Intelligent voice remote control system and method based on AI large model

Abstract

The invention relates to an intelligent voice remote control system based on an AI large model, and the system constructs a dual-mode architecture of a local lightweight model and a cloud large model through a model integration module, and guarantees the local low-delay response and offline availability of a basic instruction. The cloud large model is utilized to realize accurate analysis of a fuzzy instruction, a multi-intention instruction and multiple rounds of dialogues, so that the naturalness and accuracy of interaction are remarkably improved; through a unified protocol layer and an infrared learning unit in the device control module, unified control of intelligent devices such as Bluetooth, Wi-Fi and ZigBee and traditional infrared devices is realized, and the problem of device compatibility is solved. In addition, AES encryption and authority management are carried out on the data through the security protection module, and user privacy and equipment security are effectively guaranteed.

Description

Technical Field

[0001] This application relates to the field of smart home technology, and in particular to a smart voice remote control system and method based on an AI big data model. Background Technology

[0002] Currently, smart home devices are becoming increasingly popular, and voice remote controls, as a crucial entry point for human-computer interaction, directly impact user experience. However, existing voice remote control technology suffers from the following limitations: Traditional voice remote controls mostly use basic voice recognition and keyword matching technology, which can only recognize preset fixed phrase commands. When users issue ambiguous requests, complex commands with multiple intentions, or engage in natural language dialogue, these remote controls often cannot accurately interpret them. For example, traditional devices have difficulty understanding and executing commands such as "I want to watch a heartwarming family movie" or "Turn on the air conditioner and set the temperature to 26 degrees Celsius, and turn off the living room lights."

[0003] Some remote controls rely entirely on cloud processing, which can achieve good recognition results, but are limited by network transmission, resulting in high latency, unavailability when the network is down, and the risk of leakage of user privacy data; while some remote controls that rely solely on local processing, although they have fast response, have limited computing power, cannot handle complex semantic logic, and have low intelligence.

[0004] Most remote controls are designed for devices from a single brand or using a specific protocol, making it difficult to achieve unified control of multiple protocol devices in the home, such as those using infrared, Bluetooth, Wi-Fi, and ZigBee. Traditional home appliances (such as infrared-controlled TVs and air conditioners) often require separate remote controls, leading to cumbersome operation for users. Summary of the Invention

[0005] This application provides an intelligent voice remote control system, method, computer device, and computer-readable storage medium based on an AI large model, to at least solve the problem of poor intelligence in remote controls in related technologies.

[0006] In a first aspect, embodiments of this application provide an intelligent voice remote control system based on an AI large-scale model, the system comprising: The voice processing module is used to collect the user's voice signal and preprocess it, converting the voice signal into text command information; The AI ​​model integration module is used to select a target model to perform semantic parsing and task planning on the text instruction information based on the instruction complexity of the text instruction information in order to generate control instructions. The target model includes a local lightweight model and a cloud-based large model. The device control module is used to send control signals to the target intelligent device according to the control instructions based on a multi-protocol communication network in order to execute corresponding operation tasks.

[0007] In some embodiments, the AI ​​model integration module includes a model selection and evaluation unit and a dual-mode processing unit, wherein: The model selection and evaluation unit is used to evaluate the complexity of the text instruction information. The dual-mode processing unit is used to call the local lightweight model integrated locally to perform inference and generate basic control instructions when the result of the complexity evaluation is a simple instruction or a real-time requirement. The dual-mode processing unit is further configured to, when the complexity assessment result is a complex requirement or a vague expression, transmit the text instruction information to the cloud via a communication module to invoke the cloud-based large model for semantic reasoning and task decomposition, and receive complex control instructions returned by the cloud-based large model.

[0008] In some embodiments, the AI ​​model integration module further includes a semantic understanding unit, which is used for: Based on the context awareness capability of the target model, during the semantic parsing process, the text instruction information is parsed for multiple intents to identify multiple parallel sub-task intents contained in the text instruction information. Furthermore, based on the multi-turn dialogue memory function, during the semantic parsing process, the current text instruction information is correlated with historical interaction records to generate the control instruction with contextual information.

