Brain-computer interface interaction system based on generative large model

Abstract

The invention relates to the technical field of information interaction, in particular to a brain-computer interface interaction system based on a generative large model. The system comprises a user interaction interface module for dynamically generating an SSVEP control interface; the electroencephalogram equipment is used for collecting an electroencephalogram signal of the user; the data transmission module is used for transmitting an electroencephalogram signal acquired by electroencephalogram equipment to the data processing and analyzing module; the data processing and analyzing module is used for processing and analyzing the electroencephalogram signals; the large model end adopts a generative large model to perform semantic understanding and reasoning recognition on the natural language request instruction, and generates feedback information of a current task scene according to user requirements; and the equipment control module is used for executing corresponding control operation on the equipment according to the equipment control instruction selected by the user. Through deep integration with a large model, the SSVEP control interface adaptive to different task scenes can be dynamically generated, and the feasibility and expansibility of the brain-computer interface technology in practical application are improved.

Description

Technical Field

[0001] The present invention relates to the field of information interaction technology, and more specifically, to a brain-computer interface interaction system based on a generative large model. Background Art

[0002] Brain-Computer Interface (BCI) technology directly connects the brain to external devices to achieve two-way information transmission between brain signals and devices. Traditional brain-computer interfaces mainly rely on sensors to capture brain electrical activity and use technologies such as electroencephalogram (EEG) to convert brain signals into commands that can control computers or other devices.

[0003] The popularization of existing brain-computer interface systems in daily application scenarios is limited by the following aspects: First, the functional limitations are significant. Most brain-computer interface systems in the existing technology are mainly limited to certain specific functions, such as simple letter spelling, gesture control, or lightweight device control. Although the Steady-State Visual Evoked Potentials (SSVEP) technology performs well in spelling systems, the existing system functions are insufficient for complex device control requirements such as smart home control, drone operation, or complex mission planning.

[0004] Secondly, the paradigm design has limitations. Most existing SSVEP paradigm designs rely on preset flashing frequencies and position layouts, which makes it difficult to dynamically generate personalized control interfaces. This requires users to rely on fixed and limited interfaces to interact in different scenarios, reducing the flexibility and practicality of the system.

[0005] Thirdly, there is insufficient multi-language support. Most existing brain-computer interface systems are limited to specific languages, especially for non-English users who often lack complete localization support. Cross-language support is very limited, especially the lack of integration with natural language understanding and generation.

[0006] Finally, the level of intelligence is low. Although some brain-computer interface systems in the prior art can execute relatively simple commands, such as controlling electrical switches or robot movement, the system cannot reason based on complex semantics, automatically generate interfaces that adapt to different tasks, and lacks deep integration with intelligent assistants or large language models.

[0007] Therefore, in view of the above challenges, the development and application of brain-computer interface technology requires in-depth research and technological innovation in many aspects. Summary of the invention

[0008] The purpose of the present invention is to provide a brain-computer interface interaction system based on a generative large model to solve the problem that the brain-computer interface interaction system in the prior art is difficult to execute complex task scenarios and has a low level of intelligence.

[0009] In order to achieve the above-mentioned purpose, the present invention provides a brain-computer interface interaction system based on a generative large model, including a user interaction interface module, an electroencephalogram acquisition module, a data transmission module, an SSVEP dynamic interface generation module, a large model terminal and a device control module: The user interaction interface module is used to obtain the user's control intention using the SSVEP technology, and dynamically generate the SSVEP control interface based on the feedback information of the large model end for the user to select to generate the corresponding device control instructions; The electroencephalogram device is used to collect the electroencephalogram signal of the user; The data transmission module is used to transmit the EEG signals collected by the EEG device to the data processing and analysis module; The data processing and analysis module processes and analyzes the EEG signal, and then sends the decoded natural language request instruction to the large model end, and sends the decoded device control instruction to the device control module; The large model end uses a generative large model to perform semantic understanding and reasoning recognition on the input natural language request instructions, generates feedback information of the current task scenario according to user needs, and sends it to the user interaction interface module via the data processing and analysis module; The device control module performs corresponding control operations on the device according to the device control instruction selected by the user.

