Monitoring system and method based on hierarchical agent architecture and dual hardware confirmation

Abstract

The invention discloses a monitoring system and method based on hierarchical agent architecture and two-stage hardware confirmation, and belongs to the technical field of Internet of Things and artificial intelligence. The system comprises a bracelet end which is provided with a main key and a touch key and is internally provided with a security chip for generating a digital signature; the mobile phone end deploys a private agent and an encryption sandbox and only responds to the instruction confirmed by the hardware; and the cloud end is used for processing non-sensitive tasks by adopting a Manager / Worker multi-tenant architecture. The method comprises the steps that the bracelet collects a voice instruction, a mobile phone end analyzes an intention, and sensitive operation is issued to the bracelet; the old person selects through the main key and confirms through the touch key, and the security chip generates a signature and returns the signature; after verifying the signature, the mobile phone decrypts the certificate and calls the API to execute operation; and broadcasting and feeding back a result, and performing key operation on a chain for evidence storage. Through cooperation of two-stage hardware confirmation and the layered intelligent agent, hardware right confirmation of sensitive operation and local isolation of privacy data are achieved, full thinking time is given to old people while operation authenticity is guaranteed, and misoperation is avoided.

Description

Technical Field

[0001] This invention relates to the field of smart wearable device technology, and specifically to a monitoring system and method based on a hierarchical intelligent agent architecture and dual hardware verification. Background Technology

[0002] The demand for health monitoring, safety protection, and emotional companionship among the elderly is growing. Currently, smart bracelets mainly fall into two categories: one is a "passive" device focused on health monitoring, which only collects data and reports it to an app, lacking proactive intelligent services; the other is an "interactive" device based on voice assistants, but usually requires the elderly to manually operate a mobile app or rely on screen touch control, which is extremely unfriendly to elderly people with declining vision or dexterity.

[0003] In recent years, the emergence of intelligent agent frameworks such as OpenClaw has enabled AI to autonomously operate applications to complete complex tasks. However, existing intelligent agents have two major drawbacks: First, the security boundaries are blurred. If an intelligent agent wants to operate sensitive functions such as WeChat and payment, it needs to obtain user privacy credentials, which poses a risk of leakage. Second, the interaction confirmation mechanism is simple, usually relying only on a single button press or biometric recognition. Elderly people are prone to misoperation and lack the time to think about secondary confirmation.

[0004] Furthermore, while payment wristbands from manufacturers like Infineon support biometric payments, they rely on biometric features such as electrocardiograms and lack a collaborative architecture with cloud-based intelligent agents and mobile-side private intelligent agents. How to enable senior citizens to securely and conveniently utilize AI capabilities to complete tasks such as WeChat communication and utility bill payments through simple voice commands and two-level hardware button confirmation is a pressing technological challenge. Summary of the Invention

[0005] To address the problems existing in the prior art, the present invention aims to provide a monitoring system and method based on a hierarchical intelligent agent architecture and dual hardware confirmation, which addresses the issues of ambiguous security boundaries and a single interactive confirmation mechanism in existing intelligent monitoring systems. This system enables the elderly to safely and conveniently complete daily operations such as WeChat communication and utility bill payment through voice command triggering and two-level physical button confirmation. While ensuring the authenticity of the operation and privacy security, it also gives the elderly sufficient time to think to prevent misoperation.

[0006] To solve the above-mentioned technical problems, the present invention adopts the following technical solution: A monitoring system based on a hierarchical intelligent agent architecture and dual hardware verification, the system comprising: Wearable hardware verification layer, configured in wristband devices, includes physical interaction units and security chips; The physical interaction unit includes a main button and a touch button. The main button is configured to cycle through operation options, switching between options each time it is triggered. The touch button is configured to confirm the currently selected operation option. The security chip has a built-in independent storage area and a computing unit for storing a unique, non-exportable device private key, and a corresponding device public key. In response to the final confirmation command triggered by the touch key, the security chip internally calls the private key to perform digital signature calculation on sensitive operation data, generates a signature result with non-repudiation, and sends it to the mobile terminal through the near-field wireless communication module. A private intelligent agent layer for mobile terminals, deployed on smartphones for the elderly, includes a background service module, an encrypted storage module, and a verification execution module; The background service module runs as a system background service process, and achieves seamless interaction and status monitoring of the target application through an automated interaction interface or an application programming interface (API). The encrypted storage module is a hardware-level encrypted sandbox used to isolate and store highly sensitive privacy data, including long-term tokens for third-party services and payment account credentials, and to set access barriers for all external process access requests. The verification execution module is equipped with a hardware confirmation trigger threshold. It continuously listens for operation commands issued by the wearable device and only authorizes the use of the corresponding credentials from the encrypted storage module to execute sensitive transactions after verifying that the command is accompanied by a valid digital signature signed by the security chip. The cloud-based public intelligent agent cluster is deployed using a Manager / Worker multi-tenant architecture. The Manager node is responsible for global task scheduling, tenant lifecycle management, and the elastic scaling of Worker nodes; The Worker container dynamically allocates an independent running instance to each registered senior citizen, encapsulating their exclusive public intelligent agent. This intelligent agent only has the permission to access de-identified data and call general computing resources, and is dedicated to handling non-sensitive computing tasks that do not require the combination of local privacy data. The processing results of the cloud-based public intelligent agent are encrypted or de-identified and then synchronized to the private intelligent agent layer of the mobile terminal for decision-making reference. The wearable device, the mobile terminal, and the cloud interact with each other through an encrypted communication link, forming a layered intelligent agent architecture with end-edge-cloud collaboration.

