Emergency pager automatic alarm system based on multi-mode perception

Abstract

The invention relates to the technical field of medical first-aid equipment, and discloses a first-aid pager automatic alarm system based on multi-modal perception, the system is internally provided with eight modules, a multi-modal perception module collects user state data online, a special group data collection module pays attention to information of three types of crowds of deaf-mute, blind and cognitive impairment, and a first-aid pager alarm module is provided with a first-aid alarm module and a second-aid alarm module. The data analysis module calculates a session key, a permission compliance rate, a fault influence degree and a success rate of simplified operation of a cognitive impairment, the sensitive information encryption protection module performs end-to-end encryption on collected data, the fine-grained permission management and control module performs dynamic access control judgment, the distributed fault self-healing module performs self-healing strategy judgment, and the distributed fault self-healing module performs self-healing strategy judgment. The special crowd interaction optimization module performs interaction strategy evaluation, and the instruction execution module receives encryption, authority, self-healing and interaction optimization instructions issued by all the modules, uniformly controls underlying hardware and network resources to complete actions after priority arbitration, and returns an execution result log to form a closed loop.

Description

Technical Field

[0001] This invention relates to the field of medical emergency equipment technology, specifically to an automatic alarm system for emergency call devices based on multimodal perception. Background Technology

[0002] Currently, traditional emergency call systems face multiple technical bottlenecks in practical applications: First, their perception dimensions are limited, relying heavily on manual alarm triggering by users (such as pressing a call button). This makes it difficult to automatically identify non-active alarm scenarios such as falls, sudden physiological abnormalities (such as sudden changes in heart rate caused by myocardial infarction), and painful groans, easily delaying emergency response. Second, they have poor adaptability to special populations. The accuracy of sign language recognition for the deaf and mute is insufficient; the voice interaction logic for the blind is complex (requiring multi-step command confirmation); and the user interface for those with cognitive impairments is cumbersome, making it difficult to complete alarm calls, resulting in low success rates for emergency calls for these three groups. Third, there are significant security risks associated with sensitive information. Most systems use static keys to encrypt patient audio, video, and physiological data. With the development of quantum computing technology, static keys are easily cracked, posing a risk of leaking patient identity and medical details. Fourth, access control is coarse-grained. There is a lack of hierarchical authorization mechanisms when medical staff access historical data, posing a risk of unauthorized personnel (such as interns and administrative staff) unauthorized access to sensitive data. Fifth, the architecture lacks reliability. In a centralized architecture, interruptions to the main server or communication links can easily lead to system-wide paralysis, making it difficult to guarantee the continuity of emergency services. To address the aforementioned pain points, there is an urgent need to develop an automatic alarm system for emergency call devices that integrates multimodal perception, adaptation to special populations, quantum security protection, fine-grained access control, and distributed self-healing capabilities, in order to improve emergency response efficiency, data security, and service adaptability for special populations. Summary of the Invention

[0003] (a) Technical problems to be solved To address the shortcomings of existing technologies, this invention provides an automatic alarm system for emergency call devices based on multimodal perception. It features multimodal intelligent perception, precise adaptation to special populations, quantum-level data encryption, dynamic fine-grained access control, and distributed disaster recovery and self-healing capabilities, thus solving the problems of existing technologies in terms of data security, access control, system reliability, and inclusion of special populations.

[0004] (II) Technical Solution To achieve the above objectives, the present invention provides the following technical solution: an automatic alarm system for emergency call devices based on multimodal perception, comprising a multimodal perception module, a special group data acquisition module, a data analysis module, a sensitive information encryption and protection module, a fine-grained access control module, a distributed fault self-healing module, a special group interaction optimization module, and an instruction execution module; The multimodal perception module collects user status data online through three types of sensors: video, audio, and physiological sensors, and generates a multimodal feature vector with a unified timestamp, which is then output to the data analysis module. The special group data collection module targets three groups: deaf-mute, blind, and cognitively impaired individuals. It collects sign language video streams, low-volume voice commands, and one-click trigger signals, converts them into standardized intent tags, and outputs them to the data analysis module. The data analysis module receives output data from the multimodal perception module and the special group data acquisition module, and calculates the session key. Permission compliance rate Fault impact Simplified operation success rate for people with cognitive impairment The calculation results are sequentially output to the sensitive information encryption protection module, the fine-grained access control module, the distributed fault self-healing module, and the special population interaction optimization module; The sensitive information encryption protection module receives the session key output by the data analysis module. End-to-end encryption is performed on video, audio, raw physiological data, and intermediate analysis results; The fine-grained permission control module is based on the permission compliance rate. Perform dynamic access control judgments; The distributed fault self-healing module is based on the fault impact. Determine self-healing strategies; The special population interaction optimization module simplifies operations based on the success rate of individuals with cognitive impairment. Conduct interaction strategy assessments and remotely push assistance prompts to caregivers; The instruction execution module receives encryption, permission, self-healing, and interaction optimization instructions from the above modules, arbitrates them according to priority, and uniformly controls the underlying hardware and network resources to complete the actions, and sends back the execution result log to form a closed loop.

