A television machine internet of things application intelligent control system for the elderly

Abstract

The application provides a television internet-of-things application intelligent control system for the elderly, and relates to the technical fields of wisdom-based care for the elderly, smart home, internet of things and smart television, and comprises a multi-modal sensing module, a television end, an edge internet-of-things gateway, a cloud platform and a mobile end; the edge internet-of-things gateway is integrated and arranged in the television end or is independently arranged and is in communication connection with the television end; the edge internet-of-things gateway is in communication connection with the multi-modal sensing module; the television end is in interaction with the edge internet-of-things gateway and the cloud platform respectively; and the mobile end establishes communication with the cloud platform; the application takes the television end as an old-adapted interaction and home internet-of-things control center, and through old-adapted interaction, device control and a low-power consumption monitoring module, the television can monitor multi-modal events and wake up in a high-risk state in the process of starting or standby, so that the operation difficulty of the elderly is solved, and continuous monitoring and energy saving are considered.

Description

Technical Field

[0001] This invention relates to the fields of smart elderly care, smart home, Internet of Things and smart TV technology, and in particular to a smart control system for TV IoT applications serving the elderly. Background Technology

[0002] As my country's population ages, the needs of seniors for device operation, safety monitoring, health reminders, and family communication in their daily lives are becoming increasingly prominent. Existing age-friendly smart TVs mostly only offer large fonts, large icons, voice control, and video calling functions, primarily serving as audio-visual entertainment and not acting as a central control hub for the home's Internet of Things (IoT). Most existing smart home systems rely on mobile apps or smart speakers for operation, which presents high learning costs, complex operation levels, and difficulties for seniors to master.

[0003] Meanwhile, most existing elderly monitoring systems rely on cameras or wearable devices for monitoring. On the one hand, cameras raise serious privacy concerns; on the other hand, wearable devices suffer from low compliance, with elderly people frequently forgetting to wear them. Furthermore, traditional monitoring systems suffer from significant false alarms and missed alarms. In addition, existing alarm systems are mostly single-point alarms, lacking a hierarchical closed-loop alarm mechanism that includes local confirmation, family notification, and community linkage. Traditional cloud platform monitoring solutions are highly dependent on external networks; the entire monitoring system fails when the network is interrupted. Moreover, the large amount of raw video and audio data uploaded to the cloud poses a significant risk of privacy leaks. Therefore, this invention proposes a smart control system for television IoT applications serving the elderly to address the above problems. Summary of the Invention

[0004] To address the aforementioned issues, this invention proposes a smart control system for television IoT applications designed for the elderly. The television serves as the central hub for elderly-friendly interaction and home IoT control. Through elderly-friendly interaction, device control, and low-power monitoring modules, the television can monitor multimodal events both when powered on and in standby mode, and be activated in high-risk situations. This solves the problem of difficult operation for the elderly while simultaneously ensuring continuous monitoring and energy efficiency.

[0005] To achieve the objectives of this invention, the invention is implemented through the following technical solutions: A smart control system for a TV IoT application for elderly services includes a multimodal sensing module, a TV terminal, an edge IoT gateway, a cloud platform, and a mobile terminal. The edge IoT gateway is integrated and deployed inside the TV terminal or deployed independently and communicates with the TV terminal. The edge IoT gateway communicates with the multimodal sensing module. The TV terminal interacts with the edge IoT gateway and the cloud platform respectively. The mobile terminal establishes communication with the cloud platform. The multimodal sensing module includes millimeter-wave radar, camera, microphone array, environmental sensors, and health device interface; The television end comprises an age-friendly interaction module, a device control module, a low-power monitoring module, an alarm interaction module, a remote assistance module and a multi-channel alarm local confirmation sub-module. When the television end is in a power-on or standby state, it receives events output by the multi-modal perception module through the low-power monitoring module, and wakes up the screen and the television main control module when a wake-up condition is satisfied; The edge IoT gateway comprises a protocol adaptation module, a local rule engine, an edge AI reasoning module, a local storage module, a network disconnection autonomy module, and a multi-modal perception conflict arbitration and dynamic confidence fusion module; The cloud platform comprises an elderly person portrait module, an alarm grading module, an elderly person portrait baseline adaptive updating and disturbance identification sub-module, a grading authorization module, and an alarm linkage automatic family review sub-module; The mobile end is used by relatives, community staff and medical care personnel, for receiving graded alarm messages, viewing desensitized event records, initiating video calls, and performing remote assistance operations after being authorized by the elderly person.

