time period of the disconnected area, and statistically analyze the changing trends of reference signal received power (RSRP), reference signal received quality (RSRQ), and signal-to-noise ratio (SNR). When the signal quality continuously declines and falls below the preset communication quality threshold in multiple consecutive inspection cycles, it is determined that there is continuous signal attenuation. When the terminal repeatedly reattaches to the network, re-establishes the RRC connection, or switches the serving cell in a short period of time, it is determined that there is frequent reconnection. When there are continuous access failures, inability to attach to the network for a long time, or inability to establish a communication connection in multiple inspection cycles, it is determined that there is network access interruption. Statistically analyze the duration and frequency of abnormal phenomena to indicate the degree of communication abnormality. Based on the characteristics of changes in response behavior, it is determined whether there are missing touch responses, missing voice responses, or a continuous increase in response latency. The interaction results of each inspection cycle in the area of ​​no contact are tracked and analyzed, and the number of incomplete touch confirmations, incomplete voice responses, and consecutive unresponsive cycles are counted. When no effective touch confirmation results are generated in multiple consecutive inspection cycles, it is determined that there is a missing touch response. When no effective voice response results are obtained in multiple consecutive inspection cycles, it is determined that there is a missing voice response. The trend of response latency is analyzed. When the response latency gradually increases in multiple consecutive inspection cycles and exceeds the historical average response time, it is determined that there is a continuous increase in response latency. The degree of behavioral change is analyzed in combination with historical response habits to reflect the changes in the interaction ability of people living alone. When communication anomalies are significant and historical behavior patterns remain stable, the state of disconnection is determined to be caused by communication anomalies. The characteristics of communication status changes and response behavior changes are analyzed together. When communication anomalies such as continuous signal attenuation, frequent reconnection, or network access interruption reach the preset anomaly threshold, and historical inspection records show that the person living alone maintained a stable response habit before the loss of contact and was able to re-establish normal interaction after communication was restored, the loss of contact is considered to be mainly caused by changes in the NB-IoT network environment. The corresponding area is marked as a state of disconnection caused by communication anomalies, and the corresponding communication anomaly type, duration, and scope of impact are recorded as data basis for subsequent inspection and scheduling optimization. When the communication status remains normal but the response behavior is continuously absent, it is determined to be a disconnection state caused by potential disability, and a corresponding candidate result for disability determination is generated. When the communication link remains in a normal attachment state, the signal quality meets the preset requirements, the number of message retransmissions and uplink latency remain within the normal range, but the absence of touch response, absence of voice response or continuous non-response persists, the main influence of communication abnormal factors on the disconnection result is excluded. The recent changes in the response ability of the person living alone are analyzed in conjunction with the inspection history. When the number of consecutive missing responses continues to increase and the response behavior deviates from the historical behavior pattern, it is determined that there is a potential disability risk, a corresponding candidate result for disability determination is generated, and the risk level, duration, abnormal behavior characteristics and communication status characteristics of the candidate result are recorded.

[0030] Strategy Correction Module: Corrects the inspection task scheduling strategy based on the new communication status, effective response results and candidate results of disability determination generated in subsequent inspection cycles, re-executes the communication influence zone division process, and outputs the potential disability status identification results corresponding to the current inspection cycle.

