An all-unattended hospital AI registration system

Abstract

The application discloses a full-unattended hospital AI registration system, which is used for the unified interaction of an in-hospital self-service terminal (self-service machine) and a patient mobile terminal (including WeChat / alipay applet and App). The system takes multi-modal identity verification and natural language understanding as an entrance, and penetrates a closed-loop process of registration-approval-list identification-charging-one-key payment-task distribution-navigation / calling number. In the background, state machines / workflow are used to arrange each module, and automatic degradation and compensation can be realized in the case of abnormality. The innovation lies in that the structured analysis of the "clinic electronic list" is coupled with real-time load sensing into the task distribution decision, and the in-transit navigation and reminding of the mobile terminal and the self-service machine, and even the extendable AR glasses are combined, so that the patient can directly reach the target window / department according to the shortest path after completing payment, and the intermediate queuing and window changing are avoided.

Description

Technical Field

[0001] This invention relates to the fields of medical informatics and artificial intelligence, and in particular to an unattended registration and post-diagnosis task distribution system suitable for hospital settings, with multimodal AI and process orchestration as its core. Background Technology

[0002] Traditional medical processes commonly face long-standing pain points such as multiple queues (registration, payment, receipt collection, medication / examination collection), errors from manual data entry, fragmented information flow, and uneven utilization of equipment / windows. Existing systems mostly achieve partial automation in appointment or triage, failing to deeply integrate approval, billing, and post-treatment task distribution. Furthermore, they remain inadequate in multimodal identity verification, security and privacy, and emergency response.

[0003] While existing technologies have made attempts at intelligent guidance and triage, they have not formed an end-to-end closed loop of "consultation room list → instant distribution → on-the-way navigation". Patients still need to queue twice at the cashier or manual window, and the overall efficiency improvement is limited. Summary of the Invention

[0004] 4.1 Purpose of the Invention This paper proposes an end-to-end unattended system solution: from the patient initiating the order at a self-service machine or mobile device, to the doctor completing the order, the system automatically calculating the bill, and the mobile / self-service machine making one-click payment, to the intelligent distribution of the bill items to the most appropriate window / department and the instant generation of on-the-go navigation / call number, the entire process requires no manual intervention or repeated queuing. Furthermore, it can automatically degrade and record audits in case of anomalies.

[0005] A fully unattended hospital AI-based registration and post-diagnosis task distribution system is deployed on in-hospital self-service terminals and patient mobile devices. The system includes a processor, a memory, and a program running on the memory. The system includes at least: a) A multimodal identity recognition and verification module, used for fusion verification based on two or more of the following: optical document recognition, near-field card reading, and facial recognition.

[0006] b) Natural Language Understanding and Human-Computer Interaction Module, used to receive voice or text input and convert dialogue content into structured fields in intent-slot format.

[0007] c) The registration and approval integrated processing module is used to output structured approval results and bind them to the billing context.

[0008] d) Billing and billing module, used to receive and parse the electronic bill of the clinic or its image, and generate a structured bill of charges for settlement.

[0009] e) Payment and settlement module, used to complete payments and generate electronic or paper vouchers through a third-party mobile payment platform at self-service terminals or mobile devices.

[0010] f) The list receiving and task distribution module is used to jointly evaluate candidate execution points based on list type, patient priority, real-time load of the target window or device, in-hospital spatial location, and opening hours after successful payment, determine the target execution point, and output the call number and medical route. It also performs dynamic redistribution and synchronously updates the call number and route when the status changes.

[0011] g) Process control and emergency handling module, used to orchestrate the above modules in the form of a state machine or workflow, enter degrade mode when the network or equipment is abnormal, and perform compensation processing to restore the accounting and task status after recovery.

[0012] The list receiving and task distribution module uses the list parsing result as a cross-module association key and couples it with the approval, billing and payment results to achieve real-time distribution triggered by payment and linkage with on-the-go navigation.