[0010] In some embodiments, the large model end adopts the Wenxinyiyan large model.

[0011] In some embodiments, the data processing and analysis module is a host computer or a cloud server.

[0012] In some embodiments, the data processing and analysis module uses a task discriminant component analysis algorithm to decode and identify the collected electroencephalogram signals to generate a natural language request.

[0013] In some embodiments, the task discriminant component analysis algorithm further comprises the following steps: Combining the original EEG signal with the time-delayed EEG signal to construct enhanced training data and test data; Project the augmented training and test data into the subspace spanned by the reference signal; A two-dimensional linear discriminant analysis was performed on the training trials to find the projection direction to discriminate trials across all categories; Use Fisher's criterion to derive the projection direction; By using the idempotence of the projection matrix, the process of solving the projection direction is simplified to obtain the final projection matrix. Using the obtained projection matrix, the correlation coefficient between the single test data and the matrix generated by the universal filter is calculated to identify the frequency components of the signal; The target frequency was determined by analyzing the maximum value of the correlation coefficient, thereby identifying the stimulation frequency of the SSVEP signal.

[0014] In some embodiments, the user interaction interface module determines the number of devices by analyzing feedback information from the large model end, generates a corresponding number of SSVEP flashing blocks based on the number of devices, and assigns different flashing frequencies to each SSVEP flashing block.

[0015] In some embodiments, the user interaction interface module determines the specific control task of the device through feedback information from the large model end, and generates a corresponding dynamic SSVEP control interface according to the specific control task.

[0016] In some embodiments, the user interaction interface module is used to generate text information for controlling the device, and dynamically adjust the flashing frequency, color and position layout of the flashing blocks, and determine the optimal layout of the SSVEP control interface through automatic calculation, so that each flashing block is located in the optimal position within the user's field of view.

[0017] In some embodiments, the device control module further includes a device information collection unit: The device information acquisition unit is used to obtain device status information in real time; The device control module is connected to the large model end and sends the device status information acquired in real time to the large model end; The large model side generates corresponding real-time feedback information based on the real-time acquired device status information, and sends it to the user interaction interface module via the data processing and analysis module; The user interaction interface module dynamically adjusts the steady-state visual evoked potential control interface based on real-time feedback information from the large model end.

[0018] In some embodiments, the large model side supports voice input control instructions in multiple languages.

[0019] In some embodiments, the device control module includes a home device control module: The home appliance control module is used to obtain detailed information of connectable home appliances and send it to the large model end, and perform corresponding control operations on the home appliances according to the device control instructions selected by the user; The large model side generates feedback information matching the current task scenario based on the detailed information of the device and the historical operation data of the user, and the feedback information includes the device name, device function and device status.

[0020] In some embodiments, the device control module includes a robotic arm control module: The robot arm control module is used to obtain detailed information of the robot arm and send it to the large model end, and perform corresponding control operations on the robot arm according to the device control instructions selected by the user; The large model end generates feedback information matching the current task scenario based on the detailed information of the robotic arm combined with natural language request instructions, and the feedback information includes movement commands, posture adjustments, grasping instructions and placement instructions.

[0021] In some embodiments, the device control module includes a drone control module: The drone control module is used to obtain detailed information of the drone and send it to the large model end, and perform corresponding control operations on the drone according to the device control instructions selected by the user; The large model side generates feedback information matching the current mission scenario based on the detailed information of the drone and natural language request instructions. The feedback information includes flight path planning, speed adjustment, and mission execution.