[0031] The process of outputting the potential disability status identification result corresponding to the current inspection cycle in the strategy correction module is as follows: Acquire newly generated communication status data, effective response results, and candidate results for disability assessment in subsequent inspection cycles. Statistically analyze the trend of response credibility changes and the frequency of disability assessments in each communication impact zone. After the current inspection cycle ends, continuously receive communication status data, effective response results, and candidate results for disability assessments output by the NB-IoT terminal in subsequent inspection cycles. Establish a time series data set according to the correspondence of communication impact zones. Continuously analyze the direction, magnitude, and duration of change in response credibility scores within each communication impact zone to form a trend of response credibility changes. Statistically analyze the cumulative number of occurrences, consecutive occurrences, and frequency of occurrence of candidate results for disability assessments within the corresponding communication impact zone. When a candidate result for disability assessment appears in a certain area for multiple consecutive inspection cycles, record the corresponding frequency of occurrence in the risk tracking queue to reflect the development trend of potential disability risks. Adjust the trigger frequency of inspection tasks, touch confirmation priority, and voice inspection interval for the corresponding area according to the changing trend; dynamically adjust the inspection control parameters according to the changing trend of response reliability and the frequency of disability judgment. When the response reliability continues to decline and the frequency of disability judgment continues to increase, increase the trigger frequency of inspection tasks in the corresponding area, shorten the time interval between adjacent inspection tasks, and increase the scheduling priority of touch confirmation tasks in the inspection task queue in order to quickly obtain the most direct interactive feedback information from users. Appropriately shorten the voice inspection interval and increase the status confirmation density. When the response reliability remains stable and the frequency of disability judgment is low, maintain the original inspection parameter configuration to reduce terminal power consumption and communication resource occupation, and form an inspection task scheduling strategy adapted to the risk status. The communication impact zone division process is re-executed using the revised inspection task scheduling strategy, and the communication impact zone level is updated. The revised inspection task scheduling strategy is applied to subsequent inspection cycles, and communication status data and interaction result data within the corresponding cycle are re-collected. The communication impact index is recalculated according to the communication impact zone division rules. Based on the latest obtained network access status, cell handover status, signal quality level, and connection interruption status, the original communication impact zone is dynamically corrected. When network quality improves and response reliability increases, the corresponding communication impact zone level is reduced. When network fluctuations further intensify or response results continue to deteriorate, the corresponding communication impact zone level is increased. By periodically updating the communication impact zone level, the system can continuously maintain its adaptability to changes in the actual network environment. By comprehensively analyzing the updated communication impact zone level, response credibility score, and disability assessment candidate results, the system outputs the potential disability status identification results corresponding to the current inspection cycle and sends the identification results to the monitoring terminal. The updated communication impact zone level is used as the basis for evaluating communication reliability, the response credibility score as the basis for evaluating response effectiveness, and the disability assessment candidate results as the basis for evaluating behavioral abnormalities. Correlation analysis is performed on the three types of data. When the communication impact zone level is low, the response credibility score continues to decline, and disability assessment candidate results appear consecutively, the potential disability risk is determined to be high. When the communication impact zone level is high and the decline in response credibility score is mainly related to communication abnormalities, the disability risk assessment level is reduced. Based on the comprehensive analysis results, potential disability status identification results are generated, including normal status, attention status, warning status, and emergency status. The identification results are sent to the monitoring terminal, family terminal, or management platform through NB-IoT network, cellular mobile communication network, or Internet communication link, along with the corresponding risk level, occurrence time, and suggested treatment information, to achieve timely early warning and remote intervention for potential disability risks of people living alone.

[0032] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0033] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An AI-powered, intelligent, companion-style touch-screen emergency voice inspection system, characterized in that: include: Inspection and filing module: The NB-IoT terminal periodically reports touch interaction records, voice inspection results and network access status, and counts uplink latency, message retransmission times and response completion status in each inspection cycle. The module establishes corresponding inspection record units based on communication status and interaction results. Impact Classification Module: Tracks changes between adjacent inspection record units. When the response completion rate decreases and the number of message retransmissions continues to increase, it identifies uncertain segments in the corresponding inspection link and divides them into different levels of communication impact zones based on changes in network access status. The scheduling and control module adjusts the triggering method and message sending order of subsequent inspection tasks for each communication impact zone. It prioritizes issuing touch confirmation commands to high-level communication impact zones and maintains voice inspection mode for low-level communication impact zones. It records the effective response results formed under different scheduling methods and establishes the correspondence between communication impact zones and effective response results. Trustworthy discrimination module: Calculates the trustworthiness of response results in each communication impact area based on the correspondence, performs correlation analysis between low-trust response areas and continuously missing response areas, identifies the difference characteristics between the disconnection state caused by communication anomalies and the disconnection state caused by potential disability, and forms candidate results for disability determination; Strategy Correction Module: Corrects the inspection task scheduling strategy based on the new communication status, effective response results and candidate results of disability determination generated in subsequent inspection cycles, re-executes the communication influence zone division process, and outputs the potential disability status identification results corresponding to the current inspection cycle.

2. The AI-powered intelligent companion-style touch-screen emergency voice inspection system according to claim 1, characterized in that, The process of calculating uplink latency, message retransmission count, and response completion status within each inspection cycle is as follows: NB-IoT terminal devices with touch interaction and voice inspection functions are deployed in the living environment of people living alone, and a communication connection is established with the remote inspection platform. According to the preset inspection cycle, the inspection task is sent to the terminal, triggering the terminal to perform touch confirmation interaction or voice question and answer interaction. Record the time when the inspection task is issued, the time when the terminal responds, and the time when the data is transmitted back, and calculate the uplink latency within the corresponding inspection cycle; Statistics were compiled on the number of message retransmissions, successful responses, and failed responses during the NB-IoT communication process for the same inspection task. Based on the completion status of touch confirmation, voice response, and data transmission, the response completion status of the corresponding inspection task is determined, and communication status data for the corresponding inspection cycle is generated.

3. The AI-powered intelligent companion-style touch-screen emergency voice inspection system for single-person living as described in claim 1, characterized in that, The process of establishing corresponding inspection record units based on communication status and interaction results is as follows: Call the communication status data to extract the uplink latency, message retransmission count and network access status corresponding to each inspection cycle; Extract touch interaction records, voice inspection results, and response completion status within the corresponding inspection cycle; The communication status parameters and interaction result parameters within the same inspection cycle are time-aligned to establish a unified data index, and corresponding inspection record units are generated according to the inspection cycle order.

