Cross-platform data processing system and method for AI healthy ecology and medium

Abstract

The invention discloses an AI healthy ecology-oriented cross-platform data processing system and method and a storage medium, and belongs to the technical field of Internet of Things and artificial intelligence. The method comprises the following steps: an intelligent wearable device (A1) sends a standardized data packet to a terminal (B1) through a predefined GATT service; a terminal loads a Web application through a browser (B1), a JavaScript SDK (B2) and a WebAssembly module (D1) are used for carrying out analysis (C1), quality evaluation (C4) and safety processing (C5) on data, and an uplink data stream (E1) is formed and uploaded to a cloud end (F1) through WebSocket; the cloud side analyzes the AI model (H1) corresponding to the data packet type routing call (G1), generates a result (I1) and returns the result (I1) in a downlink mode; and the result is visually displayed through a browser or an SDK (D3) embedded in a third-party page (D4). Data islands are broken through a standardized protocol, cross-platform installation-free deployment is realized by utilizing browser capability, low-delay and high-reliability communication is guaranteed by means of WebSocket and edge calculation, a model is updated through federal learning (S1), privacy is protected, AI is rapidly endowed for wearable equipment, and an open AI healthy ecosystem is constructed.

Description

[0001] The present application relates to the technical fields of Internet of Things, artificial intelligence and mobile computing, and in particular to a system, method and storage medium for AI health ecology, supporting cross-platform access, data processing and AI services of heterogeneous intelligent wearable devices.

[0002] With the popularity of intelligent wearable devices (A1), the multi-modal health data generated thereby provides the possibility for AI health services. However, the current solutions have inherent defects:

[0003] 1. Data island problem: Different manufacturers and product categories of devices use private data protocols and communication interfaces, resulting in difficulties in unified access and analysis of data.

[0004] 2. Deployment and compatibility problems: Users (U1) need to download and install special Apps, which is a cumbersome process and poses cross-platform compatibility challenges, increasing the user usage threshold and manufacturer development costs.

[0005] 3. AI empowerment path complexity: The link from the device to the AI model (H1) involves multiple technical stacks such as embedded, mobile and cloud, making it difficult for small and medium-sized manufacturers to quickly integrate AI capabilities.

[0006] 4. Communication efficiency and real-time performance are insufficient: Traditional HTTP polling and other methods cannot meet the low-latency and high-reliability transmission requirements of real-time health monitoring.

[0007] 5. Data security and privacy concerns: Health data is sensitive, and how to ensure security and privacy during transmission and processing, and comply with increasingly stringent regulations, is a key challenge.

[0008] 6. Ecological closure: Existing systems are mostly closed ecosystems, making it difficult to open AI capabilities to third-party applications (D4) in a standardized and low-cost manner, limiting business model innovation.

[0009] Therefore, there is an urgent need in the art for a new technical solution that can break through the above bottlenecks and achieve integration of device access, data processing, AI analysis and ecological collaboration.

[0010] The present application relates to the technical fields of Internet of Things, artificial intelligence and mobile computing, and in particular to a system, method and storage medium for AI health ecology, supporting cross-platform access, data processing and AI services of heterogeneous intelligent wearable devices.

[0011] To achieve the above-mentioned purposes, the present application adopts the following technical solutions:

[0012] ​​​In a first aspect, the present application provides a cross-platform data processing method for an AI health ecosystem, comprising the following steps:

[0013] A device data standardization step: the smart wearable device (A1) sends a plurality of standardized data packets with a unique preset packet header identifier (P1) to the terminal device (B1) through a predefined set of Bluetooth GATT services; the smart wearable device (A1) also reports its device type code to the terminal device (B1); a terminal web processing step: the terminal device (B1) loads a web application embedded with a JavaScript SDK (B2) by accessing a uniform resource locator URL; the terminal device (B1) receives the standardized data packets through a Web Bluetooth API in the browser environment (B1), calls a WebAssembly module (D1) to parse (C1), verify, quality assess (C4), and securely process (C5) the received binary data packets, and forms an uplink data stream (E1) in a unified format; and then uploads the uplink data stream (E1) to a cloud service platform (F1) through an established WebSocket long connection;

