Physiotherapy device dual engine question and answer system, method, physiotherapy device and storage medium

Abstract

The application relates to the technical field of intelligent question and answer systems, and provides a physiotherapy equipment double-engine question and answer system and method, physiotherapy equipment and a storage medium. The physiotherapy equipment double-engine question and answer system identifies a user intention category through a trained classification model, accurately routes a question to a corresponding first question and answer engine or a second question and answer engine; the first question and answer engine searches for an entity relationship based on a physiotherapy field special knowledge graph, inputs a large language model as a basis according to a search result, the second question and answer engine references real-time hardware parameters provided by an equipment state monitoring module when generating troubleshooting guidance; the second question and answer engine can link and adjust equipment output power through an equipment communication interface when generating troubleshooting guidance, automatically reduces power when detecting that an electrode piece is in poor contact, avoids discomfort or safety risks caused by blindly increasing power due to the fact that a user does not perceive contact abnormalities, and improves the safety of equipment use.

Description

Technical Field

[0001] This application relates to the field of intelligent question-answering system technology, and in particular to a dual-engine question-answering system, method, physiotherapy device and storage medium for physiotherapy equipment. Background Technology

[0002] With the rapid development of generative AI technology, intelligent question-answering systems based on Large Language Models (LLMs) have become an important research direction in the field of natural language processing. These systems provide accurate and reliable answers to users by asking questions in natural language and combining them with the context, and have been widely used in many fields.

[0003] However, in the application of existing intelligent question-and-answer solutions in physiotherapy equipment, when the electrode pads are not making proper contact, users may blindly increase the output intensity because they cannot perceive normal electrical stimulation. Once the contact improves, the previously set high-intensity output will directly act on the skin, potentially causing stinging or even discomfort. Existing systems can only provide textual guidance such as "Please check if the electrode pads are attached tightly," and cannot proactively intervene in device control to eliminate safety hazards. This disconnect between "question and answer" and "control" is the core technical problem currently facing intelligent question-and-answer systems for physiotherapy equipment. Summary of the Invention

[0004] To address the aforementioned shortcomings, this application provides a dual-engine question-and-answer system, method, physiotherapy equipment, and storage medium for physiotherapy devices. The system categorizes user questions through an intent recognition module. For knowledge consultation intents, it leverages a knowledge graph and a large language model to collaboratively generate professionally grounded answers. For operational malfunction intents, it combines real-time equipment status parameters to generate specific troubleshooting guidance. Furthermore, it proactively issues control commands when specific abnormal states are detected, achieving a secure closed loop of "question-answer – diagnosis – control." This solves the technical problem of the disconnect between intelligent question-and-answer systems and equipment control in physiotherapy equipment usage scenarios.

[0005] In a first aspect, embodiments of this application provide a dual-engine question-and-answer system for physiotherapy equipment, including: The intent recognition module is configured to receive user input statements and identify user intent categories based on a trained classification model. The user intent categories include at least knowledge consultation intent and operation failure intent.

[0006] The device status monitoring module is configured to acquire the hardware operating parameters of the physiotherapy device in real time through a communication interface. The hardware operating parameters include at least the electrode contact impedance and the current output intensity.

[0007] The first question-answering engine, in response to the intent recognition module determining that the intent is for knowledge consultation, retrieves entity relationships based on the knowledge graph of the physiotherapy field, inputs the retrieval results into the large language model, and generates a target answer containing professional basis.

[0008] The second question-answering engine, in response to the intent recognition module determining that the operation is faulty, matches the preset operation guidance library with the real-time hardware parameters obtained by the device status monitoring module to generate a fault troubleshooting guide that combines the current device status. The second question-and-answer engine is also configured to: when the device status monitoring module detects that the electrode contact impedance exceeds a preset threshold, the second question-and-answer engine generates a prompt to the user to check the skin contact status of the physiotherapy accessories, and sends a power adjustment command to the physiotherapy device host through the device communication interface to reduce the current output power to a preset safe value.