[0005] Preferably, the multimodal perception module includes a video perception unit, an audio perception unit, and a physiological sensing unit; the data analysis module includes a sensitive information leakage risk analysis unit, a fine-grained permission compliance analysis unit, a key component failure risk analysis unit, and a special population needs adaptation analysis unit.

[0006] Preferably, the video sensing unit uses a wide-angle color-depth dual camera to monitor user posture, fall events, facial pain expressions, and gestures for help online; the audio sensing unit uses a microphone array, combined with sound event detection and speech recognition algorithms, to monitor ambient sound and user voice online; and the physiological sensing unit uses a wristband-type multi-parameter sensor to monitor heart rate, blood oxygen, skin conductance, and body temperature online.

[0007] Preferably, the special group data acquisition module includes a sign language acquisition unit for deaf-mute people, a voice interaction unit for blind people, and a simplified interaction unit for cognitive impairment. The deaf and mute sign language acquisition unit collects deaf and mute sign language information, including continuous hand gestures, hand shapes, movement trajectories, and facial expressions, through the real-time video stream provided by the video perception unit and by calling a dedicated sign language recognition model. The blind voice interaction unit captures speech through an audio perception unit and runs an optimized voice interaction process to collect blind voice interaction information centered on short preset commands. The simplified interaction unit for cognitive impairment collects the user's direct trigger signals through a large touch button or physical switch on the device, supplemented by eye-catching icons and light prompts.

[0008] Preferably, before analysis, the sensitive information leakage risk analysis unit encrypts the data using a post-quantum cryptography algorithm and calculates the session key. The calculation formula is as follows: ; In the formula, Indicates the session key. Indicates the current timestamp. This indicates the preset master seed. This represents a single random number generated by the device. This represents a post-quantum secure key derivation function conforming to NIST SP 800-56C.

[0009] Preferably, the fine-grained permission compliance analysis unit calculates the permission compliance rate. The calculation formula is as follows: ; In the formula, Indicates the permission compliance rate. This indicates the number of times healthcare professionals have been authorized to correctly access historical records. This indicates the number of unauthorized access attempts. This indicates the total number of access requests.

[0010] Preferably, the critical component failure risk analysis unit calculates the failure impact. The calculation formula is as follows: ; In the formula, Indicates the degree of impact of the fault. This represents the probability of failure of a centralized component. Indicates the backup switch success rate. This represents the service interruption duration coefficient.

[0011] Preferably, the special population needs adaptation analysis unit calculates the success rate of simplified operation for individuals with cognitive impairment. The calculation formula is as follows: ; In the formula, This indicates the success rate of simplified procedures for individuals with cognitive impairment. This indicates the number of times a person with cognitive impairment has successfully completed a simplified task. This indicates the total number of attempts made by individuals with cognitive impairment.

[0012] Preferably, the sensitive information encryption protection module is based on a session key. End-to-end post-quantum encryption is applied to raw video, audio, physiological data, and intermediate analysis results. The fine-grained permission control module is based on the permission compliance rate. Dynamic access control is performed, and the compliance rate of permissions is assessed. Maintain normal authorization policy at 95% compliance rate; when the permission compliance rate is... When the compliance rate is between 80% and 95%, initiate secondary identity verification; this will be implemented when the permission compliance rate is [not specified]. If the error rate is ≤80%, immediately downgrade to read-only mode and push an abnormal audit alert.