[0022] The brain-computer interface interaction system based on the generative big model provided by the present invention supports dynamic interface generation and intelligent task reasoning through the deep integration of the SSVEP speller and the big model, significantly improves the interaction efficiency, can optimize the interaction experience according to user habits, and promotes the development of brain-computer interface technology from simple operation to intelligent task execution, laying a solid foundation for the popularization and wide application of technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The above and other features, properties and advantages of the present invention will become more apparent through the following description in conjunction with the accompanying drawings and embodiments, in which the same reference numerals always represent the same features, wherein: Figure 1 A principle block diagram of a brain-computer interface interaction system based on a generative large model according to an embodiment of the present invention is disclosed; Figure 2 A workflow diagram of a brain-computer interface interaction system based on a generative macro model according to an embodiment of the present invention is disclosed; Figure 3 A schematic diagram of a work flow of a home control system according to an embodiment of the present invention is disclosed; Figure 4 A workflow diagram of a robot / drone control system according to an embodiment of the present invention is disclosed.

[0024] The meanings of the reference numerals in the figures are as follows: 100 user interface module; 200 EEG equipment; 300 data transmission module; 400 data processing and analysis module; 500 large model end; 600 device control module. DETAILED DESCRIPTION

[0025] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the invention and are not used to limit the invention.

[0026] Figure 1 The principle block diagram of the brain-computer interface interaction system based on the generative large model according to an embodiment of the present invention is disclosed, such as Figure 1 As shown, the brain-computer interface interaction system based on the generative large model proposed in the present invention includes a user interaction interface module 100, an electroencephalogram device 200, a data transmission module 300, a data processing and analysis module 400, a large model terminal 500 and a device control module 600: The user interaction interface module 100 is connected to the data processing and analysis module 400, and is used to obtain the user's control intention using the SSVEP technology, and dynamically generate an SSVEP control interface based on the feedback information of the large model end 500 for the user to select to generate corresponding device control instructions; The EEG device 200 is connected to the data transmission module 300 and is used to collect the EEG signal of the user; The data transmission module 300 is connected to the data processing and analysis module 400 and is used to transmit the EEG signal collected by the EEG device 200 to the data processing and analysis module 400; The data processing and analysis module 400 is connected to the large model end 500 and the device control module 600, respectively, to process and analyze the EEG signal, and then send the decoded natural language request instruction to the large model end 500, and send the decoded device control instruction to the device control module 600; The large model end 500 uses a generative large model to perform semantic understanding and reasoning recognition on the input natural language request instructions, generates feedback information of the current task scenario according to user needs, and sends it to the user interaction interface module 100 via the data processing and analysis module 400; The device control module 600 performs corresponding control operations on the device according to the device control instruction selected by the user.

[0027] It should be noted that there are many kinds of generative big models, such as GPT, Claude, etc., but in the present invention, the Wenxinyiyan big model is used as the optimal recommendation.

[0028] The Wenxinyiyan model has significant advantages in Chinese processing, especially in natural language understanding and generating Chinese instructions. Wenxinyiyan's Chinese semantic understanding and reasoning capabilities can provide higher accuracy and user experience. Although other generative large models can also provide similar functions, their capabilities in Chinese processing may not be as good as the Wenxinyiyan model, especially in complex Chinese instruction parsing and task reasoning.

[0029] The Wenxinyiyan model's deep understanding and task decomposition capabilities in the Chinese context provide key support for the intelligent control of the system. Therefore, although other large models can replace the Wenxinyiyan model, considering the accuracy of language processing and the specific needs of the project. In this embodiment, the Wenxinyiyan model is the most appropriate choice.

[0030] Figure 2 The following discloses a workflow diagram of a brain-computer interface interaction system based on a generative large model according to an embodiment of the present invention. Figure 2 As shown, the brain-computer interface interaction system based on the generative large model proposed in the present invention includes processes such as EEG signal acquisition, data processing and analysis, large model interaction, dynamic generation of SSVEP control interface and device control.

[0031] Among them, the EEG signal collection process is mainly implemented by the EEG device 200; The dynamic generation process of the SSVEP control interface is mainly implemented by the user interaction interface module 100; The data processing and analysis process is mainly implemented by the data processing and analysis module 400; The large model interaction process is mainly implemented by the large model end 500; The device control process is mainly implemented by the device control module 600.