[0009] Furthermore, the preset safety value corresponding to the power adjustment command issued by the second question-and-answer engine is 40% to 60% of the current output intensity.

[0010] Furthermore, the device status monitoring module is also used to obtain the type information of the currently connected physiotherapy accessories; the second question-and-answer engine fills in the corresponding accessory name in the fault diagnosis guide according to the accessory type information.

[0011] Furthermore, the intent recognition module is a classifier built based on a BERT pre-trained language model. This classifier is obtained through supervised fine-tuning using user question data labeled with intent categories from the historical usage logs of the physiotherapy equipment.

[0012] Furthermore, the knowledge graph in the physiotherapy field includes anatomical entity relationship data and EMS waveform parameter library data; the relationship type of the knowledge graph is at least one of the following types predefined for the use scenario of physiotherapy equipment: disease-recommended waveform relationship, symptom-applicable frequency relationship, acupoint-electrode attachment position relationship, and waveform-physiological effect relationship.

[0013] This application also provides a dual-engine question-answering method for physiotherapy equipment, applied to the above-mentioned system, including the following steps: Obtain the user's input in natural language; The intent category of the question statement is identified by the intent recognition module; If the intent category is knowledge consultation intent, the first question answering engine retrieves entity relationships based on the knowledge graph of the physiotherapy field, inputs the retrieval results into the large language model, and generates a target answer containing professional basis; If the intent category is an operational malfunction intent, the second question-and-answer engine reads the real-time hardware parameters obtained by the device status monitoring module, matches them with the preset operation guidance library, and generates a troubleshooting guide that combines the current device status. If the impedance value exceeds the preset threshold, the second question-and-answer engine generates a guide prompting the user to check the skin contact of the physiotherapy accessories and sends an instruction to the physiotherapy device host through the device communication interface to reduce the current output power to a preset safe value.

[0014] Furthermore, the power adjustment command is to reduce the current output intensity to 50% of the current value.

[0015] Furthermore, the real-time hardware parameters read by the second question-answering engine also include Bluetooth connection status parameters and current output strength value; the second question-answering engine performs fault determination in a preset priority order: first determine the Bluetooth connection status, then determine the electrode contact impedance value, and finally determine the current output strength value.

[0016] This application also provides a physiotherapy device, including: a device host; and the dual-engine question-and-answer system as described above; the device host is communicatively connected to the device status monitoring module of the dual-engine question-and-answer system via the Bluetooth communication module.

[0017] This application also provides a computer-readable storage medium storing a computer program; when the computer program is executed by one or more processors, it causes the one or more processors to perform the steps of the dual-engine question-and-answer method for physiotherapy devices as described above.

[0018] The dual-engine question-and-answer system for physiotherapy equipment provided in this application accurately routes user questions to the corresponding question-and-answer engines through intent recognition, achieving differentiated services and avoiding performance trade-offs when a single model handles cross-type processing. The first question-and-answer engine retrieves entity relationships based on a knowledge graph specific to the physiotherapy field and then inputs them into a large language model to generate answers, ensuring professional accuracy. When generating troubleshooting guidance, the second question-and-answer engine refers to the real-time status parameters of the equipment, making the guidance real-time and operable. When the second question-and-answer engine detects abnormal electrode contact impedance, it actively sends a power adjustment command to the equipment while generating text guidance, realizing a safe closed loop of "question-and-answer - diagnosis - control". This control mechanism solves the problem that users may feel discomfort when they blindly increase the intensity of physiotherapy equipment due to poor contact and not feeling the current, even after the contact improves. No existing technology combines real-time impedance monitoring, personalized question-and-answer guidance, and automatic equipment power adjustment. Attached Figure Description

[0019] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 A structural block diagram of the dual-engine question-and-answer system for physiotherapy equipment provided in the embodiments of this application; Figure 2 A flowchart illustrating the first question-answering engine's processing of knowledge inquiry intent, provided for an embodiment of this application; Figure 3 A flowchart illustrating the intent of the second question-answering engine to handle operational failures, provided in an embodiment of this application; Figure 4 This is a schematic diagram of the structure of the physiotherapy device provided in the embodiments of this application. Detailed Implementation

[0021] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0022] The flowchart shown in the attached diagram is for illustrative purposes only and does not necessarily include all content and operations / steps, nor does it necessarily have to be performed in the order described. For example, some operations / steps can be broken down, combined, or partially merged, so the actual execution order may change depending on the actual situation.