[0013] Preferably, the distributed fault self-healing module is based on the fault impact. Perform self-healing strategy assessment, and determine the impact of the fault. When the value is less than 0.3, the local edge node is enabled to continue providing services; when the failure impact is less than or equal to 0.3... When the fault impact is ≤0.7, a primary / standby hot switch is triggered and multi-hop relay communication is initiated; when .... When the value is greater than 0.7, it is determined to be a global fault, and offline caching mode is automatically enabled. Data is then retransmitted in batches after communication is restored. The special population interaction optimization module simplifies operations based on the success rate of individuals with cognitive impairment. Interaction strategy assessment was conducted to evaluate the success rate of simplified operations in individuals with cognitive impairment. When the success rate is ≥90%, maintain the existing interaction process; when the success rate for simplifying operations is ≤70% for individuals with cognitive impairments. When the success rate is below 90%, automatically enlarge the icon, reduce the voice prompt speed, and increase vibration confirmation; simplify the operation for cognitively impaired individuals to improve success rate. When the threshold is below 70%, switch to the minimalist mode of single button-single light-single vibration and remotely push auxiliary prompts to caregivers.

[0014] Compared with the prior art, the present invention provides an automatic alarm system for emergency call devices based on multimodal perception, which has the following beneficial effects: 1. This invention calculates the session key. It performs end-to-end encryption on video, audio, raw physiological data, and intermediate analysis results, achieving confidentiality protection throughout the entire lifecycle of data transmission, storage, and analysis. This helps resist the threat of quantum computing cracking static keys, ensuring that even if data packets are intercepted, sensitive information (such as patient identity and disease details) cannot be decrypted and restored.

[0015] 2. This invention calculates the permission compliance rate. Differentiated access control policies are implemented based on different compliance rate thresholds to achieve hierarchical authorization of medical staff and real-time interception of abnormal access. This achieves the beneficial effect of refining the granularity of permission management, strictly restricting unauthorized personnel (such as unapproved interns) from obtaining sensitive data, and ensuring the security of medical data access.

[0016] 3. This invention calculates the impact of the fault. By quantifying the combined impact of centralized component failure probability, backup switch success rate, and business interruption duration, a graded self-healing strategy is implemented based on different impact thresholds to achieve rapid response and business continuity assurance when critical components fail. This avoids the single point of failure risk of centralized architecture and prevents the system from falling into a global paralysis due to the interruption of the main server / communication link.

[0017] 4. This invention simplifies the operation success rate for individuals with cognitive impairment by calculating the success rate. The interaction strategy is dynamically adjusted based on different success rate thresholds to continuously optimize the interaction experience for special groups, thereby further reducing the difficulty of alarm operation for special groups and improving their emergency call success rate and user experience. Attached Figure Description

[0018] Figure 1 This is a system flowchart of the present invention. Detailed Implementation

[0019] 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.

[0020] Please see Figure 1 The emergency call button automatic alarm system based on multimodal perception includes a multimodal perception module, a special group data acquisition module, a data analysis module, a sensitive information encryption and protection module, a fine-grained access control module, a distributed fault self-healing module, a special group interaction optimization module, and an instruction execution module. The multimodal perception module collects user status data online through three types of sensors: video, audio, and physiological sensors, and generates multimodal feature vectors with a unified timestamp, which are then output to the data analysis module. The special group data collection module targets three groups: deaf-mute, blind, and cognitively impaired individuals. It collects sign language video streams, low-volume voice commands, and one-click trigger signals, and converts them into standardized intent labels, which are then output to the data analysis module. The data analysis module receives output data from the multimodal sensing module and the special group data acquisition module, and calculates the session key. Permission compliance rate Fault impact Simplified operation success rate for people with cognitive impairment ; The sensitive information encryption protection module receives the session key output by the data analysis module. It also performs end-to-end encryption on video, audio, raw physiological data and intermediate analysis results to ensure the confidentiality and resistance to quantum attacks of data throughout the entire lifecycle of transmission, storage and analysis, achieving the technical effect that even if the data packet is intercepted, it cannot be decrypted and restored. The fine-grained access control module is based on the access compliance rate. Perform dynamic access control judgments; The distributed fault self-healing module is based on the degree of fault impact. Determine self-healing strategies; The special population interaction optimization module simplifies operations and improves the success rate for people with cognitive impairments. Conduct interaction strategy assessments and remotely push assistance prompts to caregivers; The instruction execution module receives encryption, permission, self-healing, and interaction optimization instructions from the above modules, arbitrates them according to priority, and uniformly controls the underlying hardware and network resources to complete the actions, and sends back the execution result log to form a closed loop.