[0009] In some embodiments, the device control module includes a unified protocol layer and an infrared learning unit, wherein: The unified protocol layer is used to convert the control commands into standard control messages corresponding to the communication protocol of the target smart device through device abstraction layer technology, and send the standard control messages to the target smart device based on the communication protocol, wherein the communication protocol includes Bluetooth, Wi-Fi and ZigBee protocols; The infrared learning unit is used to record the original command waveform of the traditional infrared device paired with the target smart device, and generate a custom control protocol based on the original command waveform. The device control module is used to send infrared control signals by invoking the custom control protocol when the target smart device is a traditional infrared device.

[0010] In some embodiments, the system further includes a security protection module, which is used for: The AES encryption algorithm is used to encrypt and store the voice signals and user interaction records. Furthermore, authorized devices are connected to the control network of the intelligent voice remote control system through a preset permission management policy.

[0011] In some embodiments, the voice processing module includes a wake-up detection unit and a voice recognition unit, wherein: The wake-up detection unit is used to detect a custom wake-up word on a local low-power chip and activate the voice acquisition function when the custom wake-up word is detected. The speech recognition unit is used to perform noise reduction and echo cancellation processing on the acquired speech signal, and to identify the language of the speech signal. Furthermore, based on the language, the acquired speech signal is subjected to noise reduction and echo cancellation processing for speech-to-text conversion to obtain the text instruction information.

[0012] In some embodiments, the system further includes a multimodal feedback module, the multimodal feedback module being used for: Receive the execution result of the control command or the response content generated by the AI ​​model integration module; Based on the execution result or the response content, drive the touch screen to display visual feedback information, or drive the vibration unit to generate tactile feedback, or drive the speaker to perform voice broadcast.

[0013] Secondly, embodiments of this application provide an intelligent voice remote control method based on an AI large model, the method comprising: The voice processing module collects the user's voice signal and preprocesses it, converting the voice signal into text command information. The AI ​​model integration module is used to select a target model to perform semantic parsing and task planning on the text instruction information based on the instruction complexity of the text instruction information in order to generate control instructions. The target model includes a local lightweight model and a cloud-based large model. The device control module is used to send control signals to the target intelligent device based on the control instructions via a multi-protocol communication network in order to execute corresponding operation tasks.

[0014] Thirdly, embodiments of this application provide an intelligent voice remote control based on an AI large model. The remote control includes a microphone array, a local voice processing chip, a memory, and a processor, and is used to control a target intelligent device through the intelligent voice remote control method described in the second aspect.

[0015] Fourthly, embodiments of this application provide a computer device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method described in the second aspect above.

[0016] Compared to related technologies, the AI-based intelligent voice remote control system provided in this application constructs a dual-mode architecture of "local lightweight model + cloud-based large model" through a model integration module. While ensuring low-latency local response and offline availability of basic commands, it utilizes the cloud-based large model to accurately analyze fuzzy commands, multi-intent commands, and multi-turn dialogues, significantly improving the naturalness and accuracy of interaction. Through the unified protocol layer and infrared learning unit in the device control module, it achieves unified control of smart devices such as Bluetooth, Wi-Fi, and ZigBee, as well as traditional infrared devices, solving the device compatibility problem. In addition, the security protection module performs AES encryption and access control on the data, effectively protecting user privacy and device security. Attached Figure Description

[0017] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a structural block diagram of an intelligent voice remote control system based on an AI large model according to an embodiment of this application; Figure 2 This is a flowchart of an intelligent voice remote control method based on an AI large model according to an embodiment of this application; Figure 3 This is a schematic diagram of the internal structure of an electronic device according to an embodiment of this application. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this application clearer, the application is described and illustrated below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application. All other embodiments obtained by those skilled in the art based on the embodiments provided in this application without inventive effort are within the scope of protection of this application.

[0019] Obviously, the accompanying drawings described below are merely some examples or embodiments of this application. Those skilled in the art can apply this application to other similar scenarios based on these drawings without any inventive effort. Furthermore, it is understood that although the efforts made in this development process may be complex and lengthy, for those skilled in the art related to the content disclosed in this application, any changes to design, manufacturing, or production based on the technical content disclosed in this application are merely conventional technical means and should not be construed as insufficient disclosure of the content of this application.