[0007] The main button is also equipped with a voice feedback module, which simultaneously announces the currently selected item during the selection process; after the touch button is confirmed, a digital signature containing at least the operation identifier, transaction amount, timestamp and random number is generated upon receiving the confirmation instruction from the touch button, and then transmitted back to the mobile terminal via a wireless communication link.

[0008] The main button is also equipped with an emergency trigger mode: when pressed for more than a preset time threshold, an emergency help process is triggered. This long press function is parallel and independent of the short press selection function. The main button and the touch button adopt a concave anti-accidental touch structure, and the trigger force threshold is set to ≥2N.

[0009] The wristband also includes an automatic timeout cancellation module, configured to automatically cancel the current operation process and broadcast a prompt message if the option selection and confirmation operation is not completed within a preset time.

[0010] The mobile terminal private intelligent agent layer also includes an initial authorization guidance module, configured to guide the elderly to complete the authorization operation of third-party services when the system is first activated, and store the obtained authorization certificate in the encrypted sandbox after encrypting it with the wristband's public key.

[0011] The cloud-based public intelligent agent cluster achieves resource isolation based on containerization technology. Each Worker container has an independent file system, network space, and operating environment, realizing data isolation and session isolation between different users.

[0012] A two-level hardware verification and monitoring method based on a hierarchical intelligent agent architecture, applied to the system, includes the following steps: S1: Voice Acquisition and Transmission: The wristband collects the elderly's voice commands and transmits them in real time to the private intelligent agent on the mobile phone via a wireless communication link. S2: Intent Recognition and Task Triage: The private intelligent agent performs semantic parsing on the voice command to identify the operation intent; it determines whether the operation belongs to a sensitive operation type. If it is a non-sensitive operation, a task request is constructed and forwarded to the cloud public intelligent agent for execution; if it is a sensitive operation, it proceeds to step S3. S3: Hardware Confirmation Request Generation and Issuance: The private intelligent agent generates an operation confirmation request based on the parsing results and sends it to the wristband via a wireless communication link; after receiving the request, the wristband prompts the elderly with the operation content through voice broadcast and enters the hardware confirmation waiting state; S4: Cyclic Selection and Voice Feedback: Seniors can cycle through operation options by triggering the main button. Each time the main button is triggered, the wristband will simultaneously broadcast the semantic description of the currently selected item. S5: Hardware Confirmation and Signature Generation: When the elderly select the desired operation option, the touch key is triggered for final confirmation; if it is to confirm the execution of a sensitive operation, the security chip on the wristband calls the private key stored in the chip to generate a digital signature for the sensitive operation data containing the operation identifier, timestamp, and random number, and sends the signature result and confirmation instruction back to the mobile phone; if it is to cancel the operation, only the cancellation instruction is sent back. S6: Signature Verification and Credential Retrieval: The private intelligent agent on the mobile device receives the returned data and verifies the validity of the digital signature; after successful verification, it decrypts the corresponding authorization credential from the encrypted sandbox and calls the official application programming interface of the third-party service to complete the sensitive operation. S7: Result Feedback and Evidence Preservation: The operation execution result is transmitted back to the wristband via wireless communication link and fed back to the elderly in the form of voice broadcast; at the same time, the operation log and signature data of key operations are preserved and recorded.

[0013] Sensitive operations in step S2 include, but are not limited to: sending instant messaging messages, initiating voice calls, authorizing payments, and confirming electronic signatures; non-sensitive operations include, but are not limited to: checking household bills, generating health reports, playing entertainment content, and checking the weather.

[0014] In step S3, the wristband broadcast uses dialect speech synthesis technology, supporting no less than 40 dialects; the timeout period for the elderly to select and confirm after the broadcast is preset to 30 seconds, and the cancellation process is automatically triggered after the timeout.

[0015] The digital signature generated in step S5 includes at least: a unique operation identifier, transaction amount, timestamp, anti-replay random number, and current wristband status parameters to prevent replay attacks and forged requests.

[0016] The cloud-based public intelligent agent cluster adopts a Manager / Worker architecture, where the Manager node is responsible for task semantic decomposition and Worker resource allocation, and each Worker container runs independently; the Manager centrally manages the global API credentials of third-party services, and each Worker only holds a temporarily issued access token with a limited validity period.

[0017] A computer-readable storage medium having a computer program stored thereon, characterized in that the program, when executed by a processor, implements the method as described in any one of claims 7 to 11.

[0018] The beneficial effects of this invention are: 1. A dual confirmation mechanism based on physical separation eliminates the risk of misoperation. The main button and touch keys are functionally decoupled, defined as "option cycle selection key" and "final confirmation execution key" respectively. This two-step physical operation of "selection-confirmation" provides double protection in both time and operation dimensions, fundamentally solving the problem of accidental touches in single-point touch control and ensuring the prudence and authenticity of the operation.

[0019] 2. Edge-cloud layered and decoupled architecture to achieve a balance between privacy and performance. The intelligent agent is broken down into a "wristband-side hardware verification layer," a "mobile phone-side private data layer," and a "cloud-based public computing power layer." Through the physical isolation and functional division of the three-layer architecture, it ensures that core privacy data such as WeChat tokens and payment credentials always reside in the secure sandbox on the mobile phone and are never leaked. At the same time, it utilizes the elastic computing power of the cloud to handle large-scale and complex computing tasks, thus resolving the inherent contradiction between privacy protection and computing power expansion in the traditional single intelligent agent architecture.