[0006] The further improvement lies in that: the age-friendly interaction module is configured to provide an age-friendly interface with large fonts and large icons, and support voice interaction; the device control module is configured to receive instructions from the edge IoT gateway to realize linkage control of smart home devices; the low-power monitoring module includes a television main control unit and a low-power co-processing unit; the alarm interaction module is configured to receive alarm events issued by the edge IoT gateway, and execute television pop-up windows and voice broadcast of alarm prompts; the remote assistance module is configured to receive remote assistance instructions forwarded by the cloud platform, and perform remote interface operations; the multi-channel alarm local confirmation sub-module is configured to execute multi-mode alarm confirmation logic.

[0007] The further improvement lies in that: in the low-power monitoring module, when the television is in a standby state, the screen and the television main control unit are turned off, and the low-power co-processing unit continuously monitors various events from the multi-modal perception module; when a preset high-risk event is identified, the television main control unit and the screen are woken up; and the low-power monitoring module supports time-segmented sampling control, and automatically controls the environmental sensors of the multi-modal perception module to reduce the sampling frequency during the nighttime sleep period.

[0008] The further improvement lies in that: when the multi-channel alarm local confirmation sub-module triggers a local inquiry for an alarm, it first executes a television pop-up window and a high-volume voice inquiry. If no remote control confirmation instruction is received after the first timeout, a millimeter-wave radar is used to recognize the elderly person's hand-raising and head-swinging postures as a contactless confirmation or denial response signal. If there is still no response after the second timeout, it reports a confirmation timeout to the cloud platform and starts an alarm escalation process.

[0009] Further improvements include: the protocol adaptation module supports infrared, Zigbee, Bluetooth Mesh, WiFi, Matter, and RS485 protocols, and also supports learning the infrared codes of traditional old home appliances; the local storage module stores the raw data collected by the multimodal sensing module locally at the edge, only outputting desensitized event features externally; the network disconnection autonomy module is configured to independently complete event judgment, alarm prompts, and device linkage locally when the network is disconnected; the edge AI inference module is used to extract and fuse features from the multimodal sensing data, outputting events such as human presence, falls, prolonged sitting, getting out of bed, abnormal sounds, and failure to take medication; the local rule engine receives the event features output by the edge AI inference module and the elderly individual behavior baseline issued by the cloud platform, compares the real-time event features with the locally stored elderly individual behavior baseline, determines whether the event is a normal event, a concern event, or an abnormal event; and sends local alarm commands to the TV terminal according to preset rules to trigger smart home linkage control.

[0010] Further improvements include: the elderly profile module constructs an individual behavior baseline based on the elderly's basic attributes, sensory abilities, chronic disease information, daily routines, device usage habits, voice preferences, and historical behavior data, and sends it to the edge IoT gateway; the alarm classification module executes hierarchical alarms through local prompts, family notifications, and community or emergency contact notifications, and automatically escalates the alarm upon local confirmation of timeout.

[0011] Further improvements are made in the following ways: The multimodal perception conflict arbitration and dynamic confidence fusion module is configured to: assign dynamic confidence weights to each perception output event; automatically reduce the confidence weight when there is echo attenuation or obstruction from the millimeter-wave radar; automatically reduce the confidence weight when the camera image is dimly lit or obstructed; adjust the confidence weight of the distress call event according to the signal-to-noise ratio of the microphone array; mark the event as an event of interest when the judgment results of multiple sensor events conflict, and trigger local verification on the TV; only when the total confidence of the multi-sensor superposition exceeds a preset threshold is it judged as an abnormal event, and non-human target filtering is performed by combining radar micro-Doppler features and camera human detection to filter out false triggers caused by pet activities and falling objects.