[0020] In this application, the reference to "embodiment" means that a specific feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment that is mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described in this application may be combined with other embodiments without conflict.

[0021] Unless otherwise defined, the technical or scientific terms used in this application shall have the ordinary meaning understood by one of ordinary skill in the art to which this application pertains. The terms “a,” “an,” “an,” “the,” and similar words used in this application do not indicate quantity limitation and may indicate singular or plural. The terms “comprising,” “including,” “having,” and any variations thereof used in this application are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or device that includes a series of steps or modules (units) is not limited to the listed steps or units, but may also include steps or units not listed, or may include other steps or units inherent to these processes, methods, products, or devices. The terms “connected,” “linked,” “coupled,” and similar words used in this application are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. “Multiple” used in this application refers to two or more. “And / or” describes the relationship between related objects, indicating that three relationships may exist; for example, “A and / or B” can represent: A alone, A and B simultaneously, and B alone. The character " / " generally indicates that the preceding and following objects are in an "or" relationship. The terms "first," "second," and "third" used in this application are merely to distinguish similar objects and do not represent a specific ordering of the objects.

[0022] The AI-based intelligent voice remote control method provided in this application can be applied in smart home control environments. In this environment, users hold or wear an intelligent voice remote control and issue control commands via voice. This intelligent voice remote control acts as a control center, communicating with various target smart devices in the home environment (such as smart TVs, air conditioners, smart lights, and audio systems) via wireless networks (such as Wi-Fi, Bluetooth, and ZigBee) or infrared signals. Users can send not only explicit control commands (such as "turn on the air conditioner") but also express vague needs using natural language (such as "I feel a little hot"). The remote control analyzes the received voice information using its built-in AI model integration module and drives the corresponding devices to perform operations.

[0023] Figure 1 This is a structural block diagram of an intelligent voice remote control system based on an AI large model, according to an embodiment of this application. Figure 1 As shown, the system includes: a voice processing module 10, an AI model integration module 11, a device control module 12, a security protection module 13, and a multimodal feedback module 14, specifically: The voice processing module 10 is used to collect the user's voice signal and preprocess it, converting the voice signal into text command information. In this embodiment, the voice processing module 10 is configured with a microphone array and a local voice processing chip at the hardware level.

[0024] The microphone array is used for far-field sound pickup, ensuring clear acquisition of user commands even in noisy home environments. The local voice processing chip incorporates noise reduction algorithms, echo cancellation (AEC) algorithms, and an automatic speech recognition (ASR) module.

[0025] Specifically, the voice processing module 10 includes a wake-up detection unit and a voice recognition unit.

[0026] The wake-up detection unit operates on a local low-power chip and is used to detect preset custom wake-up words in real time. Users can set their own wake-up words according to their preferences. When the wake-up word is detected, the system is activated from the low-power sleep state and enters the working state, starting the voice acquisition function.

[0027] The speech recognition unit performs noise reduction and echo cancellation on the acquired speech signal after activation to remove interference from environmental background noise and the device's own playback sound. Simultaneously, this unit has language recognition capabilities, able to recognize multiple language modes such as Mandarin and Cantonese, and calls the corresponding acoustic model based on the recognized language to convert the speech signal into text-formatted text commands. In a clear speech environment, the recognition accuracy can reach over 98%.

[0028] AI model integration module 11 is used to select a target model to perform semantic parsing and task planning on the text instruction information based on the instruction complexity of the text instruction information in order to generate control instructions.

[0029] This embodiment employs a dual-mode hybrid architecture of "local lightweight model + cloud-based large model" to balance response speed and intelligent interaction capabilities. The target models include a local lightweight model and a cloud-based large model.