[0020] 3. Hardware-level trust anchors ensure the authenticity and non-repudiation of operations. Using the built-in security chip (SE) in the wristband as the root of trust, combined with the trigger events of the physical buttons, a hardware verification system of "what you see is what you sign" is constructed. When the elderly select and confirm an operation through the physical buttons, the security chip internally calls the solidified private key to generate a digital signature, ensuring the authenticity, integrity, and non-repudiation of the operation command at the hardware level, effectively resisting remote attacks, interface hijacking, and accidental triggering.

[0021] 4. A human-computer intelligent interaction paradigm, achieving true direct voice communication. The system adopts a "one-time authorization, lifetime use" model. Seniors only need their guardians to assist with third-party application authorization during the initial use. Throughout their entire lifespan, they can complete complex operations simply by issuing voice commands via the wristband, along with simple two-level button confirmation. This design completely eliminates the complex interactions of smartphone apps, achieving a barrier-free "voice-based operation" experience.

[0022] 5. Based on a containerized multi-tenant isolation architecture, it supports high-concurrency elastic scaling. The cloud cluster adopts a Manager / Worker master-slave architecture, combined with Docker / Kubernetes containerization technology, to dynamically allocate independent Worker container instances to each user. This architecture achieves tenant-level data isolation and computing power isolation, ensuring the security of data for tens of millions of users while supporting elastic scaling of computing resources to meet the needs of large-scale commercial deployments.

[0023] 6. End-to-end trusted evidence storage meets high-level compliance audit requirements. For sensitive operations such as payments and authorizations, the system automatically generates a data packet containing an operation ID, timestamp, random number, and hardware signature, and simultaneously uploads it to the blockchain for evidence storage. This mechanism creates an immutable and traceable audit log, providing legally valid evidence for dispute resolution and meeting the compliance requirements for financial payment levels. Attached Figure Description

[0024] Figure 1 This is a diagram of the overall system architecture of the present invention; Figure 2 This is a flowchart of the two-level hardware verification method of the present invention; Figure 3 This is a timing diagram of the interaction between the main button selection and the touch button confirmation in this invention. Figure 4 This is a diagram of the cloud-based public intelligent agent cluster architecture of the present invention. Detailed Implementation

[0025] To make the objectives, claims, and advantages of this invention clearer, the present invention will be clearly and completely described below with reference to specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0026] The technical solution adopted by this invention to solve its technical problem is as follows: A layered intelligent agent architecture with end-edge-cloud collaboration is constructed, and a two-level hardware verification mechanism based on physical separation is innovatively designed on the wearable device. This architecture achieves the technical effects of "hardware ownership verification - privacy protection - external computing power" through the decoupling and collaboration of three layers of functions: the wristband, the mobile phone, and the cloud. Specifically, it includes: 1. Wearable hardware verification layer (wristband end) This layer is deployed on the smart bracelets worn by the elderly, serving as the system's physical trust anchor and human-computer interaction entry point, including: Physical interaction unit: consists of main buttons and touch keys; The main button is configured as a cycle selection button for operation options. Each time it is triggered, it switches between one option in sequence and triggers the built-in voice module of the bracelet to broadcast the semantic description of the currently selected item. The touch key is configured as the final confirmation key, used to confirm the currently selected operation option or to confirm the cancellation of the current operation process; Security Chip Unit: It has a built-in independent security chip with an isolated storage area and a dedicated computing unit for storing a unique, non-retrievable device private key. In response to the final confirmation command triggered by the touch key, the security chip internally calls the private key to perform digital signature calculation on sensitive operation data, generating a signature result with non-repudiation. Voice interaction unit: integrates a voice acquisition microphone and a voice synthesis module to receive voice commands from the elderly, and to broadcast operation options, confirmation prompts and operation results; The "select-confirm" two-level hardware confirmation mechanism provides double protection in both time and operation dimensions, giving the elderly ample room for careful consideration and fundamentally solving the technical problem of accidental touches during single touch control.

[0027] 2. Mobile Terminal Private Intelligent Agent Layer (Mobile Phone) This layer is deployed on senior-friendly smartphones, serving as a privacy data security island and command verification hub for the system, including: Background service module: It runs residently as a system background service process, and achieves seamless interaction and status monitoring of the target application through automated interaction interface or application programming interface (API); Encrypted storage module: This is a hardware-level encrypted sandbox used to isolate and store highly sensitive privacy data, including long-term tokens for third-party services and payment account credentials, and to set up access barriers for all external process access requests; Verification and execution module: It has a hardware confirmation trigger threshold, continuously listens for operation commands issued by the wearable device, and only authorizes the use of the corresponding credentials from the encrypted storage module to execute sensitive transactions after verifying that the command is accompanied by a valid digital signature signed by the security chip. Communication module: Establishes an encrypted communication link with the wristband via a wireless communication link, and establishes a secure channel with the cloud via a mobile network or Wi-Fi.

[0028] 3. Cloud-based public intelligent agent cluster (cloud) This layer is deployed on a cloud server cluster, serving as the system's elastic computing power center and public service engine, and adopts a Manager / Worker multi-tenant architecture. Manager node: Responsible for global task scheduling, tenant lifecycle management, and elastic scaling of Worker nodes; Worker Container: Dynamically allocates an independent container instance to each registered user, encapsulating their own public intelligent agent runtime environment; each Worker container has an independent file system, network space and runtime environment, achieving strong isolation between users; The public intelligent agent instance only has the permission to access de-identified data and call general computing resources, and is dedicated to handling non-sensitive computing tasks that do not require the combination of local privacy data, including but not limited to: cross-time period health data trend analysis, life bill query, personalized content generation, and elderly community matching and recommendation.