[0032] In this embodiment, the user interaction interface module 100 is a SSVEP speller to implement brain-computer interface input.

[0033] Combine the following Figure 1 and Figure 2 The various components and corresponding workflows of the brain-computer interface interaction system based on the generative large model proposed in the present invention are specifically described. The various hardware modules are precisely integrated to jointly realize the core functions of the brain-computer interface.

[0034] The user interaction interface module 100 is used to interact with the user, display the SSVEP control interface, and feedback corresponding control options according to the user's intention. In the SSVEP control mode, the SSVEP control interface displays control items through different flashing frequencies and color codes, and the user issues commands by gazing at the corresponding flashing blocks.

[0035] SSVEP is a steady-state visual evoked potential response based on the frequency of visual stimulation. When the user looks at the flashing squares of different frequencies, the brain will generate electrical signals of corresponding frequencies. The system collects these EEG signals, converts the time domain signals into frequency domain signals using Fourier transform, extracts the EEG response corresponding to each frequency, and thus determines the user's gaze target.

[0036] In this embodiment, the electroencephalogram device 200 is used to collect the electroencephalogram signal of the user.

[0037] The electroencephalogram (EEG) device 200 is attached to the user's head through 8 electrodes to monitor and record the electrical activity of the cerebral cortex in real time, collect EEG signals at a high sampling rate (1000 Hz), and perform preliminary filtering, amplification and denoising on the EEG signals to ensure that the brain response of the user when looking at different frequencies is captured in real time.

[0038] In this embodiment, the data transmission module 300 is used to transmit the EEG signal collected by the EEG device 200 to the data processing and analysis module 400 .

[0039] Furthermore, the data transmission module 300 is a wireless data transmission module, which transmits the processed EEG signals and control instructions to the data processing and analysis module 400 via a wireless communication protocol (Wi-Fi).

[0040] The wireless data transmission module enables the brain-computer interface interaction system to be operated remotely, and users do not need to connect through physical cables, ensuring efficient and real-time transmission of data.

[0041] Furthermore, the wireless data transmission module may be a Xiaomi brand EEG-specific wireless router.

[0042] In this embodiment, the data processing and analysis module 400 is a host computer or a cloud server, which is responsible for receiving and processing instructions input by users.

[0043] Furthermore, the data processing and analysis module 400 performs signal preprocessing on the collected EEG signals.

[0044] Preprocessing includes removing 50 Hz power supply noise, as well as removing eye movement and muscle artifacts. The data processing and analysis module 400 uses filters and artifact removal algorithms to ensure the purity and accuracy of the signal and avoid non-target signals interfering with the analysis results.

[0045] In the signal analysis stage, the data processing and analysis module 400 uses a signal processing algorithm to decode and identify the collected EEG signals and generate natural language request instructions.

[0046] In this embodiment, the signal processing algorithm is a Task-Discriminant Component Analysis (TDCA) algorithm.

[0047] The TDCA algorithm is a dimensionality reduction algorithm for data in a multi-task learning environment. In the EEG signal decoding task, by searching for an optimal projection matrix, it can map the original high-dimensional data into a new low-dimensional space, so that the similarity of samples within the same task in this space will be retained as much as possible, and the difference between samples of different tasks will be maximized, thereby improving the distinction between tasks, accurately identifying the visual stimulation frequency that the user is concerned about, and generating effective natural language request instructions, thereby calling the Wenxin Yiyanda model.

[0048] Furthermore, the task discriminant component analysis algorithm comprises the following steps: Combining the original EEG signal with the time-delayed EEG signal to construct enhanced training data and test data; Project the augmented training and test data into the subspace spanned by the reference signal; A two-dimensional linear discriminant analysis was performed on the training trials to find the projection direction to discriminate trials across all categories; Use Fisher's criterion to derive the projection direction; By using the idempotence of the projection matrix, the process of solving the projection direction is simplified to obtain the final projection matrix. Using the obtained projection matrix, the correlation coefficient between the single test data and the matrix generated by the universal filter is calculated to identify the frequency components of the signal; The target frequency was determined by analyzing the maximum value of the correlation coefficient, thereby identifying the stimulation frequency of the SSVEP signal.