[0023] It should also be understood that the terminology used in this application specification is for the purpose of describing particular embodiments only and is not intended to limit the application. As used in this application specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0024] It should also be further understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0025] The technical solution provided in this application will be described in detail below with reference to the accompanying drawings.

[0026] Please see Figure 1 , Figure 1This is a structural block diagram of the dual-engine question-and-answer system for physiotherapy equipment provided in an embodiment of this application. Figure 1 As can be seen, the dual-engine question-and-answer system includes: an intent recognition module 200, a device status monitoring module 300, a first question-and-answer engine 400, and a second question-and-answer engine 500. This dual-engine question-and-answer system is deployed on a mobile terminal APP or a cloud server and communicates with the device host of the physiotherapy equipment through the device status monitoring module 300. For a specific structural diagram of the physiotherapy equipment, please refer to... Figure 4 The illustrated embodiment.

[0027] In this embodiment, the intent recognition module 200 is configured to receive user input statements and identify the user intent category based on a trained classification model. The user intent category includes at least two categories: knowledge consultation intent and operational fault intent.

[0028] The intent recognition module 200 has a built-in classifier based on a BERT pre-trained language model. Its training data comes from the historical usage logs of the physiotherapy equipment. It is obtained by extracting user question data and labeling the intent category of each question. This classifier receives the question text input by the user and outputs intent category labels.

[0029] The intent recognition module 200 sends the classification result to the routing unit. The routing unit then routes the user's question to either the first question-and-answer engine 400 or the second question-and-answer engine 500 based on the classification result. Specifically, when the classification result indicates a knowledge inquiry intent, the user's question is routed to the first question-and-answer engine 400; when the classification result indicates an operational malfunction intent, the user's question is routed to the second question-and-answer engine 500. Simultaneously, the routing unit also transmits the current hardware parameters obtained by the device status monitoring module 300 to the second question-and-answer engine 500.

[0030] The device status monitoring module 300 is configured to acquire the hardware operating parameters of the physiotherapy device in real time via a Bluetooth communication interface. These hardware operating parameters include the Bluetooth RSSI value (reflecting connection status), electrode contact impedance (reflecting the degree of adhesion between the electrode and the skin; an impedance exceeding 200Ω indicates poor adhesion), and the current output intensity value. Among these parameters, the electrode contact impedance is a key physical quantity for determining whether an abnormal usage state exists where "poor contact causes the user to not feel the current."

[0031] Specifically, the device status monitoring module 300 connects to the physiotherapy device via a Bluetooth communication interface and sends a status query command to the physiotherapy device at preset time intervals (e.g., once every 500 milliseconds). In response to the query command, the physiotherapy device encapsulates its current hardware operating parameters into a data packet and transmits it back via the Bluetooth link. After receiving the transmitted data packet, the device status monitoring module 300 parses it and temporarily stores the resulting structured parameters in system memory for use by the second question-and-answer engine 500.

[0032] The first question-answering engine 400 responds to the intent recognition module 200's determination of a knowledge consultation intent, retrieves entity relationships based on the knowledge graph of the physiotherapy field, and inputs the retrieval results into a large language model to generate a target answer containing professional evidence.

[0033] In this embodiment, the knowledge graph upon which the first question-answering engine 400 relies is stored using a graph database. The knowledge graph contains anatomical entity relationship data and EMS waveform parameter library data. The relationship types in the knowledge graph are predefined for the usage scenarios of the physiotherapy equipment, including relationship types such as "symptom-recommended waveform," "symptom-applicable frequency," "acupoint-electrode attachment location," and "waveform-physiological effect," enabling the search results to be directly mapped to control parameters such as waveforms and frequencies that the physiotherapy equipment can execute.