[0029] Based on the above system architecture, the present invention also provides a two-level hardware verification and monitoring method based on a hierarchical intelligent agent architecture, comprising the following steps: Step S1: Initial Authorization and Trust Establishment Phase When the system is first activated, the elderly person or their assistant completes the authorization operation for third-party services (such as WeChat and Alipay) on their mobile phone. After obtaining the authorization credential, the private intelligent agent on the mobile phone calls the public key of the security chip on the wristband to encrypt the credential and stores the encrypted credential data in the hardware-level encryption sandbox on the mobile phone to establish an initial trust relationship.

[0030] Step S2: Voice command acquisition and intent recognition stage The wristband uses its built-in microphone to collect the elderly person's voice commands in real time. After initial local noise reduction, the commands are transmitted to the mobile phone's private intelligent agent via an encrypted wireless communication link. The mobile phone's private intelligent agent performs semantic analysis on the voice commands, identifies the intention of the operation, and determines whether the operation is a sensitive type. If it is a non-sensitive operation (such as checking living bills, generating health reports, telling stories, or checking the weather), a task request is constructed and forwarded to the cloud public intelligent agent through a secure channel for execution, and then proceed to step S5; If it is a sensitive operation (such as sending WeChat messages, initiating voice calls, authorizing payments, or confirming electronic signatures), then proceed to step S3 to initiate the hardware confirmation process.

[0031] Step S3: Hardware Confirmation Request Generation and Issuance Phase The mobile-side private intelligent agent generates an operation confirmation request based on the parsing results. This request includes key information such as the operation type, the operation target, and the transaction amount (if applicable), and is sent to the wristband via a wireless communication link. After receiving the request, the wristband uses a voice synthesis module to announce the operation content (e.g., "There is a water bill of 200 yuan that needs to be paid, do you confirm?") and enters a hardware confirmation waiting state, prompting the elderly to "press the main button to select and press the touch button to confirm".

[0032] Step S4: Two-level hardware selection and confirmation stage This stage represents the core innovative step of the invention, achieving careful confirmation through two physical button presses: S4-1 Cyclic Selection: Seniors can cycle through preset operation options by triggering the main button (the options include at least "Confirm Execution" and "Cancel Operation"). Each time the main button is triggered, the wristband will simultaneously broadcast the semantic description of the currently selected item via voice. S4-2 Final Confirmation: When the elderly select the desired operation option, the touch key is triggered for final confirmation; If the elderly person confirms the execution of a sensitive operation, the security chip in the wristband will call the private key stored in the chip to generate a digital signature on the sensitive operation data, which includes the operation identifier, transaction amount, timestamp, and anti-replay random number, and send the signature result and confirmation instruction back to the mobile phone. If the elderly person confirms the cancellation, only the cancellation instruction is sent back; no digital signature needs to be generated. S4-3 Timeout Protection: If the selection and confirmation operation is not completed within the preset time (preferably 30 seconds), the wristband will automatically cancel the current operation process and broadcast a prompt to prevent the process from getting stuck.

[0033] Step S5: Signature Verification and Sensitive Operation Execution Phase The mobile app's private AI agent receives data transmitted back from the wristband: If a cancellation command is received, the process will be terminated directly, with the option to announce "Operation has been cancelled"; If a confirmation instruction with a digital signature is received, the validity of the digital signature is first verified (including signature verification, timestamp freshness check, and random number anti-replay check). After successful verification, the corresponding authorization credential is decrypted from the encrypted sandbox, and the official application programming interface (API) of the third-party service is called to complete the sensitive operation. The operation result is transmitted back to the wristband via wireless communication link and fed back to the elderly in the form of voice broadcast (e.g., "Water fee of 200 yuan has been successfully paid").

[0034] Step S6: Cloud Execution Phase of Non-Sensitive Tasks For the non-sensitive operations in step S2, the Manager node of the cloud-based public intelligent agent cluster receives the task request, parses the task type, and distributes it to the corresponding elderly person's Worker container for execution. The public intelligent agent instance in the Worker container calls general computing resources to complete the processing. The processing result is encrypted or desensitized and then synchronized to the private intelligent agent on the mobile phone. Finally, the elderly person is given feedback via voice broadcast on the wristband.

[0035] Step S7: Operation Evidence Preservation and Traceability Stage For critical and sensitive operations such as payment and authorization, the system automatically packages operation logs, digital signatures, timestamps, and other data, uploads them to the blockchain for evidence storage, and forms an immutable and traceable audit log to meet financial-grade compliance audit requirements.

[0036] Preferably, the main button and touch button adopt a concave anti-accidental touch structure, and the trigger force threshold is set to ≥2N to reduce accidental triggering in daily activities; the main button is also equipped with an emergency trigger mode: when the main button is pressed for more than a preset time threshold (preferably 3 seconds), an emergency help process is triggered. This long press function and the short press selection function run in parallel and independently without interfering with each other.

[0037] Preferably, the cloud-based public intelligent agent cluster uses Docker or Kubernetes containerization technology to achieve resource isolation. Each Worker container has an independent file system, network space, and operating environment, achieving memory isolation and session isolation between users. The Manager node centrally manages the global API credentials of third-party services, and each Worker only holds a temporarily issued access token with a limited validity period, further reducing security risks.

[0038] Preferably, the two-level hardware confirmation mechanism of the wristband also includes an operation preview function: when the elderly press the main button to select in a cycle, the wristband not only announces the option name, but also announces a preview of the execution consequences of the option (such as "Confirm payment of 200 yuan" or "Cancel this payment"), to help elderly people with weaker cognitive abilities understand the meaning of the options.