[0014] A cloud AI routing and analysis step: the cloud service platform (F1) receives the uplink data stream (E1), parses the data packet type and device type, and routes and calls (G1) the corresponding preset AI analysis model (H1) according to the parsing result, analyzes and processes the uplink data stream (E1), and generates processing result data (I1); a result visualization and ecosystem interaction step: the cloud service platform (F1) returns the processing result data (I1) to the terminal device (B1) through a downlink communication channel, and the terminal device (B1) visualizes the data through a browser interface; and / or, embeds the processing result into a third-party webpage or application (D4) for display through a preset Web SDK (D3).

[0015] Preferably, the device data standardization step comprises:

[0016] The predefined set of GATT services includes multi-modal data services, device control services, and device status services; the smart wearable device (A1) switches the working mode and data collection frequency based on the device type code according to terminal instructions or built-in strategies.

[0017] Preferably, the terminal web processing step comprises:

[0018] The WebAssembly module (D1) performs real-time filtering, motion artifact identification, and signal quality assessment (C4) on the data packets, and adds a quality flag bit to the data packets; and / or,

[0019] The WebAssembly module (D1) desensitizes sensitive information in the data packet and performs end-to-end encryption (C5) on the data.

[0020] Preferably, the method further comprises a federated learning model updating step (S1):

[0021] The WebAssembly module (D1) of the terminal device (B1) calculates the update gradient (C6) of the AI model (H1) based on local data;

[0022] The encrypted update gradient (C7) is uploaded to the cloud through the WebSocket long connection;

[0023] The cloud service platform (F1) aggregates the update gradients uploaded by multiple terminals and performs incremental updating (C10) on the preset AI analysis model (H1).

[0024] Preferably, the method further comprises a connection and cache management step (S2):

[0025] When the connection between the terminal device (B1) and the cloud network (F1) is interrupted (N1), the smart wearable device (A1) caches the standardized data packet locally (A4);

[0026] When the terminal device (B1) detects network recovery, it automatically pulls the cached data packet from the smart wearable device (A1) through the Web Bluetooth API and continues transmission (B2).

[0027] Preferably, the result visualization and ecological interaction step comprises:

[0028] After the JavaScript SDK (B2) receives the processing result (I1), it performs UI adaptive rendering on the display interface according to the configuration information of the third-party web page (D4);

[0029] The processing result (I1) and the user's (U2) identity on the third-party website are bound and stored. In the second aspect, the application provides a cross-platform data processing system for an AI health ecosystem, which comprises:

[0030] The smart wearable device module (MD1) is used to generate and send standardized data packets with unique packet header identifiers (P1) through a predefined GATT service set and report device type codes;

[0031] A terminal processing module (MD2) is configured to receive data through a Web Bluetooth API, parse (C1), verify, quality assess (C4), securely process (C5), and encapsulate (C3) through a WebAssembly module (D1), and upload the data to the cloud through a WebSocket long connection;

[0032] A cloud service module (MD3) is configured to receive an uplink data stream (E1), route and call (G1) a corresponding AI model (H1) for analysis according to a data packet type and a device type, and return a result through a downlink channel; and an ecological interaction module (MD4) is configured to visually display and interact with the processing result (I1) in a terminal browser or a third-party webpage (D4).

[0033] In a third aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, when the program is executed by a processor, the method of the first aspect is implemented.

[0034] The present application has the following advantages:

[0035] 1. Ecological access: through standardized GATT services and data packet formats (S3), the access barriers of heterogeneous wearable devices (A1) are broken down, laying a technical foundation for building a "VitalPrint Enabled" open ecosystem.

[0036] 2. Cross-platform and installation-free: taking full advantage of the capabilities of modern browsers (B1), users (U1) can use complete functions without downloading and installing Apps, greatly improving user experience and reducing the access cost of ecological partners.