[0012] Further improvements are made in the following: The elderly portrait baseline adaptive update and disturbance recognition submodule is used to divide the elderly individual behavior baseline into long-term baseline and short-term disturbance label. When multiple visitors are detected, the elderly leave home, or short-term abnormal physical events are detected, a disturbance label is applied, part of the threshold comparison logic is temporarily frozen, and the baseline comparison is restored after the disturbance ends. Then, a sliding time window is used to realize the gradual iterative update of the baseline.

[0013] Further improvements are made in that: the hierarchical authorization module is used to divide remote assistance permissions into multiple levels: Level 1 is only for receiving alarms and viewing de-identified event records, without device operation permissions; Level 2 is for initiating video calls, but not for controlling TVs and smart home devices; Level 3 is for temporary authorization, allowing remote operation of the TV interface and control of home appliances, with an automatic expiration time set for the authorization; Level 4 is for time-limited temporary authorization for community maintenance personnel.

[0014] A further improvement is that the alarm linkage automatic family verification submodule is configured to: when receiving a concern event or a primary abnormal event from the edge IoT gateway, and with the authorization of the elderly, automatically initiate a family verification call to the relative's mobile terminal. During the process, the de-identified event abbreviated feature frame is pushed first, and the relatives can choose to initiate a two-way video call. Then, the time confirmation mark is performed, and finally the mark result is sent back to the edge IoT gateway to terminate the alarm or continue the alarm escalation process.

[0015] The beneficial effects of this invention are as follows: This invention uses the television set as the central hub for elderly-friendly interaction and home IoT control. Through elderly-friendly interaction, device control, and low-power monitoring modules, the television can monitor multimodal events when it is powered on or in standby mode and wake up in case of high risk, solving the problem of difficult operation for the elderly while taking into account continuous monitoring and energy saving.

[0016] By integrating multimodal conflict arbitration with dynamic confidence, the weights of radar, camera, and microphone are dynamically adjusted according to their status. In case of conflict, the data is transferred to the local TV for review. Anomalies are only judged when the total confidence exceeds the threshold. Non-human targets such as pets and falling objects are filtered out to reduce false alarms and missed alarms.

[0017] By using multi-channel alarm local confirmation, the system first uses a pop-up window on the TV and voice inquiry. If the timeout is exceeded, millimeter-wave radar is used to identify raising a hand or moving a head as a contactless response. If the timeout is exceeded again, the alarm is upgraded, forming a local confirmation closed loop. This solves the problems of traditional single-point alarms lacking local confirmation and being inconvenient for the elderly to operate.

[0018] By outputting only de-identified event characteristics through edge local storage, the network disconnection self-governance module can complete event judgment, alarm prompts and device linkage locally when the network is disconnected, reducing the risk of privacy leakage when the original audio and video are uploaded to the cloud, and avoiding monitoring failure due to network disconnection.

[0019] By constructing individual behavioral baselines based on elderly profiles and sending them to the edge, a local rule engine compares and determines the event level; combined with disturbance labels, threshold freezing, and sliding window updates, it adapts to individual differences among the elderly and short-term life changes, reducing misjudgments.

[0020] Through tiered alarms, tiered authorization, and family verification, the system achieves a progressive escalation from local alerts to family notifications and community collaboration. With the elderly's authorization, relatives can be called, de-identified feature frames can be pushed, and video confirmation and result transmission are supported, forming a closed loop of "local confirmation - family verification - community collaboration" that balances privacy, permissions, and processing efficiency. Attached Figure Description

[0021] Figure 1 This is a schematic diagram of the system framework of the present invention. Detailed Implementation

[0022] To enhance understanding of the present invention, the present invention will be further described in detail below with reference to embodiments. These embodiments are only used to explain the present invention and do not constitute a limitation on the scope of protection of the present invention.

[0023] according to Figure 1 As shown, this embodiment proposes a smart control system for IoT applications in televisions to serve the elderly, including a multimodal sensing module, a television terminal, an edge IoT gateway, a cloud platform, and a mobile terminal. The edge IoT gateway is integrated and deployed inside the television terminal or deployed independently and communicates with the television terminal. The edge IoT gateway communicates with the multimodal sensing module. The television terminal interacts with the edge IoT gateway and the cloud platform respectively, and the mobile terminal establishes communication with the cloud platform.