[0039] Preferably, the wristband's voice broadcast supports multiple dialects and can be configured to the corresponding dialect according to the elderly's habits, reducing the language comprehension threshold; the broadcast speed is adjustable to suit elderly people with different hearing conditions.

[0040] Example 1: Two-level hardware confirmation process for sending WeChat voice messages This embodiment describes the specific implementation process of sending WeChat messages using the system described in this invention, focusing on the complete process of voice command triggering, two-level hardware confirmation, and secure signature.

[0041] Initial authorization phase: When the system is first activated, the guardian or elderly person opens the accompanying management APP on their mobile phone and clicks the "WeChat Authorization" entry. The system calls the official WeChat authorization interface and jumps to the WeChat QR code page to complete the authorization operation. After the private smart agent on the mobile phone obtains the WeChat long-term refresh token, it calls the public key of the security chip on the wristband to encrypt the token and stores the encrypted token data in the hardware-level encryption sandbox on the mobile phone, thus completing the establishment of the initial trust relationship.

[0042] Daily use phase: Step 1: Voice command acquisition and intent recognition When an elderly person wears the bracelet, they can trigger the voice wake-up function (or press the voice button) and say the command: "Send my son a WeChat message saying that my blood pressure is normal today." The built-in microphone on the bracelet collects the voice signal, which is then processed locally for noise reduction and transmitted in real time to the private smart agent on the mobile phone via an encrypted wireless communication link.

[0043] The mobile-side private AI agent invokes a locally deployed lightweight natural language understanding model to perform semantic parsing of voice commands and identify: Operation type: Send WeChat message (sensitive operation) Recipient: Son (mapped from contacts to a specified contact ID) Message content: "My blood pressure is normal today." Step 2: Hardware Confirmation Request Generation and Issuance The private agent determines that the operation is sensitive and initiates a hardware verification process. It generates an operation verification request containing the following information: Operation summary: "Send a WeChat message to [son]" Content preview: "My blood pressure is normal today" Optional actions: [Confirm Send] / [Cancel] The request was sent to the wristband via a wireless communication link.

[0044] Step 3: Two-level hardware selection and confirmation After receiving the request, the wristband will announce the following via its voice synthesis module: "Send a WeChat message to your son with the message 'My blood pressure is normal today.' Press the main button to select, and press the touch button to confirm." At the same time, the corresponding options will be displayed on the wristband screen (if equipped).

[0045] The senior citizen can cycle between the "Confirm Send" and "Cancel" options using the main button. Each time the main button is pressed, the bracelet simultaneously announces the currently selected item via voice. After selecting "Confirm Send," the senior citizen presses the touch button for final confirmation.

[0046] The moment the touch button is triggered, the security chip on the wristband responds with an interrupt, and internally retrieves the unique device private key stored in the chip to generate a digital signature for the operation information containing the following data fields: Operation ID: A unique UUID that identifies this operation. Operation type: Sending WeChat message Receiver ID: Encrypted contact identifier Timestamp: Unix timestamp (milliseconds) Anti-replay random numbers: Random numbers generated by the security chip The security chip generates a digital signature and confirmation command, which is then transmitted back to the private intelligent agent on the mobile phone via a wireless communication link.

[0047] Step 4: Signature Verification and Message Execution The mobile-side private agent receives the returned data and performs the following verification: Verify the validity of the digital signature using the public key on the wristband; Check if the timestamp is within the allowed time window (to prevent replay); Verify whether the random number has already been used (to prevent duplicate attacks); After successful verification, the private agent decrypts the corresponding WeChat long-term refresh token from the hardware-level encrypted sandbox, obtains a short-term access token through the official WeChat API, and calls the message sending interface to complete the message sending.

[0048] Step 5: Result Feedback and Evidence Storage The WeChat API returns the sending result, and the private intelligent agent transmits the result back to the wristband via a wireless communication link. The wristband then announces in voice: "WeChat has been sent to my son." At the same time, the system automatically packages the operation log, digital signature, timestamp and other data of this operation and uploads them to the blockchain evidence storage node to generate an unalterable audit record.

[0049] Cancellation procedure: If, in step 3 above, the elderly person selects the "Cancel" option via the main button and then presses the touch button to confirm, the wristband will only send back the cancellation command and will not generate a digital signature; the private intelligent agent on the mobile phone will terminate the process after receiving the cancellation command and may announce "Operation has been cancelled".

[0050] Example 2: Two-level hardware verification process for utility bill payment This embodiment describes the specific implementation process of utility bill payment using the system described in this invention, focusing on the cloud-triggered active triggering, dual confirmation of payment-sensitive operations, and blockchain evidence storage mechanism.

[0051] Scenario setting: An elderly person's water bill account has an outstanding amount of 200 yuan, and the system proactively initiates a payment reminder and payment process.

[0052] Step 1: Cloud-based detection and task distribution In the cloud-based public intelligent agent cluster, the public intelligent agent instance running in the worker container dedicated to the elderly person queries the water utility interface according to the preset scheduled task (10:00 every day). It detects that the elderly person's account number (pre-bound) is in arrears of 200 yuan, and the payment deadline is 3 days later.

[0053] The public intelligent agent determines that the elderly person needs to confirm the matter and generates a payment reminder request containing the following information: Item type: Utility payment reminder Payment item: Water fee Payment amount: 200.00 yuan Payment account number: Last four digits **** (anonymized) Action options: [Pay Now] / [Remind Me Later] / [Ignore] The request is sent to the private intelligent agent on the mobile device via an encrypted secure channel.

[0054] Step 2: Mobile phone forwarding and wristband broadcast The private intelligent agent on the mobile phone receives the request, determines that the operation involves payment (a sensitive operation), and forwards the request to the wristband through the wireless communication link.