[0034] Specifically, after receiving a user question distributed by the routing unit, the first question-answering engine 400 performs entity recognition and relation extraction on the user question. Based on the recognition results, it constructs a graph database query statement, retrieves matching triple information from the physiotherapy domain knowledge graph, and concatenates the retrieved triple information with the user question to form prompt words, which are then input into a large language model (such as GPT-4, Wenxin Yiyan, or open-source LLaMA). The large language model generates a fluent natural language answer based on the professional evidence in the prompt words, and this answer explicitly includes the retrieved professional evidence. The generated answer is finally presented to the user through the APP interface or device display screen.

[0035] The second question-answering engine 500 responds to the intent recognition module 200's determination of an operation malfunction intent, and generates a fault-finding guide that combines the current device status by matching the preset operation guidance library with the real-time hardware parameters obtained by the device status monitoring module 300.

[0036] In this embodiment, the operation guidance library is pre-stored in the system storage medium in the form of a structured data table. This operation guidance library contains multiple fault determination rules, each including triggering conditions and a corresponding troubleshooting guide text template.

[0037] Specifically, after receiving user questions and device status parameters distributed by the routing selection unit, the second question-answering engine 500 parses the real-time hardware parameters provided by the device status monitoring module 300, extracts the Bluetooth RSSI value, electrode contact impedance, and current output strength value; compares each parameter value with the preset fault judgment threshold in the operation guidance library, and determines whether each rule is triggered in sequence according to the preset priority order; based on the matched rule, it extracts the corresponding fault troubleshooting guide template from the operation guidance library, dynamically fills the template with the specific values ​​of the device status parameters and accessory type information, and generates personalized response content.

[0038] The core difference between the second question-and-answer engine 500 and general question-and-answer systems lies in its ability to automatically reduce the current output power to a preset safe value when the electrode contact impedance exceeds a preset threshold (e.g., 200Ω). This is achieved by generating instructions for the user to check the skin contact of the therapeutic accessories and simultaneously sending a power adjustment command to the main unit of the therapeutic device via the device communication interface. The safety logic of this linked power adjustment is based on the fact that when the electrode contact is poor, the user may blindly increase the intensity because they cannot perceive normal current stimulation. Once the contact improves, the previously set high-intensity output will directly act on the skin, potentially causing discomfort. Automatically reducing the power when high impedance is detected eliminates this safety hazard. The technical essence of this mechanism is to transform the user's "subjective lack of sensation" (inability to feel the current) into an "objective physical parameter anomaly" (impedance exceeding the standard) detectable by the device, and simultaneously completing "information notification" (text guidance) and "risk elimination" (automatic power reduction) in a closed-loop manner, thus extending the question-and-answer system from pure information provision to proactive safety control.

[0039] The following example illustrates the collaborative relationship between the modules described above.

[0040] Suppose a user inputs the question "Why can't I feel any current in my therapy pants?" via the app. The user's input is transmitted to the intent recognition module 200 via the app's internal communication interface. The intent recognition module 200 classifies the question statement based on a trained BERT classification model, determining it to be an "operational malfunction intent." Based on the classification result, the routing unit routes the question to the second question-answering engine 500, and simultaneously reads the current hardware operating parameters (Bluetooth RSSI value -65dBm, electrode contact impedance value 280Ω, current output intensity value 5) from the device status monitoring module 300, and transmits them to the second question-answering engine 500.

[0041] The second question-and-answer engine 500 analyzes the device status parameters, compares the electrode contact impedance value of 280Ω with the preset threshold of 200Ω in the operation guidance library, and determines that the impedance value exceeds the preset threshold, matching rule R002. The second question-and-answer engine 500 extracts the fault diagnosis guide template corresponding to rule R002 from the operation guidance library, dynamically fills the current impedance value "280" and accessory type "therapeutic pants" into the template, and generates a personalized response: "High electrode contact impedance detected (current impedance value is 280Ω). Please check if the therapeutic pants are close to the skin to ensure full contact between the electrode pads and the skin. If there is hair or clothing, please clean it and wear them again. The system has automatically reduced the output power to a safe value. Please manually adjust the intensity after the contact improves." At the same time, the second question-and-answer engine 500 sends a power adjustment command to the physiotherapy device through the device communication interface, automatically reducing the current output power to 50% of the current intensity (i.e., from intensity 5 to intensity 3). This linkage operation can be completed without additional user intervention, effectively avoiding discomfort caused by blindly increasing the intensity due to poor contact when the contact improves.