[0013] A method for unattended registration and post-treatment task distribution in hospital patient flow includes: A) Receive and parse the electronic list of clinics to obtain structured list items.

[0014] B) Bind the structured results of registration and approval to the list items to form a unified data context.

[0015] C) Generate a bill of expenses and trigger the task distribution process after the patient completes payment through a self-service terminal or mobile device on a third-party mobile payment platform.

[0016] D) Based on the list type, patient priority, real-time load of the target window or device, in-hospital spatial location and opening time, the candidate execution points are jointly evaluated to determine the target execution points and output the call number and medical treatment path.

[0017] E) When congestion, downtime, or patient path deviation is detected at the target execution point, dynamic redistribution is performed and call numbers and paths are updated synchronously.

[0018] F) Arrange steps A) to E) using a state machine or workflow, enter a degraded mode in case of an anomaly, and perform compensation processing after recovery to restore the accounting and task status.

[0019] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method as described in claim 2.

[0020] The multimodal identity recognition and verification module integrates the results of optical document recognition, near-field card reading and facial recognition. When the confidence of any channel is insufficient, it triggers an alternative channel or manual backup. The identity verification result is written into a unified data context for subsequent billing and distribution reference.

[0021] The natural language understanding and human-computer interaction module automatically extracts the dialogue between the patient and the system into a structured form draft of intent-slot. The slot includes at least the department, examination / test items, prescription points, allergy history and insurance type. After the patient confirms, it is written into the electronic health record and participates in billing and distribution decisions.

[0022] When the payment and settlement module receives a successful payment receipt from a third-party mobile payment platform, it directly triggers the list receiving and task distribution module to generate the target execution point, call number, and medical treatment path, without requiring the patient to confirm again at a manual window, thus forming an instant closed loop between payment and execution.

[0023] The list receiving and task distribution module considers complementary features such as queue length, equipment capacity matching, intra-hospital spatial distance, opening time period adaptation, and patient priority when determining the target execution point. Upon detecting congestion, downtime, or patient deviation from the path, it triggers dynamic redistribution and synchronizes new call numbers and updated paths to self-service terminals and mobile devices.

[0024] The process control and emergency handling module outputs downgrade slips and manual instructions in case of anomalies, and completes the accounting, distribution and receipt status through compensation transactions after the network or equipment is restored, so as to ensure process consistency and traceability.

[0025] The system provides unified interaction and voucher display for both in-hospital self-service terminals and patient mobile devices. After either terminal completes payment or receives the distribution result, the other terminal automatically synchronizes the call status, medical route, and voucher information.

[0026] (i) The lists, prescriptions, and examination requests are preferably exchanged using an HL7 / FHIR compatible data structure. And / or (ii) Further includes an augmented reality glasses interaction interface for displaying the medical route and call progress in a head-up display manner, and converting voice input into structured text to update electronic health records and task distribution context.

[0027] 4.2 Overview of Technical Solution (Modular and Integrated) The system adopts an overall architecture of "ten modules + orchestration engine" (only the core features are listed below. See the implementation method for details): Multimodal identity recognition and verification: OCR / NFC / face fusion verification, unified strategy for mobile terminals and self-service machines. Automatically triggers a fallback process in case of failure.

[0028] Natural Language Understanding and Human-Computer Interaction: Dual-channel (voice / text). Dialect / Accent Adaptation. Semantic Parsing and Multi-turn Interaction for Medical Intent.

[0029] Integrated registration and approval process: Connects with hospital policies / medical insurance rules, outputs structured approval results, and links with billing.

[0030] Patient self-service information entry: touch screen + voice, real-time EHR write-back (preferably compatible with HL7 FHIR), and completeness / consistency verification.

[0031] List Recognition and Billing: Perform OCR / structured parsing on the doctor's list, combine the approval results to automatically calculate billing and generate a cost list.