[0030] In some embodiments, the AI ​​model integration module 21 includes a model selection and evaluation unit and a dual-mode processing unit, wherein: The model selects an evaluation unit to assess the complexity of text command information. This evaluation process is based on a pre-defined rule base or a lightweight classifier. If the text command matches a locally stored control term (e.g., "volume up," "next track," "power off"), or is an operation with extremely high real-time requirements, it is determined to be a simple command; if the text command contains ambiguous expressions, logical reasoning requirements, content search requests, or newly discovered words that were not matched, it is determined to be a complex requirement.

[0031] In this embodiment, the dual-mode processing unit performs traffic splitting based on the above evaluation results. If the complexity evaluation result indicates a simple instruction or real-time requirement, the local lightweight model integrated into the local memory is invoked for inference. This local lightweight model is a quantized and compressed AI model with low latency and offline usability. It can quickly generate basic control instructions, ensuring that the response latency of basic functions is controlled within an extremely low range (e.g., ≤1 second), while simultaneously protecting user privacy data from leaving the local machine.

[0032] Furthermore, when the complexity assessment results in complex requirements or ambiguous expressions, the text instruction information is encrypted and transmitted to the cloud server via a communication module (such as a Wi-Fi module). The cloud server deploys a large-scale cloud model with a large number of parameters, possessing powerful semantic understanding and knowledge base retrieval capabilities. The cloud model performs deep semantic reasoning and task decomposition on the text, generating complex control instructions containing operational parameters, and returns them to the local system.

[0033] In addition, the system supports manual switching between two models. Users can choose between precise control mode (force the local model to be used first) or intelligent interaction mode (prioritize the use of the cloud model to enhance the experience) in the settings interface according to their needs.

[0034] In some embodiments, the AI ​​model integration module 11 further includes a semantic understanding unit. This unit, based on the context-aware capabilities of the target model (especially large cloud-based models), performs the following operations during semantic parsing: 1) Multi-intent parsing. Identifies multiple parallel sub-task intents contained in text command information. For example, when a user says "Turn on the air conditioner and set the temperature to 26 degrees, and turn off the living room light," the system can break it down into three independent control commands: "Air conditioner - turn on," "Air conditioner - temperature setting - 26℃," and "Living room light - turn off," and distribute them in parallel or sequentially.

[0035] 2) Multi-turn dialogue memory and association analysis. Based on the multi-turn dialogue memory function, the current text command information is associated with historical interaction records. For example, the user first issues the command "turn on the air conditioner," and then says "lower it by 2 degrees." Traditional systems may not understand the referent of "lower it by 2 degrees." The semantic understanding unit of this system, combined with the context, determines that the command belongs to the previous interaction object "air conditioner," thereby generating a control command to lower the air conditioner temperature by 2 degrees from the current temperature.

[0036] 3) Fuzzy instruction parsing. For non-standardized natural language (such as "I want to watch a heartwarming family movie" or "I close the curtains because it's raining"), the semantic understanding unit uses knowledge graphs and logical reasoning to convert "heartwarming family movie" into a content search instruction of "search type = family / warmth, rating > 8.0", or to map the causal relationship between "raining" and "closing the curtains" into the closing action of the curtain motor.

[0037] The device control module 12 is used to send control signals to the target intelligent device according to the control instructions based on the multi-protocol communication network in order to execute the corresponding operation tasks.

[0038] This module aims to address the issues of inconsistent device protocols and poor compatibility in existing technologies. The device control module 12 includes a unified protocol layer and an infrared learning unit.

[0039] The unified protocol layer employs Device Abstraction Layer (DAL) technology. It integrates the underlying differences between different brands and communication protocols (Bluetooth, Wi-Fi, ZigBee, etc.), converting standardized control commands generated by AI models into standard control messages corresponding to the communication protocols of specific target smart devices.

[0040] For example, for Wi-Fi-connected smart TVs and ZigBee-connected smart light bulbs, the unified protocol layer encapsulates TCP / IP packets and ZigBee frames respectively for transmission. An infrared learning unit is used for compatibility with traditional non-smart home appliances (such as older TVs and fans). This unit records the original command waveforms of the traditional infrared remote control paired with the target smart device through an infrared receiver tube, encodes and stores the waveform characteristics, and generates a custom control protocol.