[0021] The advantages are: through multimodal perception fusion, adaptive collection for special groups, dynamic risk assessment and execution closed loop, it systematically solves the four core problems of existing technologies in data security, access control, system reliability and inclusion of special groups, and achieves a technological leap from passive response to proactive early warning and from general services to personalized protection.

[0022] The multimodal perception module includes a video perception unit, an audio perception unit, and a physiological sensing unit; the data analysis module includes a sensitive information leakage risk analysis unit (corresponding to sensitive information leakage issues), a fine-grained permission compliance analysis unit (corresponding to coarse-grained permission management issues), a key component failure risk analysis unit (corresponding to single point of failure issues), and a special population needs adaptation analysis unit.

[0023] The video sensing unit uses wide-angle color-depth dual cameras to monitor user posture, fall events, facial pain expressions, and gesture distress signals online, including the coordinates of key points of the human skeleton, the velocity of the center of gravity projected on the ground, the amplitude of facial movement changes, the trajectory of hand waving, and the duration of cessation. The audio sensing unit uses a microphone array, combined with sound event detection and speech recognition algorithms, to monitor ambient sounds and user speech online, including but not limited to keywords for help, abnormal sounds, painful groans, and specific voice commands from blind users. The physiological sensing unit monitors heart rate, blood oxygen, skin conductance, and body temperature online via a wristband-type multi-parameter sensor, including instantaneous heart rate variability (LF / HF ratio), pulse oxygen saturation drop gradient, skin conductance surge magnitude, and body temperature ≥38℃ for several seconds, and generates a comprehensive physiological abnormality index.

[0024] The advantages are: by complementing and co-verifying multi-source information from video, audio and physiological sensors, it can achieve the beneficial effect of capturing abnormal user states (such as falls, signs of sudden death) in a comprehensive and high-precision manner, thereby reducing the false alarm and false alarm rates of the system.

[0025] The special group data collection module includes a sign language collection unit for the deaf and mute, a voice interaction unit for the blind, and a simplified interaction unit for the cognitive impairment. The sign language acquisition unit for deaf and mute people uses the real-time video stream provided by the video perception unit and calls a dedicated sign language recognition model to collect sign language information of deaf and mute people, including continuous hand gestures, hand shapes, movement trajectories and facial expressions, including alarm intentions, basic needs (such as water, pain) and simple descriptions of symptoms. The blind voice interaction unit captures speech through the audio perception unit and runs an optimized voice interaction process to collect blind voice interaction information centered on short preset commands (such as "alarm" or "I'm not okay"), including active alarm confirmation and simple feedback to system voice prompts (such as "yes" or "no"). The simplified interaction unit for cognitive impairment uses a large touch button or physical switch on the device, along with eye-catching icons and light prompts, to collect the user's direct trigger signals, realize one-click alarm, and simplify complex information collection into a simple "yes / no" answer or status selection.

[0026] The advantages are: by using a group-specific customized data collection strategy, standardized intent tags, and isolated interaction channels, it can accurately cover the differentiated needs of deaf, blind, and cognitively impaired users, thereby breaking down communication barriers and ensuring that all groups can effectively use emergency services.

[0027] Before analysis, the sensitive information leakage risk analysis unit encrypts the data using a post-quantum cryptography algorithm. The encryption key is not static but dynamically generated, and the session key is calculated. The calculation formula is as follows: ; In the formula, Indicates the session key. This represents the current timestamp (accurate to the second), used to ensure the key is used only once. This refers to the preset master seed, which varies depending on the device and is written to the secure storage area at the factory. This indicates a single random number generated by the device, ensuring that the key is different each time. This represents a post-quantum secure key derivation function conforming to NIST SP 800–56C (such as Kyber-based KDF), and the dynamic key... The working key for this risk analysis unit is used to encrypt video, audio, and physiological data before subsequent calculations, ensuring that there is no risk of static key leakage during the analysis process.