[0032] Payment and Settlement (Mobile / Self-service Machine Dual Stack): Supports multiple payment methods including WeChat Pay, Alipay, and bank cards, with features such as "Confirmation → One-click Payment → Voucher Push / Print".

[0033] Hardware integration and automation: Unified driver management and self-testing for cameras / microphones / scanners / card readers / payment and receipt devices / printers, with a multimodal interactive interface.

[0034] Process control and emergency handling: state machine / workflow scheduling, log monitoring, and automated degradation. Supports compensatory transactions and manual fallback.

[0035] Multifunctional processing: One-stop processing for special diseases, chronic diseases, referrals and examination appointments, etc.

[0036] List reception and task distribution (core): Receives electronic lists from clinics (preferably HL7 / FHIR), performs intelligent distribution and dynamic rerouting based on features such as list type, patient priority, real-time window / equipment load, hospital space topology and opening hours, and outputs navigation and call numbers on mobile devices / self-service machines.

[0037] 4.3 Non-obviousness (creativity element) Cross-process "data-decision-execution" closed-loop coupling: deeply coupling list parsing and real-time load to the instant distribution after billing and payment are completed, and linking with queuing / route navigation, which is different from conventional systems that only do triage or static scheduling.

[0038] The unified human-machine collaboration between self-service machines and mobile terminals allows for seamless route guidance, electronic vouchers, and on-the-go reminders on the other terminal after payment is completed on either device. This avoids the need for secondary queuing and duplicate registration caused by fragmented multi-terminal systems.

[0039] Integrated orchestration and emergency response: Using finite state machines / workflows to connect states such as lists, billing, payment, distribution, and receipts, abnormal situations will enter a degraded mode (such as only printing receipts and guiding manual intervention), and automatic compensation will be provided after recovery. Compared with "functional module co-location", it has unpredictable continuity and recoverability. Attached Figure Description

[0040] Figure 1 System overall architecture (mind map).

[0041] A Tree-like Mind Map Explaining a Fully Unmanned Hospital AI Registration System System Overall Architecture and Deployment This fully unmanned hospital AI registration system is deployed on in-hospital self-service terminals and patient mobile devices, encompassing processors, memory, and the programs running on them, building a comprehensive and flexible registration service architecture.

[0042] This deployment method allows patients to operate through self-service terminals at the hospital or complete registration and other related processes anytime, anywhere using mobile devices, greatly improving the convenience of medical treatment.

[0043] Multimodal identity recognition and verification module This module is a crucial component for secure access control within the system. It integrates optical document recognition, near-field card reading, and facial recognition for verification. Optical document recognition quickly retrieves basic information from the document; near-field card reading accurately reads detailed identity data stored in the chip; and facial recognition confirms identity through biometrics. This fusion of three methods, combined with liveness detection, effectively prevents identity theft and other problems.

[0044] When any channel lacks sufficient confidence, an alternative channel or manual backup will be triggered to ensure the accuracy and reliability of identity verification. The verification results will also be written into a unified data context for subsequent reference.

[0045] Natural Language Understanding and Human-Computer Interaction Module This module enables natural and fluent communication between patients and the system. It can receive voice or text input, convert speech into text through speech recognition, and then accurately understand the patient's needs using intent recognition and slot filling technology, mapping the dialogue results into a structured form.

[0046] For example, if a patient says, "I want to register for an internal medicine appointment tomorrow morning," the system can recognize the intention to "register" and fill in the slot information such as "Time: Tomorrow morning" and "Department: Internal Medicine," providing an accurate basis for the subsequent registration process.

[0047] Integrated Registration and Approval Module This module tightly integrates the registration and approval process. It outputs structured approval results and binds them to the billing context. This means that while completing the registration approval, the system automatically associates relevant fee information, preparing for the subsequent billing process, improving the efficiency and accuracy of the entire registration process, and avoiding duplicate data entry and errors.