[0041] When it is necessary to control such devices, the device control module 12 calls the custom control protocol to drive the infrared transmitter to send the corresponding infrared control signal, thereby realizing the intelligent transformation of existing home appliances.

[0042] Security protection module 13 is used to ensure data security and network security.

[0043] At the data level, this module uses industrial-grade encryption standards such as AES (Advanced Encryption Standard) to encrypt and store the collected user voice signals, generated text commands, and user historical interaction records, preventing privacy leaks during local storage or cloud synchronization.

[0044] At the network level, a device access whitelist is established through preset permission management policies. Only authorized and authenticated devices can access the control network of the smart voice remote control system, preventing malicious external devices from intruding into the home network or sending forged control commands.

[0045] The multimodal feedback module 14 is used to enhance the human-computer interaction experience. This module receives the execution results of control commands (such as "the air conditioner is on") or response content generated by the AI ​​model integration module (such as weather query results or encyclopedia Q&A answers).

[0046] Depending on the content type, different hardware units are driven to provide feedback: specifically, the touch screen can be driven to display visual feedback information (such as displaying the current temperature icon or movie poster); the vibration unit can be driven to generate tactile feedback (such as providing a short vibration confirmation when the operation is successful); or the speaker can be driven to perform TTS (Text-to-Speech) voice broadcast.

[0047] The system constructs a dual-mode architecture of "local lightweight model + cloud-based large model," ensuring low-latency local response and offline availability for basic commands while leveraging the cloud-based large model to accurately interpret fuzzy commands, multi-intent commands, and multi-turn dialogues, significantly improving the naturalness and accuracy of interaction. Through the unified protocol layer and infrared learning unit in the device control module, unified control of smart devices such as Bluetooth, Wi-Fi, and ZigBee, as well as traditional infrared devices, is achieved, resolving device compatibility issues. Furthermore, the security protection module performs AES encryption and access control on data, effectively protecting user privacy and device security.

[0048] This application also provides an intelligent voice remote control method based on an AI large model. Figure 2 This is a flowchart of an intelligent voice remote control method based on an AI large model according to an embodiment of this application, such as... Figure 2 As shown, the process includes the following steps: Step S201: System initialization and device connection.

[0049] After the smart voice remote control is powered on, it automatically activates the communication module and scans for nearby smart devices. For paired devices, it automatically establishes a connection channel. Simultaneously, it connects to the cloud server via Wi-Fi to load, verify, and initialize the local lightweight model, ensuring the system is in standby mode.

[0050] Step S102, voice wake-up and data collection.

[0051] The system operates in low-power monitoring mode. When a user's custom wake-up word is detected via the microphone array (or the user presses the physical voice button), the wake-up mechanism is triggered. The voice processing module then begins acquiring the user's voice signal and uses a local chip for noise reduction and echo cancellation to obtain clear voice data.

[0052] Step S103, instruction parsing and model allocation.

[0053] The speech processing module converts the processed speech signal into text command information. The AI ​​model integration module analyzes this text information in real time to evaluate its command complexity. If the evaluation result indicates that the command is a simple command (such as standard control words) or a high real-time requirement, it is routed to the local lightweight model. The local model quickly fills semantic slots to generate basic control commands. If the evaluation result indicates that the command is a complex requirement (such as fuzzy intent, multi-task concurrency, or knowledge-based question answering), the text information is encrypted and uploaded to the cloud. The large model in the cloud performs deep semantic reasoning, task planning, and decomposition to generate complex control commands or service response content, which is then sent back to the local machine.

[0054] Step S104, Protocol conversion and execution.

[0055] After receiving a control command, the device control module identifies the target device's communication protocol type (infrared, Bluetooth, Wi-Fi, or ZigBee) through the device abstraction layer. If the target device is a traditional infrared device, it calls the pre-learned and stored infrared waveform data and transmits an infrared signal. If the target device is a smart connected device, it encapsulates the control command into a standard message of the corresponding protocol and sends it. If the command is a personalized service request (such as asking "What's the weather like today?" or "Recommend a movie?"), it directly retrieves content data generated in the cloud.

[0056] Step S105, multimodal feedback.