[0024] The multimodal perception module includes millimeter-wave radar, a camera, a microphone array, environmental sensors, and a health device interface. Specifically: the millimeter-wave radar detects human presence, posture, falls, prolonged sitting, getting out of bed, and respiratory and body movement trends; the camera extracts information about human skeletal points, posture, and activity areas; the microphone array detects far-field speech, abnormal sounds, and calls for help; environmental sensors include door sensors, smoke detectors, gas detectors, water immersion sensors, light sensors, and temperature and humidity sensors; and the health device interface connects to peripheral health devices such as blood pressure monitors, blood glucose meters, smart pillboxes, and smart mattresses.

[0025] The TV terminal includes an elderly-friendly interaction module, a device control module, a low-power monitoring module, an alarm interaction module, a remote assistance module, and a multi-channel alarm local confirmation sub-module. When the TV is powered on or in standby mode, it receives events output by the multimodal sensing module through the low-power monitoring module and wakes up the screen and TV main control module when wake-up conditions are met. Specifically: the elderly-friendly interaction module provides a user interface with large fonts and icons suitable for the elderly and supports voice interaction; the device control module receives commands from the edge IoT gateway to achieve smart home device linkage control; the low-power monitoring module includes the TV main control unit and a low-power coprocessing unit. When the TV is in standby mode, the screen and TV main control unit are turned off, and the low-power coprocessing unit continuously monitors various events from the multimodal sensing module. The monitoring period can be set to poll the event queue once every 100ms. When a preset high-risk event is detected (wake-up conditions include: a fall confidence score greater than 0.85, detection of SOS keywords such as "Help / Help!", or being out of bed for more than a preset time), the TV is woken up. The system includes a main control unit and a screen; a low-power monitoring module that supports time-segmented sampling control, automatically controlling the environmental sensors of the multimodal perception module to reduce sampling frequency during nighttime sleep (radar at 10Hz, camera off or reduced to 1fps during nighttime sleep, environmental sensor sampling interval adjusted from 1 minute to 5 minutes; if the radar detects leaving the bed, the camera and microphone sampling frequency is temporarily increased); an alarm interaction module to receive alarm events from the edge IoT gateway and execute TV pop-ups and voice broadcast alarm prompts; a remote assistance module to receive remote assistance instructions forwarded by the cloud platform and execute remote interface operations; and a multi-channel alarm local confirmation submodule to execute multi-mode alarm confirmation logic. When an alarm triggers a local query, it first executes a TV pop-up and a loud voice query. If no confirmation instruction is received from the remote control after the first timeout, it activates millimeter-wave radar to recognize the elderly person's hand raising and head shaking posture as a contactless confirmation or denial response signal. If there is still no response after the second timeout, it reports a confirmation timeout to the cloud platform and initiates the alarm escalation process. By using the television as the central hub for elderly-friendly interaction and home IoT control, and through an elderly-friendly interaction module, a device control module, and a low-power monitoring module, the television can receive events output by the multimodal sensing module in both power-on and standby states. When the wake-up conditions are met, the screen and the main control module of the television are awakened. This solves the problem that existing elderly-friendly televisions are mainly for audio-visual entertainment and that the elderly have difficulty operating them by relying on mobile apps or smart speakers. At the same time, it achieves both continuous monitoring in standby mode and energy saving.

[0026] The edge IoT gateway includes a protocol adaptation module, a local rules engine, an edge AI inference module, a local storage module, a network outage autonomy module, and a multimodal perception conflict arbitration and dynamic confidence fusion module. Specifically: the protocol adaptation module supports infrared, Zigbee, Bluetooth Mesh, WiFi, Matter, and RS485 protocols, and can learn the infrared codes of traditional home appliances. The local storage module stores the raw data collected by the multimodal perception module locally at the edge, only outputting anonymized event features. The network outage autonomy module is configured to independently handle event determination, alarm notifications, and device linkage when the network is disconnected. The edge AI inference module extracts and fuses features from the multimodal perception data, outputting events such as human presence, falls, prolonged sitting, getting out of bed, abnormal sounds, and failure to take medication. The local rule engine receives event features output by the edge AI inference module and baselines of individual elderly behavior sent by the cloud platform. It compares the real-time event features with the locally stored baselines of individual elderly behavior to determine whether the event is a normal event, a concern event, or an abnormal event. Based on preset rules, it sends local alarm commands to the TV to trigger smart home linkage control.