[0048] List Identification and Billing Module This module is responsible for parsing electronic medical records or their images. For electronic records, they are mapped to task items and fee items using a standardized interface; for paper records, key fields are extracted using technologies such as text detection and recognition, layout analysis, and table structure recovery.

[0049] Meanwhile, term alignment is achieved through named entity recognition and relation extraction, and code normalization is achieved by combining domain knowledge graphs when necessary, ultimately generating a structured cost list for settlement.

[0050] Payment and Settlement Module After a patient completes payment through a third-party mobile payment platform at a self-service terminal or on a mobile device, this module receives a payment success receipt and converts it into a unified business event. This conversion allows the system to promptly detect the payment status, providing trigger signals for subsequent task distribution and other processes, ensuring the smooth operation of the entire medical treatment process.

[0051] Task distribution and path generation module This is the core module of the system for optimizing the patient's medical experience. After receiving the business event corresponding to a successful payment, it determines the target execution point based on a comprehensive cost model, taking into account multiple factors such as the queue status of candidate execution points, equipment capacity, hospital topology, time schedule, and patient attributes. It then generates a medical path on the indoor topology map that takes into account capacity and time constraints.

[0052] For example, priority will be given to selecting execution points with fewer people in the queue, high equipment compatibility, and close proximity to the patient's current location, so as to plan the optimal medical route for the patient.

[0053] Dynamic redistribution module When congestion, equipment unavailability, schedule changes, or patients deviating from their designated routes are detected, this module triggers secondary distribution and route updates. Simultaneously, a suppression strategy is employed to avoid frequent switching caused by short-term fluctuations, ensuring the stability and rationality of the medical route. Update results remain consistent across mobile devices, self-service terminals, and wearable devices, and are communicated and confirmed in a non-intrusive manner.

[0054] Unified messaging and consistency module, cross-platform synchronization and seamless integration module, and optional wearable extension module. The unified messaging and consistency module uses message middleware and idempotent control to transmit and deduplicate events, and employs event sourcing and compensation transactions to ensure consistency and traceability of distribution, queuing, path and accounting status.

[0055] The cross-platform synchronization and seamless connection module enables self-service terminals and mobile devices to share a unified context. Once an operation is completed on one device, the other device automatically synchronizes the relevant information. The optional wearable extension module can display navigation and task prompts on wearable devices, maintaining consistency with the system context, further enhancing the patient's medical experience. Detailed Implementation

[0056] 6.1 Front-end design and unified interaction 6.1.1 Self-service machine (unattended) The input method is multimodal: card reading (NFC / second-generation ID card), ticket / list scanning OCR, camera facial recognition, and simultaneous touch screen and voice input. If the input fails, it automatically prompts for an alternative channel (e.g., ID card + facial recognition → ID card + SMS verification).

[0057] Payment integration: After the fee is confirmed, payment can be completed on-site by scanning a code / swiping a card / facial recognition + binding a card, and an electronic voucher can be printed or sent to the system.

[0058] 6.1.2 Mobile devices (WeChat / Alipay / APP) Direct connection via mini-program / SDK: Complete fee confirmation → One-click payment → Electronic voucher / reimbursement details → Automatically redirect to the "My Medical Treatment Path" tab, displaying the target window and call progress. If the patient first starts the process at the self-service machine, the mobile device can scan the code to continue.

[0059] Message notification: If a redistribution occurs during transit (such as when the device is temporarily out of service), the mobile app will immediately display an updated path and new call number to prevent patients from backtracking.

[0060] 6.1.3 Future-Oriented AR Glasses Expansion The system displays the patient's medical route and call progress, and provides spatial anchor points to indicate "Next step to window ××" as the patient moves within the hospital.

[0061] Voice-to-text conversion → structured forms: The voice conversation between the patient and the system is automatically extracted into structured items (such as symptoms / allergy history / past medical history) and written directly into the EHR form draft, reducing the time spent inputting on self-service machines / mobile phones.