[0037] 3. End-to-cloud intelligent collaboration: through the architecture of "browser edge processing (WASM, D1) + cloud AI routing (G1)", data quality control (C4), low-latency communication, and flexible scheduling of AI capabilities (H1) are realized, balancing real-time performance and computational complexity.

[0038] 4. Security and privacy protection: through data desensitization and encryption (C5) before transmission, and the introduction of a federated learning mechanism (S1), user privacy is strictly protected while the value of data is fully realized, meeting compliance requirements.

[0039] 5. Business model innovation: through the SDK (D3) design that can be embedded in third-party pages (D4), the core AI capabilities (H1) are output as a service (AIaaS), creating a new channel for monetizing health data value.

[0040] 6. System robustness: Through the device-side cache (A4) and terminal resuming transmission (B2) mechanism (S2), effectively deal with network unstable scenarios, ensure data integrity, and improve system reliability. BRIEF DESCRIPTION OF DRAWINGS

[0041] Figure 1 is the overall method flowchart of the present application, which shows the complete process from device data collection to result display, including device data standardization (A1), terminal Web processing (B1, D1), cloud AI routing and analysis (F1, G1, H1), result visualization and ecological interaction (U1, U2, D4) and other core steps, as well as federated learning update (S1) and connection cache management (S2) two enhanced processes.

[0042] Figure 2 is the system module schematic diagram of the present application, which clearly depicts the four core modules: intelligent wearable device module (MD1), terminal processing module (MD2), cloud service module (MD3) and ecological interaction module (MD4) and the data interaction relationship between their internal units (such as A1, B2, D1, F1, H1, etc.).

[0043] Figure 3 is the data packet structure and processing flow schematic diagram, which specifically explains the structure of the standardized data packet (S3, P1-P4) and the detailed process of processing (C1-C5) by the WebAssembly module (D1) on the terminal side and finally encapsulating as the uplink data stream (E1).

[0044] Figure 4 is the federated learning model update step flowchart, which details the collaborative mechanism of incrementally updating (C10) the global AI model (H1) through terminal local calculation (S4, C6, C7) and cloud aggregation (C8, C9) under the premise of protecting data privacy.

[0045] The above drawings together constitute a complete, multi-level visual description of the technical solution of the present application. DETAILED DESCRIPTION

[0046] The technical solution of the present application will be further described in detail below by specific examples combined with the drawings. It should be understood that the specific examples described herein are only used to explain the present application and do not limit the present application.

[0047] Example 1: ECG monitoring and analysis ecological service based on intelligent ring

[0048] This example takes a smart ring (A1) that complies with the "VitalPrint Enabled" specification, a terminal (B1) that is a smartphone browser, and a cloud (F1) that is a VitalPrint Core platform integrated with multiple AI models (H1).

[0049] 1. Device data standardization

[0050] • The ring firmware (A2) loads predefined GATT services, with a multi-modal data service UUID of A3D2F000-50CE-11EE-BE56-0242AC120002.

[0051] • The ring sensor array (A1) collects heart rate, blood oxygen, and single-lead ECG data, encapsulated as:

[0052] o Basic monitoring package (package header (P1) 0x55AA01): contains heart rate, blood oxygen, and acceleration data.

[0053] o ECG data package (package header (P1) 0x55AA05): sampling rate of 125 Hz, containing 80 bytes of waveform data (P3).

[0054] • The ring reports device type code (e.g., 0x01 for "Ring - Standard Edition") through the device status service.

[0055] 2. Terminal web processing

[0056] • The user (U1) accesses the URL:

[0057] https: / / health.vitalprint.top / analyzer in the phone's Chrome browser (B1), loading the web application and JavaScript SDK (B2).

[0058] • The SDK (B2) discovers and connects to the ring through the navigator.bluetooth API, subscribing to data characteristic values. • The browser (B1) receives binary data packets and hands them over to the preloaded WASM module (D1).