[0042] For the processing flow of knowledge consultation intent, taking the user input "What frequency of EMS should be used for lumbar muscle strain" as an example: The intent recognition module 200 determines that the question is a knowledge consultation intent and routes it to the first question-answering engine 400. The first question-answering engine 400 performs entity recognition on the user's question, extracting "lumbar muscle strain" (disease entity), "frequency" (parameter entity), and "EMS" (technical entity); and identifies the relationship between entities as "treatment plan". The first question-answering engine 400 constructs a query statement, retrieves matching triplet information in the physiotherapy knowledge graph, and obtains professional evidence such as ("lumbar muscle strain", "recommended frequency", "2-10Hz low frequency"), ("lumbar muscle strain", "recommended waveform", "square wave"), and ("low-frequency pulse", "physiological effect", "promotes local blood circulation"). The first question-answering engine 400 concatenates the retrieved triplet information with the user's question to form structured prompt words, and inputs them into the large language model. The large language model generates the target answer based on the professional information in the prompt words: "According to professional knowledge in the field of physiotherapy, EMS treatment for lumbar muscle strain recommends using low-frequency pulses of 2-10Hz, combined with square wave waveforms. Low-frequency pulses can penetrate the epidermal tissue and act on the deep muscle fibers of the lumbar region, promoting local blood circulation through rhythmic contractions, thereby relieving the pain and stiffness caused by lumbar muscle strain. It is recommended that each treatment session be controlled at 20-30 minutes, with the intensity set according to the patient's comfort level." This answer is presented to the user through the APP interface.

[0043] It should be noted that, in this embodiment, the intent recognition module is a binary classifier built based on a BERT pre-trained language model. The training process of this binary classifier includes: extracting user question data from the historical usage logs of the physiotherapy equipment; having professionals in the physiotherapy field label each question data with intent categories, labeling it as either knowledge consultation intent or operational malfunction intent; obtaining the BERT pre-trained model; adding a binary classification output layer on top of it; and using the labeled data for supervised fine-tuning to obtain the trained classification model. During the inference phase, the user-inputted question text is input into the trained classification model. The model outputs the probability distribution of knowledge consultation intent and operational malfunction intent, uses the category with the higher probability as the intent category label, and outputs the corresponding confidence score for use by the subsequent routing unit.

[0044] Please see Figure 2 , Figure 2 This is a flowchart illustrating the first question-answering engine's processing of knowledge inquiry intent, as provided in an embodiment of this application.

[0045] In this embodiment, the knowledge graph includes anatomical entity relationship data and EMS waveform parameter library data. The relationship types of the knowledge graph are predefined for the usage scenarios of physiotherapy equipment, including relationship types such as "symptom-recommended waveform", "symptom-applicable frequency", "acupoint-electrode attachment location", and "waveform-physiological effect". The knowledge graph is stored in a graph database to support efficient graph traversal and relationship retrieval.

[0046] like Figure 2 As shown, when the first question-answering engine 400 receives a user's question, it performs the following steps: S210: Perform entity recognition and relation extraction on user questions, and identify key entities and their semantic relationships in the questions.

[0047] Specifically, named entity recognition models can be used to identify entities such as disease names, parameter names, and equipment names in the problem, and syntactic analysis methods can be used to identify semantic relationships between entities.

[0048] S220: Construct a graph database query statement based on the recognition results.

[0049] S230: Execute the query statement to retrieve matching triple information in the physiotherapy knowledge graph.

[0050] S240: Concatenate the retrieved triplet information with the user's question to form prompt words, and input them into a large language model to generate a target answer containing professional evidence.