[0062] Gestures / gaze confirmations replace touchscreen clicks, maintaining a contactless experience.

[0063] (The AR form follows the core of this invention: multimodal understanding + distribution / navigation closed loop. The key technical points remain unchanged, only the interaction end is expanded.) 6.2 Capability Service Layer 6.2.1 Multimodal Identity Recognition and Verification OCR + NFC + Face fusion: Passage is granted if any two or more of these methods pass. Secondary verification or manual verification is triggered if confidence is insufficient. The same strategy applies to mobile devices (OCR for document capture / liveness detection + facial comparison).

[0064] Data minimization and logging: Only fields necessary for decision-making are saved, and key operations are written to the audit log to meet internal compliance requirements.

[0065] 6.2.2 Natural Language Understanding and Human-Computer Interaction Semantic parsing for the medical intent domain: Supports intents such as "I want a follow-up appointment / prescription / CT scan / reimbursement rate / how to get there". Robust to dialect / noisy environments.

[0066] Dialogue as Form: Convert voice conversations into structured fields (department, project, allergy history, insurance type, etc.) and write them into the EHR and billing context after user confirmation, reducing repetitive input.

[0067] 6.2.3 Integrated registration and approval Unified data structure: Registration results and approval fields (catalog, percentage, limit) are written back to the billing context and synchronized to the mobile / self-service machine interface. Failures automatically revert to alternative solutions (self-pay / rescheduling, etc.).

[0068] 6.2.4 Inventory Identification and Billing List parsing: Receive electronic lists (preferably HL7 / FHIR), or OCR paper lists. Extract item codes / quantities / usages, and combine them with approval information to form a structured expense list.

[0069] Billing rules: Without changing the hospital's internal settlement logic, automatically combine reimbursement / self-paid items and generate payable bills.

[0070] 6.2.5 Payment and Settlement Mobile-first approach + self-service machine complement: Supports payments via WeChat, Alipay, and bank cards. Upon payment completion, electronic / paper vouchers and reimbursement details are generated, triggering task distribution and path generation.

[0071] 6.2.6 List Receiving and Task Distribution (Core Algorithm Perspective) Inputs include: structured items in the list, approval / billing results, real-time load of windows / devices, spatial location and open hours, etc.

[0072] Strategy: Calculate the overall cost of the candidate window for each item on the list (considering queue length, equipment matching, spatial distance, time period adaptation, patient priority, etc.), and select the target with the better cost.

[0073] Dynamic rerouting: When the target window fails / becomes congested or the patient deviates from the path, the system automatically recalculates and synchronizes the new call number / path to the mobile terminal and self-service machine.

[0074] Output: The combined result of "shortest medical treatment path + queuing progress + voucher QR code".

[0075] This module integrates list parsing, payment settlement, and in-transit navigation, forming a coupled closed loop that differs from conventional triage.

[0076] 6.3 Arrangement and Emergency Response The state machine / workflow engine is responsible for connecting "receiving the list → billing → payment → distribution → receipt / completion".

[0077] Abnormal Branch: When a network / device malfunctions, it immediately enters a degraded mode (e.g., only printing receipts and providing manual guidance), and performs compensation transactions to update the accounts and status after recovery.

[0078] Auditing and monitoring: Log key points are recorded to facilitate auditing and review.

[0079] 6.4 Data and Interfaces In-hospital standard: Preferred HL7 / FHIR description list / prescription / examination request, interacting with the in-hospital bus or REST API.

[0080] Message distribution: MQTT / message queues can be used to push "distribution results / call status / path updates" to mobile devices / self-service machines.

[0081] Security and privacy: Minimize data collection, implement tiered authorization, end-to-end encryption, and retain audit logs to meet the hospital's management requirements.

[0082] 6.5 Typical Process (From "Pre-visit" to "Post-visit") Before seeking medical treatment: Complete identity verification, registration, and filling in necessary information via mobile device or self-service machine.