[0055] The wristband's voice announcement reads: "There is a water bill of 200 yuan to be paid. The account number ends in 1234. Press the main button to select, and press the touch button to confirm." Simultaneously, the wristband screen (if equipped) displays the following options: ① Confirm Payment ② Remind Me Later ③ Ignore.

[0056] Step 3: Two-level hardware selection and confirmation Seniors can cycle through three options using the main button, and the wristband will simultaneously announce the currently selected item via voice ("Confirm Payment", "Remind Me Later" or "Ignore").

[0057] Scenario A: The elderly person selects and confirms payment. After selecting "Confirm Payment," the elderly person presses the touch button to confirm. The security chip on the wristband responds, generating a digital signature containing the following fields: Transaction ID: A unique transaction serial number generated by the system. Transaction type: Water bill payment Transaction amount: 200.00 yuan (in whole numbers or in cents) Payee Identifier: Water Company Code Timestamp: Current Unix timestamp Random number: Anti-replay attack The signature result and confirmation instruction are sent back to the mobile device.

[0058] Step 4: Signature Verification and Payment Execution After the private intelligent agent on the mobile device verifies the validity of the signature, it decrypts the payment account credentials (WeChat Pay / Alipay) bound to the elderly person from the encrypted sandbox and calls the unified order interface of the corresponding payment SDK to complete the deduction.

[0059] Step 5: Results Feedback and Notification to Multiple Parties The payment interface returns the result, and the private intelligent agent on the mobile phone transmits the result back to the wristband via a wireless communication link. The wristband then announces in voice: "Water fee of 200 yuan paid successfully." At the same time, the payment result will be synchronized to the cloud on the mobile device. The cloud-based public intelligent agent will update the payment status and can optionally push a notification to the guardian's child's app: "Your parents' water bill of 200 yuan has been paid." Step 6: Blockchain Evidence Storage The system automatically packages all data of this payment operation (including transaction ID, amount, timestamp, wristband digital signature, and payment channel return serial number), calculates the hash value, and uploads it to the blockchain evidence storage platform to generate an evidence storage number for subsequent auditing or dispute resolution.

[0060] Scenario B: The elderly person chooses "Remind me later" or "Ignore". If the elderly person selects "Remind me later" and confirms, the wristband sends a cancellation command (without a signature), and the mobile phone notifies the cloud to remind them again 2 hours later.

[0061] If the elderly person chooses to "ignore" and confirms, the wristband sends a cancellation command back, and the mobile phone notifies the cloud to stop reminding them of the bill.

[0062] Timeout protection: If the elderly person does not complete any selection or confirmation within 30 seconds, the bracelet will automatically announce "Operation timed out, cancelled" and terminate the process. The cloud will then handle the matter according to the default policy (such as reminding again the next day).

[0063] Example 3: Automatic Cancellation Mechanism for Timeout This embodiment describes the specific implementation of the automatic timeout cancellation function in the system of the present invention. This mechanism serves as a supplementary security guarantee for two-level hardware confirmation, preventing the operation process from stalling or being triggered unexpectedly.

[0064] Hardware configuration: The wristband has a built-in timer module that works in conjunction with the main button, touch button, and voice broadcast module.

[0065] Triggering scenario: The private intelligent agent on the mobile phone sends an operation confirmation request to the wristband. The wristband starts broadcasting a prompt and enters a waiting confirmation state.

[0066] Start-up timer: The moment the wristband announces "Press the main button to select, press the touch button to confirm," the built-in timer module starts, with a timeout threshold set to 30 seconds (this value can be configured by the guardian via the mobile app, and the adjustable range is 15-60 seconds).

[0067] Interaction during the waiting period: If the elderly person triggers the main button for the first time within 30 seconds, the timing module will reset and a new 30-second timer will begin (i.e., the timeout timer will be refreshed with each button press). If the elderly person confirms the action by touching the button, the timing module stops and proceeds to the next processing step. If the elderly person does not press any buttons within 30 seconds, or only presses a touch button but does not select anything (pressing a touch button alone is invalid; it is only effective when an item is selected), the timer will continue to run.

[0068] Timeout triggered: When the timer reaches the 30-second threshold, the wristband automatically triggers a timeout interrupt and performs the following operations: 1. Process Termination: Immediately exit the current operation confirmation process and clear any temporarily stored confirmation request data; 2. Voice prompt: The speech synthesis module is invoked to announce "Operation timed out, cancelled"; 3. Status Reset: Resets the status registers of the main buttons and touch keys, restoring them to standby mode; 4. Result feedback: Optionally, a "timeout cancellation" status code can be fed back to the private intelligent agent on the mobile device, which can facilitate the mobile device to update the local status or notify the cloud.

[0069] Mobile processing: When the private agent on the mobile device receives a timeout cancellation status code (or no feedback is received after the timeout), it releases the resources occupied by this operation, clears the temporary cached operation data, and optionally records an "operation timeout" log.

[0070] Cloud-based collaboration: If the operation is initiated by the cloud (such as the payment reminder in Example 2), after the cloud receives the timeout status synchronized by the mobile terminal, it will process it according to the preset retry strategy (such as pushing again after 2 hours, or reminding again the next day).

[0071] Technical effects: This mechanism effectively avoids the following situations: An elderly person accidentally activated the voice wake-up function but forgot to operate it, causing the bracelet to remain in a waiting state for an extended period. The wristband occupies the wireless communication link channel for an extended period, affecting other data transmissions. The private intelligent agent resources on the mobile device are being occupied for an extended period by invalid operation requests.