[0028] Sensitive information encryption protection module based on session key End-to-end quantum encryption is applied to video, audio, raw physiological data, and intermediate analysis results to ensure the confidentiality and resistance to quantum attacks throughout the entire lifecycle of data transmission, storage, and analysis, achieving the technical effect that even if the data packets are intercepted, they cannot be decrypted and restored.

[0029] The advantage is that it calculates the session key. It performs end-to-end encryption on video, audio, raw physiological data, and intermediate analysis results, achieving confidentiality protection throughout the entire lifecycle of data transmission, storage, and analysis. This helps resist the threat of quantum computing cracking static keys, ensuring that even if data packets are intercepted, sensitive information (such as patient identity and disease details) cannot be decrypted and restored.

[0030] The fine-grained permission compliance analysis unit (corresponding to coarse-grained permission management issues) calculates the permission compliance rate. The calculation formula is as follows: ; In the formula, Indicates the permission compliance rate. This indicates the number of times healthcare professionals have been authorized to correctly access historical records. This indicates the number of access attempts made by unauthorized personnel (such as unauthorized medical interns). This indicates the total number of access requests.

[0031] The fine-grained access control module is based on the access compliance rate. Dynamic access control is performed, and the compliance rate of permissions is assessed. Maintain normal authorization policy at 95% compliance rate; when the permission compliance rate is... When the compliance rate is between 80% and 95%, initiate secondary identity verification (including facial recognition + short-term token). If the error rate is ≤80%, immediately downgrade to read-only mode and push an abnormal audit alert.

[0032] The advantage is that it calculates the permission compliance rate. Differentiated access control policies are implemented based on different compliance rate thresholds to achieve hierarchical authorization of medical staff and real-time interception of abnormal access. This achieves the beneficial effect of refining the granularity of permission management, strictly restricting unauthorized personnel (such as unapproved interns) from obtaining sensitive data, and ensuring the security of medical data access.

[0033] The critical component failure risk analysis unit (corresponding to single point of failure problems) calculates the failure impact. The calculation formula is as follows: ; In the formula, Indicates the degree of impact of the fault. This indicates the failure probability of centralized components (master server / communication link) (historical data statistics, such as the annual failure probability of the master server). Indicates the backup switch success rate. This represents the service interruption duration coefficient.

[0034] The distributed fault self-healing module is based on the degree of fault impact. Perform self-healing strategy assessment, and determine the impact of the fault. When the value is less than 0.3, the local edge node is enabled to continue providing services; when the failure impact is less than or equal to 0.3... When the fault impact is ≤0.7, a primary / standby hot switch is triggered and multi-hop relay communication is initiated; when .... When the value is greater than 0.7, it is determined to be a global fault, and offline caching mode is automatically enabled. Data is then retransmitted in batches after communication is restored.

[0035] The advantage is that it allows for the calculation of the impact of failures. By quantifying the combined impact of centralized component failure probability, backup switch success rate, and business interruption duration, a graded self-healing strategy is implemented based on different impact thresholds to achieve rapid response and business continuity assurance when critical components fail. This avoids the single point of failure risk of centralized architecture and prevents the system from falling into a global paralysis due to the interruption of the main server / communication link.

[0036] Special population needs adaptation analysis unit calculates the success rate of simplified operation for people with cognitive impairment. The calculation formula is as follows: ; In the formula, This indicates the success rate of simplified procedures for individuals with cognitive impairment. This indicates the number of times a person with cognitive impairment has successfully completed a simplified operation (such as pressing the red alarm button to trigger an alarm). This represents the total number of operation attempts by individuals with cognitive impairments (including accidental touches of other buttons and instances of no operation being triggered). The calculation parameters for the simplified operation success rate of deaf and blind individuals are the same as those for individuals with cognitive impairments.

[0037] The special population interaction optimization module simplifies operations and improves the success rate for people with cognitive impairments. Interaction strategy assessment was conducted to evaluate the success rate of simplified operations in individuals with cognitive impairment. When the success rate is ≥90%, maintain the existing interaction process; when the success rate for simplifying operations is ≤70% for individuals with cognitive impairments. When the success rate is below 90%, automatically enlarge the icon, reduce the voice prompt speed, and increase vibration confirmation; simplify the operation for cognitively impaired individuals to improve success rate. When the threshold is below 70%, switch to the minimalist mode of single button-single light-single vibration and remotely push auxiliary prompts to caregivers.