[0051] For example, taking the question "What frequency of EMS should be used for lumbar muscle strain?" as an example: the extracted entities include "lumbar muscle strain" (disease entity), "frequency" (parameter entity), and "EMS" (technical entity), and the identified relationships are "treatment plan" and "parameter attribute". The retrieved triplet information includes ("lumbar muscle strain", "recommended frequency", "2-10Hz low frequency"), ("lumbar muscle strain", "recommended waveform", "square wave"), and ("low-frequency pulse", "physiological effect", "promotes local blood circulation"). After concatenating the above information into structured prompts, it is input into a large language model to generate a target answer containing professional evidence.

[0052] Through the above-mentioned "retrieve first, generate later" architecture, the answers generated by the first question-answering engine 400 are supported by accurate and professional evidence provided by the knowledge graph, and also have the natural and fluent expression effect generated by the large language model.

[0053] The accessory type information is identified by the device host 900 through a resistor voltage divider, and then reported to the device status monitoring module 300 via Bluetooth communication and stored. When the second question-and-answer engine 500 generates a troubleshooting guide, it reads this information from the device status monitoring module 300 and replaces the accessory name placeholder in the guide template with the corresponding accessory name.

[0054] The operation guidance library upon which the second question-answering engine 500 relies is pre-stored in the system storage medium in the form of a structured data table. For example, this operation guidance library contains the following fault judgment rules: Rule number R001 corresponds to the trigger condition: Bluetooth RSSI value < -80dBm; the corresponding troubleshooting guide is: Weak Bluetooth signal detected, please bring your phone close to the device host, ensuring the distance is within 5 meters; the corresponding linkage operation is: None. Rule number R002 corresponds to the trigger condition: Electrode contact impedance > 200Ω; the corresponding troubleshooting guide is: High electrode contact impedance detected, please check if the accessory is in close contact with the skin. The system has automatically reduced the output power to a safe value; the corresponding linkage operation is: Automatically reduce to 50% of the current strength. Rule number R003 corresponds to the trigger condition: Bluetooth is connected and impedance ≤ 200Ω and output strength = 0; the corresponding troubleshooting guide is: High electrode contact impedance detected, please check if the accessory is in close contact with the skin; the system has automatically reduced the output power to a safe value. The corresponding linkage operation is: The device is connected normally and the electrode pads are in good contact, but the current output intensity is 0. Please check if the "Start Training" button has been clicked in the APP. The trigger condition for rule number R004 is: Bluetooth not connected; the corresponding troubleshooting guide is: The device is detected to be not connected to Bluetooth. Please ensure that the device is powered on and click "Search Devices" in the APP to pair and connect; the corresponding linkage operation is: None.

[0055] The threshold values ​​in the above fault determination rules are exemplary values ​​and can be adjusted according to the specific model and performance parameters of the equipment in actual implementation.

[0056] Please see Figure 3 , Figure 3 This is a flowchart illustrating the second question-answering engine's handling of operational failure intent, provided in an embodiment of this application. When the second question-answering engine 500 receives an operational failure question, it performs the following steps: S310: Read device status parameters.

[0057] The second question-and-answer engine 500 reads the current hardware operating parameters from the device status monitoring module 300, including: Bluetooth RSSI value, electrode contact resistance value, and current output intensity value. If the device status monitoring module 300 cannot obtain any parameters, it marks the status as "Bluetooth not connected".

[0058] S320: Matches device status parameters with fault determination rules based on preset priorities.

[0059] For example, the second question-answering engine 500 will match the read device status parameters with the preset fault determination rule numbers in the operation guidance library. Specifically, the matching can be performed in the following preset priority order to obtain the fault determination rule numbers: First priority: If the device status parameter is "Bluetooth not connected", the matching fault judgment rule number is R004; Second priority: If the Bluetooth RSSI value is <-80dBm, the matching fault judgment rule number is R001; Third priority: If the impedance value is >200Ω, the fault judgment rule number for the matching is R002; Fourth priority: If the Bluetooth RSSI value is ≥ -80dBm and the impedance value is ≤ 200Ω, the output strength = 0, and the matching fault judgment rule number is R003.