[0072] Example 4: Parallel design of hardware to prevent accidental touches and emergency help mechanism This embodiment describes the hardware structure design for preventing accidental touches on the wristband of the present invention, as well as the parallel function implementation of the main button: "short press to select, long press to ask for help".

[0073] I. Hardware Structure for Preventing Accidental Touch Button physical structure: The main button and touch button on the bracelet feature a recessed, sunken design. Specific parameters are as follows: The button surface is 1.5mm ± 0.2mm lower than the surface of the bracelet's outer shell. The button is surrounded by a raised protective ring with a height of 2.0mm. Button trigger force threshold: set to ≥2.0N (approximately 200 grams of force), achieved through spring damping of the internal microswitch. Structural effect: The recessed design makes the buttons less likely to be accidentally touched during daily activities (such as dressing, washing hands, turning over in bed); The trigger force requirement of ≥2N ensures that the button will only be triggered when the elderly person consciously applies a certain amount of force; a light touch or accidental touch will not have an effect. Tests have shown that this design can reduce the daily accidental touch rate to less than 0.1 times per person per day.

[0074] II. Main Button Dual-Function Parallel Mechanism Function definition: Short press mode (trigger time < 1 second): Performs the normal "option cycle selection" function. Long press mode (trigger time ≥ 3 seconds): Triggers the "Emergency Help" process. Parallel implementation principle: The main control chip of the wristband has a built-in button interrupt handling program, which uses a time-slice polling + state machine approach to achieve parallel recognition of dual functions. Key press interruption trigger -> Record press timestamp T_press Start a timer to check the button status every 100ms. while (the key is still pressed) { Current hold time T_hold = Current time - T_press if (T_hold ≥ 3000ms && Emergency Help not triggered) { Triggering emergency assistance process Emergency assistance has been triggered. break } } When the button is released: if (emergency assistance has been triggered) { / / Emergency assistance process has been initiated; no further action will be taken. } else if (T_hold<1000ms) { Use the "short press to select" function: toggle the currently selected item. } else if (T_hold ≥ 1000ms&&T_hold<3000ms) { Ignore (presses between 1-3 seconds are considered invalid to avoid misjudgment of the critical zone). } Emergency assistance process: When the main button is pressed and held for ≥3 seconds, the wristband immediately executes the following: 1. Local notification: A high-frequency beep will be emitted for 0.5 seconds. 2. Voice announcement: "Emergency assistance has been triggered; your children are being notified." 3. Location Acquisition: Activate GPS / cell tower positioning to obtain the current location. 4. Notification Sending: Sends emergency distress signals and location data to the mobile phone's private intelligent agent via wireless communication link. 5. Mobile Forwarding: The help request is pushed to pre-set emergency contacts (children's mobile app and SMS) via mobile network. 6. Cloud-based evidence storage: The time, location, and processing status of requests for assistance are recorded in the cloud. Functional independence verification: After a long press triggers an emergency call, releasing the press will not accidentally trigger the "short press selection" function. During normal selection operations (continuous short presses), even if the interval between two short presses is less than 3 seconds, the long press function will not be triggered accidentally. The two functions operate in parallel without interfering with each other.

[0075] III. Examples of Combined Application Scenarios Scenario 1: Button usage during normal payment process The elderly person received a payment notification (as in Example 2): Press the main button three times: switch from "Confirm Payment" to "Remind Me Later" to "Ignore", then switch back to "Confirm Payment". After selecting "Confirm Payment", press the touch button to confirm. The main button is only used for selection throughout the entire process; the long-press function is not activated. Scenario 2: Emergency help needed when suddenly feeling unwell An elderly person suddenly felt unwell at home and had no mobile phone with them: The elderly man found the bracelet and pressed and held the main button for more than 3 seconds. The bracelet immediately initiates the emergency assistance process without requiring any menu selection. Even if the bracelet is in standby or operation waiting mode, a long press will trigger emergency help first. Technical effect: This embodiment achieves the dual technical effects of "preventing accidental touches in daily operation" and "one-click access in emergency scenarios" through hardware structure design and software state machine collaboration, which not only ensures the reliability of daily use, but also ensures availability in emergency situations.

[0076] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the appended claims.

Claims

1. A monitoring system based on a hierarchical intelligent agent architecture and dual hardware verification, characterized in that, The system includes: Wearable hardware verification layer, configured in wristband devices, includes physical interaction units and security chips; The physical interaction unit includes a main button and a touch button. The main button is configured to cycle through operation options, switching between options each time it is triggered. The touch button is configured to confirm the currently selected operation option. The security chip has a built-in independent storage area and a computing unit for storing a unique, non-exportable device private key, and a corresponding device public key. In response to the final confirmation command triggered by the touch key, the security chip internally calls the private key to perform digital signature calculation on sensitive operation data, generates a signature result with non-repudiation, and sends it to the mobile terminal through the near-field wireless communication module. A private intelligent agent layer for mobile terminals, deployed on smartphones for the elderly, includes a background service module, an encrypted storage module, and a verification execution module; The background service module runs as a system background service process, and achieves seamless interaction and status monitoring of the target application through an automated interaction interface or an application programming interface (API). The encrypted storage module is a hardware-level encrypted sandbox used to isolate and store highly sensitive privacy data, including long-term tokens for third-party services and payment account credentials, and to set access barriers for all external process access requests. The verification execution module is equipped with a hardware confirmation trigger threshold. It continuously listens for operation commands issued by the wearable device and only authorizes the use of the corresponding credentials from the encrypted storage module to execute sensitive transactions after verifying that the command is accompanied by a valid digital signature signed by the security chip. The cloud-based public intelligent agent cluster is deployed using a Manager / Worker multi-tenant architecture. The Manager node is responsible for global task scheduling, tenant lifecycle management, and the elastic scaling of Worker nodes; The Worker container dynamically allocates an independent running instance to each registered senior citizen, encapsulating their exclusive public intelligent agent. This intelligent agent only has the permission to access de-identified data and call general computing resources, and is dedicated to handling non-sensitive computing tasks that do not require the combination of local privacy data. The processing results of the cloud-based public intelligent agent are encrypted or de-identified and then synchronized to the private intelligent agent layer of the mobile terminal for decision-making reference. The wearable device, the mobile terminal, and the cloud interact with each other through an encrypted communication link, forming a layered intelligent agent architecture with end-edge-cloud collaboration.