[0027] The multimodal perception conflict arbitration and dynamic confidence fusion module is configured to: assign dynamic confidence weights to each perception output event; automatically reduce the confidence weight when there is echo attenuation or obstruction from the millimeter-wave radar; automatically reduce the confidence weight when the camera image is dimly lit or obstructed; adjust the confidence weight of the distress call event according to the signal-to-noise ratio of the microphone array; mark the event as an event of interest when the judgment results of multiple sensor events conflict, and trigger local verification on the TV; only when the total confidence of the multi-sensor superposition exceeds a preset threshold is it judged as an abnormal event, and non-human target filtering is performed by combining radar micro-Doppler features and camera human detection to filter out false triggers caused by pet activities and falling objects.

[0028] The formula for the total confidence score of the multimodal perception conflict arbitration and dynamic confidence fusion module is as follows: In the formula: C total The total confidence level after multi-path sensing fusion is denoted as n; the number of sensing sources participating in the fusion is denoted as c. i Let c be the confidence score of the i-th sensing event, and 0 ≤ c i ≤1; w i Let w be the dynamic confidence weight of the i-th sensing source, and 0 ≤ w i ≤1.

[0029] When a sensing source experiences echo attenuation, obstruction, dim light, low signal-to-noise ratio, or malfunction, the corresponding wi is automatically reduced. The judgment rule can be expressed as: C total A value greater than 0.8 is considered an abnormal event. 0.5≤Ctotal ≤0.8, is considered an event of concern; C total <0.5, is considered a normal time.

[0030] When event judgments from multiple sensors conflict, the final judgment is not made directly using the established rules. Instead, the event is marked as an event of interest, triggering local verification on the television. Dynamic confidence weights are assigned to events from multiple sensors, including millimeter-wave radar, cameras, and microphone arrays. These weights are automatically adjusted when radar echo attenuation or obstruction occurs, cameras are in low light or obstructed, or microphone signal-to-noise ratios change. When multiple sensor judgments conflict, the event is marked as an event of interest, triggering local verification on the television. An abnormal event is only determined when the total confidence score of the multiple sensors exceeds a preset threshold. Furthermore, radar micro-Doppler characteristics and camera human detection are used to filter out non-human targets such as pets and fallen objects. This improves the accuracy of event judgment and reduces false alarms and false negatives.

[0031] The cloud platform includes an elderly profile module, an alarm classification module, an elderly profile baseline adaptive update and disturbance recognition sub-module, a hierarchical authorization module, and an alarm linkage automatic family verification sub-module. It also includes a remote assistance service module, a device management module, and an OTA module (the OTA module enables the cloud platform to perform differential upgrades on the firmware of TVs, edge gateways, and sensors, with upgrade package signature verification, support for breakpoint resumption and failure rollback. The upgrade window can be set to off-peak hours at night, with gray-scale release initially covering 1% of devices and then gradually expanding). Specifically, the elderly profile module constructs an individual behavioral baseline based on the elderly person's basic attributes, sensory abilities, chronic disease information, daily routines, device usage habits, voice preferences, and historical behavioral data (e.g., baseline: wake-up time 6:30-7:30, afternoon nap 12:30-14:00, 1-2 nighttime toilet visits, no more than 90 minutes of continuous sitting, and triggering attention if the elderly person is out of bed for more than 30 minutes). This baseline is then distributed to the edge IoT gateway. The alarm classification module executes hierarchical alarms through local notifications, family notifications, and community or emergency contact notifications, and automatically escalates the alarm upon local confirmation of timeout. The elderly profile baseline adaptive update and disturbance recognition submodule divides the individual behavior baseline of the elderly into a long-term baseline and short-term disturbance labels. When multiple visitors are detected, the elderly leave home, or short-term abnormal physical events are identified, a disturbance label is applied, temporarily freezing part of the threshold comparison logic. The baseline comparison is restored after the disturbance ends, and then a sliding time window is used to achieve progressive iterative updates of the baseline (e.g., 7-day short-term window, 30-day medium-term window, 90-day long-term window, with weights that can be set to 0.5, 0.3, and 0.2). The hierarchical authorization module divides remote assistance permissions into multiple levels: Level 1 is only for receiving alarms and viewing de-identified event records, with no device operation permissions; Level 2 allows initiating video calls, but cannot control the TV and smart home devices; Level 3 is temporary authorization, allowing remote operation of the TV interface and control of home appliances, with an automatic expiration time set; Level 4 is a time-limited temporary authorization for community maintenance personnel. The alarm linkage automatic family verification submodule is configured to: when it receives a concern event or a primary abnormal event from the edge IoT gateway, and with the elderly person's authorization, automatically initiate a family verification call to the relative's mobile device. During the process, it prioritizes pushing de-identified event thumbnail feature frames and supports relatives to initiate two-way video calls. Then, it performs time confirmation marking and finally sends the marking result back to the edge IoT gateway to terminate the alarm or continue the alarm escalation process.