[0049] Specifically, the TDCA algorithm adopts a discriminant model. Different from the traditional strategy of optimizing spatial filters for each class, it uses discriminant analysis to search for spatial filters for all classes.

[0050] First, the enhanced training and test data are formed by combining the original signal and its time-delayed version. During the pre-training process, for a given i-th trial, , ,in, represents the number of trials, Represents the number of sampling points, and the enhanced data is: (1), in, represents the EEG signal after data enhancement, Represents the experimental data of the delay point l, that is, from time (data point) Time A delayed version of the data.

[0051] In the process of test data enhancement, in order to ensure that Data points do not enter the test data, exceeding The data points are filled with zero.

[0052] (2), in, Represents from time Time data.

[0053] The augmented data is then projected onto the subspace spanned by the reference signal.

[0054] (3), Where P is the orthogonal projection matrix of the i-th class: (4), Where Q comes from the QR decomposition of the sine and cosine reference signal Y, corresponding to the i-th stimulus frequency : (5), (6), in, is the harmonic number, is the sampling rate. Then, for both training and test data, further secondary data augmentation is constructed.

[0055] (7), Then, a two-dimensional linear discriminant analysis was performed on the training trials to find the projection direction to discriminate trials across all categories.

[0056] The inter-class difference matrix and intra-class difference matrix are defined as: (8), (9), in, and Represent the two-dimensional category center of the i-th category and the i-th sample respectively, and the superscript a represents all categories. From formula (10), we get: (10), Then, the projection direction is derived using the Fisher criterion: (11), Among them, the scatter matrix and It is expressed as: (12), (13), Finally, using the idempotence of the projection matrix, , formula (11) can be rewritten as: (14), in, and According to formula (8) (9) composition.

[0057] For the projection matrix, , ,in, Indicates direct and Indicates the number of blocks.

[0058] Finally, according to the obtained universal filter W, the single trial data generated and The correlation coefficient of is solved as: (15), Identify the target frequency based on the stimulus frequency corresponding to the maximum correlation coefficient : (16), These data processing and analysis steps ensure that the SSVEP signal can quickly and accurately reflect the user's intention, making it easier for the system to perform subsequent operations.

[0059] In this embodiment, the big model end 500 is configured as the Wenxinyiyan big model API, which provides services based on natural language understanding and reasoning.

[0060] The brain-computer interface interaction system based on the generative big model proposed in the present invention solves the shortcomings of the existing brain-computer interface interaction system in terms of flexibility, intelligence and multi-language support by combining the SSVEP control interface with the Wenxinyiyan big model.

[0061] The Wenxin Yiyan model has powerful natural language understanding and generation capabilities, and can generate SSVEP control interfaces adapted to different scenarios based on user input.

[0062] Different from the existing fixed paradigm generation method, the SSVEP control interface of the present invention is no longer fixed, but is generated by the Wenxin Yiyanda model in accordance with factors such as the number of devices and device names. The flashing color block text, flashing frequency and layout position of this paradigm are dynamically generated according to the specific scene, which improves the flexibility and adaptability of the system.

[0063] Users can generate customized SSVEP control interfaces in different scenarios, such as home appliance control, robotic arm operation, or drone mission planning, providing a more intuitive interactive experience.

[0064] In this embodiment, the SSVEP control interface of the SSVEP speller determines the number of devices according to the feedback information from the large model end, generates a corresponding number of SSVEP flashing blocks based on the number of devices, and assigns different flashing frequencies to each SSVEP flashing block.

[0065] For example, when a user inputs a command through the SSVEP speller, such as "I want to control home appliances", the system will send the command to the Wenxin Yiyanda Model API through an HTTP request.