[0038] The advantage is that it simplifies the success rate of procedures for individuals with cognitive impairments by calculating their success rate. The interaction strategy is dynamically adjusted based on different success rate thresholds to continuously optimize the interaction experience for special groups, thereby further reducing the difficulty of alarm operation for special groups and improving their emergency call success rate and user experience.

[0039] In summary, the multimodal perception-based emergency call system employs an eight-level architecture design: a multimodal perception module (data acquisition layer), a special group data acquisition module (special needs layer), a data analysis module (core computing layer), a sensitive information encryption protection / fine-grained access control / distributed fault self-healing / special group interaction optimization module (strategy execution layer), and an instruction execution module (closed-loop control layer). This architecture integrates key technologies such as dynamic post-quantum encryption, hierarchical access control, distributed hot standby self-healing, and special group-specific interaction. The core value of this system lies in: firstly, achieving automatic emergency call system response through multi-dimensional perception data collaboration. The system offers several key advantages: first, it provides dynamic and accurate identification; second, it addresses the shortcomings of traditional emergency call systems by optimizing data collection and interaction for specific population groups; third, it constructs a full lifecycle protection system for sensitive medical data through quantum-secure encryption and fine-grained permissions; and fourth, it ensures business continuity in the event of component failures through a distributed self-healing strategy. These advantages effectively solve the pain points of traditional emergency call systems in terms of perception, adaptation, security, and reliability, providing more efficient, safe, and convenient emergency call services for different populations (especially special groups). The system can be widely applied in scenarios such as home-based elderly care, community healthcare, and rehabilitation institutions, helping to improve the accessibility and reliability of emergency services.

[0040] 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 automatic alarm system for emergency call devices based on multimodal perception, characterized in that, It includes a multimodal perception module, a special group data collection module, a data analysis module, a sensitive information encryption and protection module, a fine-grained access control module, a distributed fault self-healing module, a special group interaction optimization module, and an instruction execution module; The multimodal perception module collects user status data online through three types of sensors: video, audio, and physiological sensors, and generates a multimodal feature vector with a unified timestamp, which is then output to the data analysis module. The special group data collection module targets three groups: deaf-mute, blind, and cognitively impaired individuals. It collects sign language video streams, low-volume voice commands, and one-click trigger signals, converts them into standardized intent tags, and outputs them to the data analysis module. The data analysis module receives output data from the multimodal perception module and the special group data acquisition module, and calculates the session key. Permission compliance rate Fault impact Simplified operation success rate for people with cognitive impairment The calculation results are sequentially output to the sensitive information encryption protection module, the fine-grained access control module, the distributed fault self-healing module, and the special population interaction optimization module; The sensitive information encryption protection module receives the session key output by the data analysis module. End-to-end encryption is performed on video, audio, raw physiological data, and intermediate analysis results; The fine-grained permission control module is based on the permission compliance rate. Perform dynamic access control judgments; The distributed fault self-healing module is based on the fault impact. Determine self-healing strategies; The special population interaction optimization module simplifies operations based on the success rate of individuals with cognitive impairment. Conduct interaction strategy assessments and remotely push assistance prompts to caregivers; The instruction execution module receives encryption, permission, self-healing, and interaction optimization instructions from the above modules, arbitrates them according to priority, and uniformly controls the underlying hardware and network resources to complete the actions, and sends back the execution result log to form a closed loop.

2. The emergency call system based on multimodal perception as described in claim 1, characterized in that: The multimodal perception module includes a video perception unit, an audio perception unit, and a physiological sensing unit; the data analysis module includes a sensitive information leakage risk analysis unit, a fine-grained permission compliance analysis unit, a key component failure risk analysis unit, and a special population needs adaptation analysis unit.

3. The emergency call system based on multimodal perception as described in claim 2, characterized in that: The video sensing unit uses a wide-angle color-depth dual camera to monitor user posture, fall events, facial pain expressions, and gestures for help online; the audio sensing unit uses a microphone array, combined with sound event detection and speech recognition algorithms, to monitor ambient sounds and user voice online; and the physiological sensing unit uses a wristband-type multi-parameter sensor to monitor heart rate, blood oxygen, skin conductance, and body temperature online.