[0083] During a medical visit: The doctor issues an electronic bill → The system receives and parses it instantly.

[0084] Post-visit: The system generates a bill → the patient pays with one click on their mobile device or self-service machine → the system automatically generates task assignment and path / call number.

[0085] In transit: If the window state changes, the system redistributes and updates the path.

[0086] Complete: The window receipt is written back to the system. In case of an anomaly, the workflow triggers a degradation and compensation mechanism.

[0087] 6.6 Reference Numerals 1—Multimodal identity recognition and verification. 2—Natural language understanding and interaction. 3—Registration and approval. 4—Self-service information filling. 5—List recognition and billing. 6—Payment and settlement. 7—Hardware integration and automation. 8—Process control and emergency response. 9—Multifunctional procedure processing. 10—List reception and task distribution. 11—Workflow / state machine. 12—Message queue / bus. 13—Queue management / pathing subsystem.

[0088] Advantages and technical effects A true closed loop: integrating billing, payment, distribution, and navigation / queueing into one system, eliminating the need for patients to return after payment.

[0089] Dual-terminal unification: Self-service machines and mobile terminals serve as entry points and connection points for each other, improving the continuity of medical treatment.

[0090] Multimodal robustness: OCR / NFC / face / voice / text channels are redundant with each other.

[0091] On-the-Go Intelligence: Redistribution and path updates are integrated throughout the entire process, significantly reducing non-technical losses caused by "delays / mistakes".

[0092] Evolvable: Smoothly expands to human-computer interaction for AR glasses without changing the core data and decision-making framework.

Claims

1. A fully unattended hospital AI registration system, deployed on hospital self-service terminals and patient mobile devices, the system comprising a processor, a memory, and a program running thereon, characterized in that, The program enables the system to at least achieve: (1) Multimodal identity recognition and verification: Information from sources such as document image recognition, near-field reading and face recognition is acquired in parallel, and fusion verification and liveness detection are performed. Abnormal samples are then entered into a secondary verification process. (2) Human-computer dialogue based on natural language: receive voice or text input, perform voice recognition, intent recognition and slot filling, and map the dialogue results into a structured form; (3) List identification and billing: Parse and standardize electronic medical orders or paper lists to form a cost list and a task list and bind them to the context of medical treatment; (4) Payment and Settlement: Receive payment success receipts and convert them into unified business events; (5) Task distribution and path generation: After receiving the business event, the target execution point is determined based on the multi-dimensional factors such as the queue status of the candidate execution point, equipment capacity, hospital topology, time arrangement and patient attributes, and a medical treatment path considering capacity and time constraints is generated on the indoor topology map. (6) Dynamic redistribution: When congestion, equipment unavailability, time schedule changes or patients deviating from their route are detected, secondary redistribution and route updates are triggered, and suppression strategies are adopted to avoid frequent switching; (7) Unified messaging and consistency: Event transmission and deduplication are carried out through message middleware and idempotent control, and event sourcing and compensation transactions are used to ensure the consistency and traceability of the status of distribution, calling, path and accounting. (8) Cross-terminal synchronization and seamless connection: Share a unified context between self-service terminals and mobile terminals to achieve real-time synchronization of dialogue, payment, distribution, queuing and navigation information; (9) Optional wearable extension: Present navigation and task prompts on wearable devices and keep them consistent with the above context.

2. The fully unmanned hospital AI registration system according to claim 1, characterized in that, The multimodal identity recognition and verification includes: (a) To make a fusion judgment based on probability inference or evidence theory on the results of document recognition, near-field reading and face recognition; (b) Facial features are extracted by convolutional networks or deep networks with fused attention mechanisms, and identity is determined using metric learning; (c) Liveness detection combined with depth estimation, micro-expression temporal and frequency domain features for adversarial example identification; (d) Only persist the necessary structured or template summaries, and dispose of the original images in accordance with compliance policies.