[0066] With its powerful natural language processing (NLP) capabilities, the Wenxin Yiyan model first performs semantic understanding of the input sentences, identifies user needs, and then generates feedback information suitable for the current task scenario based on semantic reasoning.

[0067] When a user requests to control a home appliance, the feedback information returned by the Wenxin Yiyanda model includes a home appliance control device list, which lists all controllable devices, such as lights, air conditioners, televisions, etc.

[0068] The semantic reasoning of the Wenxin Yiyan model is not just simple text comprehension, but also includes the judgment of task context and the deduction of logic between devices, ensuring that the output device list can accurately match user needs.

[0069] Once the Wenxin Yiyan model returns to the device home appliance control, the user interaction interface module will enter the dynamic generation link of the SSVEP control interface. This is another important innovation of the system.

[0070] The interfaces of existing brain-computer interface systems are often static and cannot change dynamically according to user needs. However, this system can dynamically generate the SSVEP control interface in real time by calling the Wenxin Yiyan model and based on the feedback information of the Wenxin Yiyan model.

[0071] Specifically, the user interaction interface module 100 analyzes the number of devices by analyzing the feedback information of the large model end, generates a corresponding number of SSVEP flashing blocks in the SSVEP control interface according to the number of devices, and assigns different flashing frequencies to each block.

[0072] Users can select the corresponding device by looking at these flashing squares.

[0073] In this embodiment, the user interaction interface module 100 determines the specific control task of the device through the feedback information of the large model terminal 500, and generates a corresponding dynamic SSVEP control interface according to the specific control task.

[0074] Specifically, the Wenxin Yiyan model returns the specific control tasks of each device, such as turning it on and off, adjusting the temperature, changing the brightness, etc.

[0075] The user interaction interface module 100 will dynamically generate the SSVEP control interface according to these tasks, and the user can select the corresponding operation command through the SSVEP speller to complete the device control.

[0076] In this embodiment, the user interaction interface module 100 is used to generate text information for controlling the device, and dynamically adjust the flashing frequency, color and position layout of the flashing blocks, and determine the best layout scheme through automatic calculation, so that each flashing block is located at the optimal position within the user's field of view.

[0077] When the SSVEP control interface is generated, the user interaction interface module will not only generate the text for device control, but also dynamically adjust the flashing frequency, color and position layout of the flashing blocks.

[0078] The user interaction interface module 100 is configured based on the automatically calculated optimal layout so that each flashing square is located at the optimal position within the user's field of vision, making it easier for the user to complete the selection by gazing.

[0079] The design of the flashing frequency is based on the complexity of device control and user habits, ensuring that there is enough difference between the frequencies to avoid incorrect selection.

[0080] The color and font size of each block will also automatically adjust to ensure clear legibility under different lighting conditions, improving the user experience.

[0081] In this embodiment, the large model end 500 supports voice input control instructions in multiple languages.

[0082] By combining with the Wenxin Yiyan big model, the system is compatible with command input in more than ten languages. Users can input control commands not only in Chinese, but also in other languages ​​such as English and French. The system automatically recognizes the input language and processes it through the big model.

[0083] The Wenxin Yiyan model can generate corresponding feedback information according to the input language to adjust the SSVEP control interface, and the system does not need additional translation or language conversion, which greatly improves the international adaptability of the system and the convenience of user use.

[0084] The brain-computer interface interaction system based on the generative large model proposed in the present invention has wide cross-domain application compatibility.

[0085] In addition to home appliance control, the system also supports the management of complex devices such as robotic arm control and drone control. These functions are achieved thanks to the universal control interface of the device control module 600. The device control module 600 consists of multiple submodules, including a home appliance control module, a robotic arm control module, and a drone control module.

[0086] The device control module 600 directly controls the movement or task execution of devices such as a robotic arm or a drone according to the control commands extracted from the EEG signals.

[0087] Typically, the device control module 600 includes hardware such as a drive motor, a sensor, and an actuator, and is capable of accurately controlling the movement or other behaviors of the device (such as completing simple motion instructions, achieving object grabbing, etc.).