[0057] Based on the execution results or response content, the system displays relevant information through the touch screen, changes the color or flashing frequency of the status indicator light, provides tactile confirmation through the vibration unit, or plays voice responses through the speaker, thus providing comprehensive feedback to the user.

[0058] Furthermore, after the interaction is completed, the system anonymizes and encrypts the interaction record and user habit data (such as commonly used air conditioning temperatures and movie viewing preferences), and synchronizes it to cloud storage when the network is idle. This data will be used for preference learning in a large cloud model and subsequent model fine-tuning to achieve a personalized experience that becomes increasingly user-friendly with repeated use. In addition, the system uses a remote update module to detect cloud update packages in real time or periodically. When a new firmware version or model parameter update is detected, the user is prompted to perform an OTA (Over-the-Air) upgrade, thereby continuously expanding the remote control's functionality, improving the accuracy of model recognition, and reducing the user's maintenance costs.

[0059] By constructing a dual-mode architecture of "local lightweight model + cloud-based large model" through the above steps, while ensuring low-latency local response and offline availability of basic commands, the cloud-based large model enables accurate parsing of fuzzy commands, multi-intent commands, and multi-turn dialogues, significantly improving the naturalness and accuracy of interaction. Through the unified protocol layer and infrared learning unit in the device control module, unified control of smart devices such as Bluetooth, Wi-Fi, and ZigBee, as well as traditional infrared devices, is achieved, solving the problem of device compatibility. In addition, the security protection module performs AES encryption and access control on data, effectively protecting user privacy and device security.

[0060] Thirdly, this application also provides an intelligent voice remote control based on an AI large-scale model. On the front panel of the remote control, the upper area features a low-power OLED touchscreen display, which visually displays the text results of AI feedback, the status icon of the currently controlled device (such as the air conditioner temperature value), and the cover information of multimedia content. The middle area houses a microphone array with built-in high-sensitivity microphones arranged in a ring, utilizing beamforming technology to achieve 360-degree omnidirectional sound pickup and accurate sound source localization. The lower area retains necessary physical buttons, including directional navigation keys, an confirmation button, and a prominent dedicated voice interaction button.

[0061] The remote control's internal circuitry integrates a high-performance, low-power system-on-a-chip (SoC) on the main control circuit board. This chip houses a central processing unit (CPU) and a dedicated neural processing unit for edge computing. The latter, as the core of the local voice processing chip, handles real-time inference tasks for the local lightweight model. Memory chips are directly soldered to the motherboard for secure storage of encrypted user configuration information and offline control command libraries. The top of the remote control features a light-transmitting infrared emission window, containing a high-power infrared emitting diode. The circuit board also integrates a Wi-Fi / Bluetooth / ZigBee three-in-one communication module and an anti-interference onboard antenna, ensuring seamless connectivity with smart devices using different protocols. Furthermore, the remote control is equipped with a rechargeable lithium polymer battery, and the bottom features a USB Type-C charging port and a power management integrated circuit (PMIC), supporting fast charging and extended standby time.

[0062] Figure 3 This is a schematic diagram of the internal structure of an electronic device according to an embodiment of this application. For example... Figure 3 As shown, an electronic device is provided, specifically the aforementioned intelligent voice remote control. Its internal structure includes a processor, memory, network interface, microphone array, infrared transceiver module, display driver circuit, and power management module connected via an internal bus. The memory includes non-volatile memory and internal memory. The non-volatile memory stores the operating system, computer programs, local lightweight AI model files, and a custom infrared protocol library. The processor provides computing and control capabilities, particularly for running the AI ​​model client and control logic. The network interface is used to communicate with external cloud servers and smart home devices. The microphone array is used to collect external voice input. When the computer program is executed by the processor, it implements the intelligent voice remote control method based on a large AI model, as described in any of the above embodiments.

[0063] Those skilled in the art will understand that Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the electronic device to which the present application is applied. The specific electronic device may include more or fewer components than shown in the figure, or combine certain components, or have different component arrangements.