4. The AI-powered intelligent companion-style touch-screen emergency voice inspection system according to claim 1, characterized in that, The process of identifying uncertain sections in the corresponding inspection link is as follows: Call the inspection record unit to continuously obtain the response completion status and message retransmission count from multiple adjacent inspection record units; Based on the change in response completion rate and the change in message retransmission frequency between adjacent inspection cycles, corresponding inspection change characteristics are formed. Filter out inspection record units where the response completion rate continues to decline and the number of message retransmissions continues to increase; The number of consecutive cycles and the intensity of change corresponding to the statistical inspection record unit are used to form abnormal change sections, and the inspection links corresponding to the abnormal change sections are marked as uncertain sections.

5. The AI-powered intelligent companion-style touch-screen emergency voice inspection system according to claim 1, characterized in that, The process of dividing communication impact zones into different levels based on changes in network access status is as follows: Extract the network access status information corresponding to each inspection record unit within the uncertain section; Statistics were collected on the number of times the NB-IoT terminal changed its dwell status, the number of cell handovers, the signal quality level, and the number of connection interruptions within the corresponding time period. The communication impact index is calculated based on the degree of fluctuation in network access status. The communication impact zone is divided into high-level, medium-level, and low-level communication impact zones based on the magnitude of the communication impact index.

6. The AI-powered intelligent companion-style touch-screen emergency voice inspection system for single-person living as described in claim 1, characterized in that, The process of recording the effective response results under different scheduling methods is as follows: For high-level communication impact areas, prioritize triggering touch confirmation inspection tasks and increase the priority of sending corresponding messages; For areas with medium-level communication impact, a method of alternating touch confirmation and voice inspection is adopted; For areas affected by low-level communication, maintain the voice inspection mode and perform inspection tasks according to the original inspection cycle. The touch response results, voice response results, and response success rates generated under each scheduling method are statistically analyzed to form the corresponding effective response results.

7. The AI-powered intelligent companion-style touch-screen emergency voice inspection system according to claim 1, characterized in that, The process of establishing the correspondence between the communication influence area and the effective response results is as follows: Based on the effective response results, the response success rate, response latency, and response integrity rate in different levels of communication impact areas were statistically analyzed. Calculate the effective response contribution value corresponding to each communication impact area, and construct an impact mapping matrix based on the correlation between communication impact level and effective response contribution value; The influence mapping matrix is ​​used to establish the correspondence between the communication influence area and the effective response results, and to form the basic data for response credibility analysis.

8. The AI-powered intelligent companion-style touch-screen emergency voice inspection system according to claim 1, characterized in that, The process of correlation analysis between low-confidence response regions and continuously missing response regions is as follows: Call upon the basic data for response reliability analysis to extract the response success rate, response integrity rate, and response latency corresponding to each communication impact zone; Response credibility score is calculated based on response success rate, response integrity rate, and response latency. The communication impact area with a credibility score lower than the preset credibility threshold is marked as a low-credibility response area; Areas that do not receive effective response results within multiple consecutive inspection cycles are identified as continuously missing response areas. The degree of overlap and temporal correlation between low-confidence response regions and continuously missing response regions are analyzed to form the correlation results of disconnection features.

9. The AI-powered intelligent companion-style touch-screen emergency voice inspection system for single-person living as described in claim 1, characterized in that, The process of forming candidate results for disability assessment is as follows: Call the loss of contact feature association results to extract communication status change features, response behavior change features and inspection history records in the corresponding area; Determine whether there is continuous signal attenuation, frequent reconnection, or network access interruption based on the characteristics of communication status changes. Based on the characteristics of changes in response behavior, determine whether there are missing touch responses, missing voice responses, or a continuous increase in response latency; When communication anomalies are significant and historical behavior patterns remain stable, the state of being out of contact is determined to be caused by communication anomalies. When the communication status remains normal but the response behavior is continuously missing, it is determined to be a disconnection state caused by potential disability, and a corresponding disability determination candidate result is generated.

10. The AI-powered intelligent companion-style touch-screen emergency voice inspection system according to claim 1, characterized in that, The process of outputting the potential disability status identification results corresponding to the current inspection cycle is as follows: Acquire new communication status data, effective response results and candidate results for disability determination in subsequent inspection cycles, and statistically analyze the trend of response reliability change and the frequency of disability determination in each communication impact area. Adjust the frequency of inspection tasks, the priority of touch confirmation, and the interval of voice inspection in the corresponding areas according to the changing trends. The communication impact zone delineation process was re-executed using the revised inspection task scheduling strategy, and the communication impact zone level was updated. By combining the updated communication impact zone level, response credibility score, and candidate results for disability determination, a comprehensive analysis is performed to output the identification results of potential disability status corresponding to the current inspection cycle, and the identification results are sent to the monitoring terminal.

Images