2. The system according to claim 1, characterized in that, The main button is also equipped with a voice feedback module, which simultaneously announces the currently selected item during the selection process; after the touch button is confirmed, a digital signature containing at least the operation identifier, transaction amount, timestamp and random number is generated upon receiving the confirmation instruction from the touch button, and then transmitted back to the mobile terminal via a wireless communication link.

3. The system according to claim 1, characterized in that, The main button is also equipped with an emergency trigger mode: when pressed for more than a preset time threshold, an emergency help process is triggered. This long press function is parallel and independent of the short press selection function. The main button and the touch button adopt a concave anti-accidental touch structure, and the trigger force threshold is set to ≥2N.

4. The system according to claim 1, characterized in that, The wristband also includes an automatic timeout cancellation module, configured to automatically cancel the current operation process and broadcast a prompt message if the option selection and confirmation operation is not completed within a preset time.

5. The system according to claim 1, characterized in that, The mobile terminal private intelligent agent layer also includes an initial authorization guidance module, configured to guide the elderly to complete the authorization operation of third-party services when the system is first activated, and store the obtained authorization certificate in the encrypted sandbox after encrypting it with the wristband's public key.

6. The system according to claim 1, characterized in that, The cloud-based public intelligent agent cluster achieves resource isolation based on containerization technology. Each Worker container has an independent file system, network space, and operating environment, realizing data isolation and session isolation between different users.

7. A two-level hardware verification and monitoring method based on a hierarchical intelligent agent architecture, applied to the system described in any one of claims 1 to 6, characterized in that, Includes the following steps: S1: Voice Acquisition and Transmission: The wristband collects the elderly's voice commands and transmits them in real time to the private smart agent on the mobile phone via a wireless communication link; S2: Intent Recognition and Task Decentralization: The private intelligent agent performs semantic parsing on the voice commands to identify the operation intent; it determines whether the operation belongs to a sensitive operation type. If it is a non-sensitive operation, it constructs a task request and forwards it to the cloud public intelligent agent for execution. If it is a sensitive operation, proceed to step S3; S3: Hardware Confirmation Request Generation and Issuance: The private intelligent agent generates an operation confirmation request based on the parsing results and sends it to the wristband via a wireless communication link; after receiving the request, the wristband prompts the elderly with the operation content through voice broadcast and enters the hardware confirmation waiting state; S4: Cyclic Selection and Voice Feedback: Seniors can cycle through operation options by triggering the main button. Each time the main button is triggered, the wristband will simultaneously broadcast the semantic description of the currently selected item. S5: Hardware Confirmation and Signature Generation: When the elderly select the desired operation option, the touch key is triggered for final confirmation; If the operation is to confirm the execution of a sensitive operation, the security chip in the wristband will call the private key stored in the chip to generate a digital signature for the sensitive operation data, which includes the operation identifier, timestamp, and random number, and send the signature result and confirmation instruction back to the mobile phone; if the operation is to cancel, only the cancellation instruction will be sent back. S6: Signature Verification and Credential Retrieval: The private intelligent agent on the mobile device receives the returned data and verifies the validity of the digital signature; after successful verification, it decrypts the corresponding authorization credential from the encrypted sandbox and calls the official application programming interface of the third-party service to complete the sensitive operation. S7: Result Feedback and Evidence Preservation: The operation execution result is transmitted back to the wristband via wireless communication link and fed back to the elderly in the form of voice broadcast; at the same time, the operation log and signature data of key operations are preserved and recorded.

8. The method according to claim 7, characterized in that, Sensitive operations in step S2 include, but are not limited to: sending instant messaging messages, initiating voice calls, authorizing payments, and confirming electronic signatures; non-sensitive operations include, but are not limited to: checking household bills, generating health reports, playing entertainment content, and checking the weather.

9. The method according to claim 7, characterized in that, In step S3, the wristband broadcast uses dialect speech synthesis technology, supporting no less than 40 dialects; the timeout period for the elderly to select and confirm after the broadcast is preset to 30 seconds, and the cancellation process is automatically triggered after the timeout.

10. The method according to claim 7, characterized in that, The digital signature generated in step S5 includes at least: a unique operation identifier, transaction amount, timestamp, anti-replay random number, and current wristband status parameters to prevent replay attacks and forged requests.

11. The method according to claim 7, characterized in that, The cloud-based public intelligent agent cluster adopts a Manager / Worker architecture, where the Manager node is responsible for task semantic decomposition and Worker resource allocation, and each Worker container runs independently; the Manager centrally manages the global API credentials of third-party services, and each Worker only holds a temporarily issued access token with a limited validity period.

12. The method according to claim 7, characterized in that, In step S7, the operation logs and signature data of key operations are packaged, hash values ​​are calculated, and uploaded to the blockchain evidence storage platform to generate tamper-proof audit logs.

13. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 7 to 11.

Images