[0064] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0065] The above embodiments merely illustrate several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. An intelligent voice remote control system based on an AI large-scale model, characterized in that, The system includes: The voice processing module is used to collect the user's voice signal and preprocess it, converting the voice signal into text command information; The AI ​​model integration module is used to select a target model to perform semantic parsing and task planning on the text instruction information based on the instruction complexity of the text instruction information in order to generate control instructions. The target model includes a local lightweight model and a cloud-based large model. The device control module is used to send control signals to the target intelligent device according to the control instructions based on a multi-protocol communication network in order to execute corresponding operation tasks.

2. The system according to claim 1, characterized in that, The AI ​​model integration module includes a model selection and evaluation unit and a dual-mode processing unit, wherein: The model selection and evaluation unit is used to evaluate the complexity of the text instruction information. The dual-mode processing unit is used to call the local lightweight model integrated locally to perform inference and generate basic control instructions when the result of the complexity evaluation is a simple instruction or a real-time requirement. The dual-mode processing unit is further configured to, when the complexity assessment result is a complex requirement or a vague expression, transmit the text instruction information to the cloud via a communication module to invoke the cloud-based large model for semantic reasoning and task decomposition, and receive complex control instructions returned by the cloud-based large model.

3. The system according to claim 2, characterized in that, The AI ​​model integration module further includes a semantic understanding unit, which is used for: Based on the context awareness capability of the target model, during the semantic parsing process, the text instruction information is parsed for multiple intents to identify multiple parallel sub-task intents contained in the text instruction information. Furthermore, based on the multi-turn dialogue memory function, during the semantic parsing process, the current text instruction information is correlated with historical interaction records to generate the control instruction with contextual information.

4. The system according to claim 1, characterized in that, The device control module includes a unified protocol layer and an infrared learning unit, wherein: The unified protocol layer is used to convert the control commands into standard control messages corresponding to the communication protocol of the target smart device through device abstraction layer technology, and send the standard control messages to the target smart device based on the communication protocol, wherein the communication protocol includes Bluetooth, Wi-Fi and ZigBee protocols; The infrared learning unit is used to record the original command waveform of the traditional infrared device paired with the target smart device, and generate a custom control protocol based on the original command waveform. The device control module is used to send infrared control signals by invoking the custom control protocol when the target smart device is a traditional infrared device.

5. The system according to claim 1, characterized in that, The system also includes a security protection module, which is used for: The AES encryption algorithm is used to encrypt and store the voice signals and user interaction records. Furthermore, authorized devices are connected to the control network of the intelligent voice remote control system through a preset permission management policy.

6. The system according to claim 1, characterized in that, The voice processing module includes a wake-up detection unit and a voice recognition unit, wherein: The wake-up detection unit is used to detect a custom wake-up word on a local low-power chip and activate the voice acquisition function when the custom wake-up word is detected. The speech recognition unit is used to perform noise reduction and echo cancellation processing on the acquired speech signal, and to identify the language of the speech signal. Furthermore, based on the language, the acquired speech signal is subjected to noise reduction and echo cancellation processing for speech-to-text conversion to obtain the text instruction information.

7. The system according to claim 1, characterized in that, The system further includes a multimodal feedback module, which is used for: Receive the execution result of the control command or the response content generated by the AI ​​model integration module; Based on the execution result or the response content, drive the touch screen to display visual feedback information, or drive the vibration unit to generate tactile feedback, or drive the speaker to perform voice broadcast.

8. A method for intelligent voice remote control based on a large AI model, characterized in that, The method includes: The voice processing module collects the user's voice signal and preprocesses it, converting the voice signal into text command information. The AI ​​model integration module is used to select a target model to perform semantic parsing and task planning on the text instruction information based on the instruction complexity of the text instruction information in order to generate control instructions. The target model includes a local lightweight model and a cloud-based large model. The device control module is used to send control signals to the target intelligent device based on the control instructions via a multi-protocol communication network in order to execute corresponding operation tasks.

9. A smart voice remote control based on an AI large-scale model, characterized in that, The remote control includes a microphone array, a local voice processing chip, a memory, and a processor, and is used to control a target smart device using the intelligent voice remote control method described in claim 8.

10. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, The processor implements the method of claim 8 when executing the computer program.

Images