[0059] This module (D1):

[0060] o Parses (C1) the package header (P1) and identifies it as ECG data.

[0061] o Performs quality assessment (C4): real-time baseline drift correction and calculation of signal quality index (SQI).

[0062] If the SQI is below the threshold, the packet is discarded or a "low quality" flag is added to the data stream.

[0063] o Security processing (C5): Hashing sensitive information such as device serial number.

[0064] o Packaging (C3) for unified JSON uplink data stream (E1): {deviceId: "hash_id",

[0065] packetType: "ECG", timestamp: 1672531200000, data: {...}, sqi: 0.85}. The JSON data stream (E1) is uploaded to the cloud (F1) in real time by the WebSocket long connection established by the SDK (B2) through the uplink communication unit (B5).

[0066] 3. Cloud AI routing and analysis

[0067] · The WebSocket server (F1) in the cloud (F1) receives the data and parses the packetType and deviceType.

[0068] · The data routing unit (C2) routes the call (G1) to the ECG analysis model (H1) optimized for single-lead rings through the AI model scheduling unit (C3) based on the packetType: "ECG" and deviceType: "Ring - Standard Edition".

[0069] · The AI model (H1) generates processed result data (I1) after analysis: {hr: 72, diagnosis: "Normal sinus rhythm", hrv: 45, alert: false}.

[0070] 4. Result visualization and ecological interaction

[0071] · The cloud (F1) pushes the analysis result (I1) to the browser (B1) through the original WebSocket downlink through the downlink communication unit (C5).

[0072] · Scenario A (self-owned platform): The browser (B1) renders the conversational interface through the visualization rendering engine (D2): "Your electrocardiogram analysis is complete, heart rate 72 beats per minute, everything is normal."

[0073] · Scenario B (ecological partners): A health blog website www.lumenring.com (D4) embeds the same WebSDK (D3). After the user (U2) logs in to the website, the SDK (D3) initializes and passes in the user ID blog_user_123. The analysis result (I1) is displayed in the form of a popup in the blog page (D4) and is bound to the blog_user_123 account, stored in the user's health profile on the blog website.

[0074] 5. Federated learning model update (periodically executed, S1)

[0075] • The terminal WASM module (D1) calculates the model update gradient (C6) using recent local ECG data. • The gradient data is encrypted (C7) and uploaded to the cloud when the system is idle.

[0076] • The cloud federated learning aggregator (C8) aggregates tens of thousands of such gradients, updates the global model (C10) with the aggregated gradient, and iteratively updates the ECG analysis model (H1) without accessing the original data, benefiting all users in the ecosystem.

[0077] The parts of the invention not described in detail are known in the art.

[0078] Finally, it should be noted that: the above examples are only used to illustrate the technical solutions of the present application, and not to limit it. Although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A cross-platform data processing method for an AI-driven health ecosystem, characterized in that, Includes the following steps: Device data standardization steps: The smart wearable device (A1) sends a variety of standardized data packets with unique preset packet header identifiers (P1) to the terminal device (B1) through a predefined set of Bluetooth Low Energy GATT services; Terminal Web-based processing steps: The terminal device (B1) loads a web application with an embedded JavaScript SDK (B2) by accessing a unified resource locator URL; it receives the standardized data packets through the Web Bluetooth API in the browser environment, calls the WebAssembly module (D1) to parse, verify, and encapsulate the received binary data packets to form a unified format uplink data stream (E1); and then uploads the uplink data stream (E1) to the cloud service platform (F1) through an established WebSocket long connection. Cloud AI routing and analysis steps: The cloud service platform (F1) receives the uplink data stream (E1), parses the data packet type, and routes and calls the corresponding preset AI analysis model (H1) according to the data packet type to analyze and process the uplink data stream (E1) and generate processing result data (I1); Results visualization and ecosystem interaction steps: The cloud service platform (F1) transmits the processing result data (I1) back to the terminal device (B1) through the downlink communication channel, and the terminal device (B1) displays it through a browser interface; and / or, the processing result data (I1) is embedded into a third-party webpage or application (D4) for display through a pre-built Web SDK (D3).