[0060] If none of the above conditions are met, the default rule will be applied, i.e., the device is in normal condition and the user is advised to check the usage.

[0061] S330: Generate troubleshooting guidance based on matching fault determination rules.

[0062] Based on the matched fault determination rule number, a specific fault troubleshooting guide is generated, and the specific values ​​of the equipment status parameters and accessory type information are dynamically filled into the template.

[0063] S340: Adjust the linkage power according to the matched fault judgment rules.

[0064] For example, when the matching rule number is R002, the second question-and-answer engine 500, while generating a guidance response, simultaneously sends a power adjustment command to the physiotherapy device host via the device communication interface, automatically reducing the current output power to 50% of the current intensity. This linked adjustment step does not rely on any additional user operation and is executed automatically by the system. The device automatically enters a safe output state while the user receives the text guidance of "poor contact," eliminating the discomfort risk that might arise from the user blindly increasing the intensity and then experiencing improved contact. Setting the preset safety value to 50% of the current intensity allows for a gentler stimulation intensity to be applied to the skin after contact improvement, avoiding the user's inability to perceive the output due to a sudden drop in intensity, thus preventing misjudgment of device malfunction. It also effectively eliminates the discomfort risk during contact improvement, representing an optimal ratio that balances safety and user experience. This four-step closed-loop mechanism of "real-time monitoring – intelligent diagnosis – text guidance – device control" is effective.

[0065] Please see Figure 4 , Figure 4 This is a schematic diagram of the structure of the physiotherapy device provided in the embodiments of this application.

[0066] like Figure 4 As shown, the physiotherapy device includes a main unit 900 and at least one detachable physiotherapy accessory 910 (such as physiotherapy pants, physiotherapy belt, physiotherapy patches, etc.). The main unit 900 is equipped with an electrode output interface 920 and a communication module 930. The electrode output interface 920 is used to connect accessory electrodes, and the communication module 930 is used to communicate with a mobile terminal.

[0067] A companion app is installed on the mobile terminal, integrating the functional modules of the dual-engine question-and-answer system as described in Example 1. After the user opens the app and establishes a Bluetooth connection with the device host 900, the app's main interface provides a text input box for the user to enter their question.

[0068] For example, in this embodiment, the intent recognition module 200, the first question-answering engine 400, and the second question-answering engine 500 are deployed on a cloud server (not shown in the figure). The user's input question is uploaded to the cloud server via the APP. After the cloud server completes the question-answering process, it sends the answer to the APP to be displayed to the user. In other embodiments, the above modules may also be deployed in whole or in part on the local side of the mobile terminal APP.

[0069] The device host 900 is equipped with multiple detachable physiotherapy accessory interfaces. Each accessory, upon connection, transmits its identification information to the host via a specific voltage divider resistor value at its physical interface. The host then reads the voltage value through an ADC interface to determine the accessory type. This accessory type information is reported via Bluetooth to the device status monitoring module 300 for storage. When the second question-and-answer engine 500 generates troubleshooting guidance, it reads the accessory type from the device status parameters and replaces the accessory name placeholder in the guidance template with the corresponding accessory name.

[0070] This application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program. When executed by one or more processors, the computer program causes the one or more processors to perform the steps of the method described above.

[0071] The computer-readable storage medium can be any tangible medium capable of storing program code, including but not limited to: USB flash drive, portable hard drive, read-only memory, random access memory, magnetic disk, optical disk, flash memory card, etc.

[0072] In one specific embodiment, the computer-readable storage medium is an installation package for a mobile terminal APP. After the user installs the APP on the mobile terminal, the APP integrates the program code of the dual-engine question-answering system.

[0073] In another specific embodiment, the computer-readable storage medium is the system disk or data disk of a cloud server. After loading the computer program stored in the storage medium, the cloud server provides an API interface for question-and-answer services to the outside world.