3. The fully unmanned hospital AI registration system according to claim 1, characterized in that, The natural language-based human-computer dialogue includes: (a) Speech recognition adopts an end-to-end streaming model; (b) Natural language understanding employs intent classification and slot labeling based on pre-trained Transformer, and combines conditional random fields when necessary to improve boundary consistency; (c) Dialogue management is based on rule-based finite state machines, and can be combined with strategy learning to improve the coherence of multi-turn dialogues; (d) "Dialogue as Form" generates fields through templates and pointer networks to output structured forms for registration, billing and distribution.

4. The fully unmanned hospital AI registration system according to claim 1, characterized in that, The inventory identification and billing includes: (a) The electronic inventory is mapped to task items and cost items based on a standardized interface; (b) Key fields were extracted from the paper list through text detection and recognition, layout analysis, and table structure restoration; (c) Term alignment is achieved through named entity recognition and relation extraction, and code normalization is achieved by combining domain knowledge graphs when necessary.

5. The fully unmanned hospital AI registration system according to claim 1, characterized in that, The task distribution and path generation include: (a) Generate a set of candidate execution points based on the list content and the equipment capability matrix; (b) Candidates are evaluated using a comprehensive cost model that takes into account queue load, equipment compatibility, spatial accessibility, time window and patient priority. (c) Generate medical treatment routes using a heuristic shortest path algorithm on the hospital topology map, and impose constraints on vertical transportation, congested areas and time windows; (d) Perform time-series smoothing or forecasting of queues and loads to improve evaluation stability.

6. The fully unmanned hospital AI registration system according to claim 1, characterized in that, The dynamic redistribution includes: (a) Perform secondary distribution and path update based on event triggering, and set suppression strategies to avoid frequent switching caused by short-term fluctuations; (b) The update results are consistent across mobile devices, self-service terminals, and wearable devices, and reminders and confirmations are provided in a non-intrusive manner.

7. The fully unmanned hospital AI registration system according to claim 1, characterized in that, The unified messaging and consistency include: (a) Pass critical events with idempotent flags for transmission and deduplication; (b) Use event tracing to record the state before and after an action, supporting audit playback; (c) Cross-system consistency is achieved through compensating transactions or the Saga pattern, and anomalies in accounting, distribution, queuing and path are rolled back and filled in a predefined order.

8. The fully unmanned hospital AI registration system according to claim 1, characterized in that, The cross-platform synchronization and seamless connection include: (a) Use context identifiers to migrate conversations between self-service terminals and mobile devices, allowing unfinished multi-turn conversations to continue on the other end; (b) Once either end completes identity verification, payment, or receives the distribution result, the other end will automatically display the synchronized call status, medical route, and voucher information; (c) Optional wearable devices can be equipped to present path overlays based on indoor positioning or visual anchors and be confirmed by gazing or simple gestures.

9. A fully unattended hospital AI registration system, used for unattended registration and post-diagnosis task distribution in the hospital visit process, characterized in that, include: A) Verify patient identity through multimodal identification and verification; B) Collect and process requests through natural language-based human-computer dialogue and output structured forms; C) Parse and standardize electronic or paper lists to form expense lists and task lists, and bind them to the context of the medical visit; D) Receive payment success receipt and trigger the distribution process; E) Based on the status of candidate execution points, equipment capacity, hospital topology, time schedule and patient attributes, comprehensively evaluate the candidates and generate medical treatment paths that take into account capacity and time constraints; F) Dynamic redistribution and route updates are performed when changes in status or patient deviations are detected, and consistency is maintained across multiple devices; G) Consistency and traceability are achieved through message middleware, idempotent control and event sourcing, and compensation and backfilling are performed in case of anomalies; H) Enable context sharing and seamless connection between self-service terminals, mobile devices, and wearable devices.

10. A fully unmanned hospital AI registration system, characterized in that, A computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the method of claim 9.

Images