[0088] The brain-computer interface interaction system based on the generative large model proposed in the present invention can interact with a variety of devices and generate a customized SSVEP control interface according to the control requirements of different devices.

[0089] For example, when controlling a robotic arm, the feedback information generated by the Wenxin Yiyan model will include a series of operation options, such as moving, grabbing, rotating, etc. After the user selects the corresponding operation through the SSVEP speller, the data processing and analysis module 400 will pass the decoded device control instructions to the robotic arm control module to perform specific actions.

[0090] In terms of drone control, the feedback information generated by the Wenxin Yiyan model will include control options such as flight path selection and speed adjustment. Users can also use the SSVEP speller to select flight modes and plan tasks by looking at flashing blocks.

[0091] During the device control process, the device control module 600 also includes a device information acquisition unit for real-time monitoring and obtaining the operating status information of the device and providing accurate feedback.

[0092] For example, the joint angle and grasping force of the robot arm, the flight altitude and speed of the drone and other key parameters can be collected in real time by the device information collection unit. The collected device status information is transmitted to the large model end 500 through the device control module 600, and the Wenxin Yiyan large model generates real-time feedback information. The user interaction interface module 100 dynamically adjusts the SSVEP control interface according to this feedback information. In this way, the user can obtain the real-time status of the device operation in a timely manner, and optimize the operation decision based on the adjusted SSVEP control interface, thereby improving the accuracy and efficiency of the device operation.

[0093] Different from traditional control methods, the present invention adopts a brain-computer interface-based control method that does not require manual operation, greatly improving the user's task execution efficiency in complex scenarios.

[0094] The brain-computer interface interaction system based on the generative big model proposed in this invention has intelligent task reasoning and execution functions. It is not limited to home appliance control, but can also control robotic arms, drones and other equipment, providing compatibility with multiple application scenarios. Through the semantic understanding ability of the Wenxinyiyan big model, users can input more complex task instructions, and the system can automatically generate the corresponding SSVEP control interface according to the instructions, which is suitable for a wide range of task execution.

[0095] The semantic understanding and reasoning capabilities of the Wenxin Yiyan model are capable of processing complex instructions. For example, when the user inputs "let the drone fly to the specified location and take pictures", the feedback information generated by the Wenxin Yiyan model will include multiple sub-tasks such as path selection and shooting mode selection.

[0096] Through the SSVEP control interface, users can easily select and execute these tasks. This automatic reasoning and dynamic generation capability makes the system not only suitable for simple tasks, but also capable of more complex work scenarios.

[0097] The brain-computer interface interaction system based on the generative big model proposed in this invention solves many technical problems of the existing brain-computer interface technology through deep integration with the Wenxinyiyan big model. Whether it is dynamic SSVEP control interface generation, multi-language support, or cross-device compatibility, it marks a new breakthrough in brain-computer interface technology. The system can be widely used in smart home, industrial automation, complex equipment control and other fields, and has strong scalability and application prospects.

[0098] Next, the functions and advantages of the brain-computer interface interaction system will be further explained by demonstrating its application examples in home control systems and robotic arm / drone control.

[0099] Figure 3A schematic diagram of the working process of a home control system according to an embodiment of the present invention is disclosed. Figure 3 In the illustrated embodiment, the brain-computer interface interaction system based on the generative large model proposed by the present invention is used as a home control system, wherein the device control module includes a home device control module:

[0100] The home appliance control module is used to obtain detailed information of connectable home appliances and send it to the large model end, and perform corresponding control operations on the home appliances according to the device control instructions selected by the user;

[0101] The large model side generates feedback information matching the current task scenario based on the detailed information of the device and the historical operation data of the user, and the feedback information includes the device name, device function and device status.

[0102] More specifically, the home device control module is a Home Assistant module.

[0103] The brain-computer interface interaction system of the present invention can automatically acquire controllable electrical appliances in the local area network through integration with the Home Assistant module, and automatically generate the types and numbers of controlled electrical appliances through a large model to form a targeted SSVEP control interface.