[0032] The mobile app is for use by relatives, community workers, and medical staff to receive tiered alarm messages, view de-identified event records, initiate video calls, and perform remote assistance operations after authorization from the elderly.

[0033] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A smart control system for a television set using the Internet of Things (IoT) to serve the elderly, characterized in that: comprising a multi-modal sensing module, a television end, an edge IoT gateway, a cloud platform, and a mobile end. The edge IoT gateway is integratedly deployed inside the television end or independently deployed and communicatively connected to the television end, the edge IoT gateway is communicatively connected to the multi-modal sensing module, the television end interacts with the edge IoT gateway and the cloud platform respectively, and the mobile end establishes communication with the cloud platform; the multi-modal sensing module comprises a millimeter-wave radar, a camera, a microphone array, an environmental sensor, and a health device interface; the television end comprises an elderly-friendly interaction module, a device control module, a low-power listening module, an alarm interaction module, a remote assistance module, and a multi-channel alarm local confirmation submodule. When the television end is in a power-on or standby state, it receives events output by the multi-modal sensing module through the low-power listening module, and wakes up the screen and the television main control module when a wake-up condition is met; the edge IoT gateway comprises a protocol adaptation module, a local rule engine, an edge AI inference module, a local storage module, a network disconnection autonomy module, and a multi-modal sensing conflict arbitration and dynamic confidence fusion module; the cloud platform comprises an elderly person profile module, an alarm grading module, an elderly person profile baseline adaptive update and disturbance recognition submodule, a hierarchical authorization module, and an alarm linkage automatic family verification submodule; the mobile end is used by relatives, community staff, and medical care personnel, for receiving graded alarm messages, viewing desensitized event records, initiating video calls, and performing remote assistance operations after being authorized by the elderly person.

2. The intelligent control system for a television set IoT application serving the elderly according to claim 1, characterized in that: the elderly-friendly interaction module is configured to provide an elderly-friendly interface with large fonts and large icons, and support voice interaction; the device control module is configured to receive instructions from the edge IoT gateway to realize linked control of smart home devices; the low-power listening module comprises a television main control unit and a low-power coprocessing unit; the alarm interaction module is configured to receive alarm events issued by the edge IoT gateway, and execute television pop-up windows and voice broadcast of alarm prompts; the remote assistance module is configured to receive remote assistance instructions forwarded by the cloud platform, and perform remote interface operations; the multi-channel alarm local confirmation submodule is configured to execute multi-mode alarm confirmation logic.

3. The intelligent control system for a television set IoT application serving the elderly according to claim 2, characterized in that: in the low-power listening module, when the television is in a standby state, the screen and the television main control unit are turned off, and the low-power coprocessing unit continuously listens to various events from the multi-modal sensing module; when a preset high-risk event is identified, the television main control unit and the screen are woken up; and the low-power listening module supports time-segmented sampling control, which automatically controls the environmental sensor of the multi-modal sensing module to reduce the sampling frequency during the nighttime sleep period.

4. The intelligent control system for a television set IoT application serving the elderly according to claim 2, characterized in that: when an alarm triggers a local inquiry, the multi-channel alarm local confirmation submodule first executes a television pop-up window and a high-volume voice inquiry. If no remote control confirmation instruction is received after the first timeout, the millimeter-wave radar is enabled to recognize the postures of the elderly person raising their hand and swinging their head, which are used as contactless confirmation or denial response signals. If there is still no response after the second timeout, a confirmation timeout is reported to the cloud platform, and an alarm escalation process is started.