4. The emergency call system based on multimodal perception as described in claim 1, characterized in that: The special group data collection module includes a sign language collection unit for deaf-mute people, a voice interaction unit for blind people, and a simplified interaction unit for cognitive impairment. The deaf and mute sign language acquisition unit collects deaf and mute sign language information, including continuous hand gestures, hand shapes, movement trajectories, and facial expressions, through the real-time video stream provided by the video perception unit and by calling a dedicated sign language recognition model. The blind voice interaction unit captures speech through an audio perception unit and runs an optimized voice interaction process to collect blind voice interaction information centered on short preset commands. The simplified interaction unit for cognitive impairment collects the user's direct trigger signals through a large touch button or physical switch on the device, supplemented by eye-catching icons and light prompts.

5. The emergency call system based on multimodal perception according to claim 2, characterized in that: Before analysis, the sensitive information leakage risk analysis unit encrypts the data using a post-quantum cryptography algorithm and calculates the session key. The calculation formula is as follows: ； In the formula, Indicates the session key. Indicates the current timestamp. This indicates the preset master seed. This represents a single random number generated by the device. This represents a post-quantum secure key derivation function conforming to NIST SP 800-56C.

6. The emergency call system based on multimodal perception according to claim 2, characterized in that: The fine-grained permission compliance analysis unit calculates the permission compliance rate. The calculation formula is as follows: ； In the formula, This indicates the permission compliance rate. This indicates the number of times healthcare professionals have been authorized to correctly access historical records. This indicates the number of unauthorized access attempts. This indicates the total number of access requests.

7. The emergency call system based on multimodal perception according to claim 2, characterized in that: The critical component failure risk analysis unit calculates the failure impact. The calculation formula is as follows: ； In the formula, Indicates the degree of impact of the fault. This represents the probability of failure of a centralized component. Indicates the backup switch success rate. This represents the service interruption duration coefficient.

8. The emergency call system based on multimodal perception according to claim 2, characterized in that: The special population needs adaptation analysis unit calculates the success rate of simplified operations for individuals with cognitive impairment. The calculation formula is as follows: ； In the formula, This indicates the success rate of simplified procedures for individuals with cognitive impairment. This indicates the number of times a person with cognitive impairment has successfully completed a simplified task. This indicates the total number of attempts made by individuals with cognitive impairment.

9. The emergency call system based on multimodal perception according to claim 1, characterized in that: The sensitive information encryption protection module is based on session keys. End-to-end post-quantum encryption is applied to raw video, audio, physiological data, and intermediate analysis results. The fine-grained permission control module is based on the permission compliance rate. Dynamic access control is performed, and the compliance rate of permissions is assessed. Maintain normal authorization strategy at 95%; When the permission compliance rate When the compliance rate is between 80% and 95%, initiate secondary identity verification; this will be implemented when the permission compliance rate is [not specified]. If the error rate is ≤80%, immediately downgrade to read-only mode and push an abnormal audit alert.

10. The emergency call system based on multimodal perception according to claim 1, characterized in that: The distributed fault self-healing module is based on the fault impact. Perform self-healing strategy assessment, and determine the impact of the fault. When the value is less than 0.3, the local edge node is enabled to continue providing services; when the failure impact is less than or equal to 0.3... When the fault impact is ≤0.7, a primary / standby hot switch is triggered and multi-hop relay communication is initiated; when .... When the value is greater than 0.7, it is determined to be a global fault, and offline caching mode is automatically enabled. Data is then retransmitted in batches after communication is restored. The special population interaction optimization module simplifies operations based on the success rate of individuals with cognitive impairment. Interaction strategy assessment was conducted to evaluate the success rate of simplified operations in individuals with cognitive impairment. When the success rate is ≥90%, maintain the existing interaction process; when the success rate for simplifying operations is ≤70% for individuals with cognitive impairments. When the percentage is less than 90%, the icons will automatically enlarge, the voice prompt speed will decrease, and vibration confirmation will be added. When simplifying procedures for people with cognitive impairment, the success rate is higher. When the threshold is below 70%, switch to the minimalist mode of single button-single light-single vibration and remotely push auxiliary prompts to caregivers.

Images