2. The method according to claim 1, characterized in that, The device data standardization steps specifically include: The firmware (A2) of the smart wearable device (A1) loads the predefined Bluetooth Low Energy GATT service set, which includes at least one multimodal data service with a unique identifier (UUID). The smart wearable device (A1) collects data through a sensor array (A1) and encapsulates it into standardized data packets according to a preset format. The standardized data packets include basic monitoring data packets and event marker data packets. Each type of data packet carries the unique preset packet header identifier (P1) in its header. After the smart wearable device (A1) establishes a low-power Bluetooth connection with the terminal device (B1), it actively pushes the standardized data packets to the terminal device (B1) in parallel using the feature values ​​of the multimodal data service.

3. The method according to claim 1, characterized in that, The terminal web-based processing steps specifically include: The WebAssembly module (D1) parses the data packets (C1) to determine the data packet type; The data packets are validated, including format integrity checks and data validity checks. Perform real-time signal quality assessment (C4) on data packets and add quality flags to the data packets; And / or, perform desensitization and end-to-end encryption on sensitive information in the data packet (C5); After verification and processing, the binary data is encapsulated into the upstream data stream (E1) in a unified JSON format.

4. The method according to claim 1, characterized in that, The method also includes a federated learning model update step: The WebAssembly module (D1) of the terminal device (B1) calculates the update gradient (C6) of the preset AI analysis model (H1) based on local data; The encrypted update gradient (C7) is uploaded to the cloud service platform (F1) via the WebSocket long connection; The cloud service platform (F1) aggregates update gradients uploaded by multiple terminals through a federated learning aggregator (C8) and performs incremental updates (C10) on the pre-built AI analysis model (H1).

5. The method according to claim 1, characterized in that, The method also includes connection and cache management steps: When the network connection between the terminal device (B1) and the cloud service platform (F1) is interrupted, the smart wearable device (A1) caches the standardized data packets in the local data cache area (A4); When the terminal device (B1) detects that the network has been restored, it automatically retrieves the cached data packets from the smart wearable device (A1) via the Web Bluetooth API and resumes the transmission.

6. The method according to claim 1, characterized in that, In the results visualization and ecological interaction steps, embedding the processing result data (I1) into a third-party webpage or application (D4) using the Web SDK (D3) for display specifically includes: After receiving the processing result data (I1), the JavaScript SDK (B2) performs UI adaptive rendering of the display interface according to the configuration information of the third-party webpage (D4). The processing result data (I1) is bound and stored with the user's (U2) identity identifier on the third-party website.

7. A cross-platform data processing system for an AI-driven health ecosystem, used to implement the method described in any one of claims 1 to 6, characterized in that, include: The Smart Wearable Device Module (MD1) is used to generate and send standardized data packets with a unique header identifier (P1) through a predefined set of GATT services. The terminal processing module (MD2) is used to receive data through the Web Bluetooth API, perform parsing, verification, quality assessment, security processing and encapsulation through the WebAssembly module (D1), and upload the data to the cloud through a WebSocket long connection; The cloud service module (MD3) is used to receive uplink data streams (E1), route and call the corresponding AI model (H1) according to the data packet type for analysis, and return the results through the downlink channel; The Ecosystem Interaction Module (MD4) is used to visualize and interact with the processing results (I1) in a terminal browser or a third-party webpage (D4).

8. The system according to claim 7, characterized in that: The WebAssembly module (D1) in the terminal processing module (MD2) is further used to perform real-time signal quality assessment (C4) on data packets and add quality flag bits, and / or to perform security processing such as data desensitization and encryption (C5). And / or, the cloud service module (MD3) also includes a federated learning aggregator (C8) for aggregating encrypted model gradients uploaded from multiple terminals to incrementally update the AI ​​model (H1).

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the cross-platform data processing method for the AI ​​health ecosystem as described in any one of claims 1 to 6.

Images