5. The intelligent control system for a television set IoT application serving the elderly according to claim 1, characterized in that: The protocol adaptation module supports infrared, Zigbee, Bluetooth Mesh, WiFi, Matter, and RS485 protocols, and also supports learning the infrared codes of traditional old home appliances; the local storage module is used to save the raw data collected by the multimodal sensing module locally at the edge, and only outputs desensitized event features externally; the network disconnection autonomy module is configured to independently complete event judgment, alarm prompts, and device linkage locally when the network is disconnected; the edge AI inference module is used to extract and fuse features from the multimodal sensing data, and output events such as human presence, falls, prolonged sitting, getting out of bed, abnormal sounds, and failure to take medication; The local rule engine is used to receive event features output by the edge AI inference module and the individual behavior baseline of the elderly issued by the cloud platform. It compares the real-time event features with the locally stored individual behavior baseline of the elderly to determine whether the event is a normal event, a noteworthy event, or an abnormal event. It also sends local alarm commands to the TV according to preset rules, triggering smart home linkage control.

6. The intelligent control system for a television set IoT application serving the elderly according to claim 1, characterized in that: The elderly profile module constructs an individual behavior baseline based on the elderly's basic attributes, sensory abilities, chronic disease information, daily routines, device usage habits, voice preferences, and historical behavior data, and sends it to the edge IoT gateway; the alarm classification module executes hierarchical alarms through local prompts, family notifications, and community or emergency contact notifications, and automatically escalates the alarm upon local confirmation of timeout.

7. The intelligent control system for a television set IoT application serving the elderly according to claim 1, characterized in that: The multimodal perception conflict arbitration and dynamic confidence fusion module is configured to: assign dynamic confidence weights to each perception output event; automatically reduce the confidence weight when there is echo attenuation or obstruction from the millimeter-wave radar; automatically reduce the confidence weight when the camera image is dimly lit or obstructed; adjust the confidence weight of the distress call event according to the signal-to-noise ratio of the microphone array; mark the event as an event of interest when the judgment results of multiple sensor events conflict, and trigger local verification on the TV; only when the total confidence of the multi-sensor superposition exceeds a preset threshold is it determined to be an abnormal event, and non-human target filtering is performed by combining radar micro-Doppler features and camera human detection to filter out false triggers caused by pet activities and falling objects.

8. The intelligent control system for a television set IoT application serving the elderly according to claim 6, characterized in that: The elderly profile baseline adaptive update and disturbance recognition submodule is used to divide the elderly individual behavior baseline into long-term baseline and short-term disturbance label. When multiple visitors are detected, the elderly leave home, or short-term abnormal physical events are detected, a disturbance label is applied, part of the threshold comparison logic is temporarily frozen, and the baseline comparison is restored after the disturbance ends. Then, a sliding time window is used to realize the gradual iterative update of the baseline.

9. The intelligent control system for a television set IoT application serving the elderly according to claim 1, characterized in that: The hierarchical authorization module is used to divide remote assistance permissions into multiple levels: Level 1 is only for receiving alarms and viewing de-identified event records, without device operation permissions; Level 2 allows initiating video calls but does not allow control of the TV or smart home devices; Level 3 is a temporary authorization, which allows remote operation of the TV interface and control of home appliances. The authorization can be set to have an automatic expiration time. Level 4 is a time-limited temporary authorization for community operations and maintenance personnel.

10. The intelligent control system for a television set IoT application serving the elderly according to claim 1, characterized in that: The alarm linkage automatic family verification submodule is configured to: when it receives a concern event or a primary abnormal event from the edge IoT gateway, and with the authorization of the elderly, automatically initiate a family verification call to the mobile terminal of the relatives. During the process, it prioritizes pushing the de-identified event thumbnail feature frame, and supports relatives to initiate two-way video calls. Then, it performs time confirmation marking, and finally sends the marking result back to the edge IoT gateway to terminate the alarm or continue the alarm escalation process.

Images