[0074] It should be noted that the specific implementation process of each module is the same as the specific implementation process of each step in the method embodiment, and will not be repeated here.

[0075] It should be understood that the terminology used in this application is for the purpose of describing particular embodiments only and is not intended to limit the application.

[0076] It should also be understood that the term “and / or” as used in this application and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0077] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A dual-engine question-and-answer system for physiotherapy equipment, characterized in that, include: The intent recognition module is configured to receive user input statements and identify user intent categories based on a trained classification model. The user intent categories include at least knowledge consultation intent and operation failure intent. The equipment status monitoring module is configured to acquire the hardware operating parameters of the physiotherapy equipment in real time, and the hardware operating parameters include at least the electrode contact impedance and the current output intensity. The first question-answering engine, in response to the intent recognition module determining that the intent is to seek knowledge, retrieves entity relationships based on the knowledge graph of the physiotherapy field, inputs the retrieval results into the large language model, and generates a target answer containing professional evidence. The second question-answering engine, in response to the intent recognition module determining that the operation is faulty, matches the preset operation guidance library according to the real-time hardware parameters and generates a fault troubleshooting guide that combines the current device status. The second question-and-answer engine is also configured to: when the device status monitoring module detects that the electrode contact impedance exceeds a preset threshold, while generating the fault troubleshooting guide, send a power adjustment command to the physiotherapy device host through the device communication interface to reduce the current output power to a preset safe value.

2. The system according to claim 1, characterized in that, The preset safety value is 40% to 60% of the current output power intensity.

3. The system according to claim 1, characterized in that, The device status monitoring module is also used to obtain the type information of the currently connected physiotherapy accessories; the second question-and-answer engine fills in the corresponding accessory name in the fault diagnosis guide according to the accessory type information.

4. The system according to claim 1, characterized in that, The intent recognition module is a classifier built on a BERT pre-trained language model. This classifier is trained under supervision using user question data labeled with intent categories from the historical usage logs of the physiotherapy equipment.

5. The system according to claim 1, characterized in that, The knowledge graph in the field of physiotherapy includes anatomical entity relationship data and EMS waveform parameter library data; the relationship type of the knowledge graph is at least one of the following types predefined for the use scenarios of physiotherapy equipment: disease-recommended waveform relationship, symptom-applicable frequency relationship, acupoint-electrode attachment position relationship, and waveform-physiological effect relationship.

6. A dual-engine question-and-answer method for physiotherapy equipment, applied to the system as described in any one of claims 1-5, characterized in that, Includes the following steps: Obtain the user's input in natural language; The intent category of the question statement is identified by the intent recognition module; If the intent category is knowledge consultation intent, the first question answering engine retrieves entity relationships based on the knowledge graph of the physiotherapy field, inputs the retrieval results into the large language model, and generates a target answer containing professional basis; If the intent category is an operation failure intent, the second question-and-answer engine reads the real-time hardware parameters, matches them with the preset operation guidance library, and generates a fault diagnosis guide that combines the current device status. When the electrode contact impedance value read exceeds a preset threshold, the second question-and-answer engine generates the troubleshooting guide and sends a power adjustment command to the physiotherapy device host to reduce the current output power to a preset safe value.

7. The method according to claim 6, characterized in that, The power adjustment command is to reduce the current output intensity to 50% of the current value.

8. The method according to claim 6, characterized in that, The real-time hardware parameters read by the second question-answering engine also include Bluetooth connection status parameters and current output strength value; the second question-answering engine performs fault determination in a preset priority order: first determine the Bluetooth connection status, then determine the electrode contact resistance value, and finally determine the current output strength value.

9. A physiotherapy device, characterized in that, include: The main unit of the device is equipped with an electrode output interface and a communication module; And a dual-engine question-answering system as described in any one of claims 1 to 5; The device host is connected to the device status monitoring module of the dual-engine question-and-answer system via the communication module.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program; When the computer program is executed by one or more processing modules, the one or more processing modules perform the steps of the dual-engine question-and-answer method for physiotherapy devices as described in any one of claims 6 to 8.

Images