[0104] Next, the technical implementation details of the home control system will be introduced in detail. These details cover several key steps, including automatic device discovery, interaction with the Wenxin Yiyanda model, and dynamic generation of the SSVEP control interface.

[0105] First, the system realizes the automatic discovery function of devices in the local area network through the Home Assistant module.

[0106] Home Assistant is an open source smart home control platform that can integrate with multiple protocols (Zigbee, Z-Wave, and Wi-Fi) and a large number of brands and models of smart devices.

[0107] When the system starts, the API of the Home Assistant module is used to automatically scan the connectable devices in the local area network to obtain the type, status, device name and control method (such as switching, adjusting temperature, etc.) of all controllable devices.

[0108] The API of the Home Assistant module provides a standardized interface for device discovery, through which the system can obtain detailed information about the device in real time, including the device's IP address, control protocol, function description, etc.

[0109] After the device is automatically discovered, the system will pass the device information to the Wenxin Yiyan large model on the large model side. The Wenxin Yiyan large model on the large model side automatically generates the device list and corresponding control options required by the SSVEP control interface by analyzing the device information.

[0110] The Wenxin Yiyan big model on the big model side can infer the control options and interaction logic suitable for the current scenario based on the functions, types and quantity of the equipment.

[0111] Specifically, the Wenxin Yiyan big model on the big model side generates one or more control commands for each device (such as "on / off", "adjust temperature", etc.), and determines the required number of control blocks based on the number of devices.

[0112] The Wenxin Yiyan big model on the big model side not only generates basic control commands for the device, but also optimizes these commands based on the user's historical behavior and semantic input, such as giving priority to displaying commonly used functions on devices that users habitually operate.

[0113] Next, the user interface module generates a new SSVEP control interface through these device control information.

[0114] During this process, the system first assigns a flashing square to each device, with each square representing a device or a control function of the device.

[0115] The allocation of flashing frequencies is done dynamically, and the system ensures that there is sufficient differentiation between the flashing frequencies of different devices so that users can select specific control functions by looking at different blocks.

[0116] In order to achieve a better interactive experience, the system will also dynamically adjust the layout of the SSVEP control interface, automatically arrange the position and size of the blocks according to the number of devices, and ensure that users can clearly see each control option and make selections.

[0117] In addition, the system will dynamically update the SSVEP control interface based on the real-time status information provided by the Home Assistant module.

[0118] If the status of a device in the local area network changes (such as a device going online, offline, or updating its operating status), the Home Assistant module will push notifications of the device status change in real time. The large model will generate new feedback information through the Wenxin Yiyan large model based on the real-time status of the device and pass it to the user interaction interface. The user interaction interface module will update the SSVEP control interface in real time based on this feedback information, add or remove corresponding control blocks, and ensure that the user interface always reflects the actual situation of the current controllable device.

[0119] Through the above process, the brain-computer interface interaction system proposed by the present invention, the SSVEP control interface can always accurately reflect the status of the controllable devices in the current LAN environment, without the need for manual configuration by the user, and automatically generate an SSVEP control interface suitable for the current LAN environment, and allow users to achieve efficient control of the device through the SSVEP control interface. The system not only simplifies the operating steps of smart home control, but also improves the dynamic adaptability of the SSVEP control interface and the intelligence level of the system through the automatic generation and reasoning functions of large models, so that it can adapt to more scenarios and device types. Users can achieve efficient and accurate device control through the optimized SSVEP control interface, thereby further improving the overall user experience.

[0120] In this system, the big model end is embedded through the Wenxin Yiyan big model API, which is used to process user instructions, parse device information, generate the SSVEP control interface and perform intelligent reasoning.

[0121] The embedding of the Wenxinyiyan model realizes multiple core functions of the system, such as semantic understanding, device control reasoning, and dynamic interface generation. The following is a technical description of the Wenxinyiyan model in the brain-computer interface interaction system proposed in this invention: First, after the system receives the Chinese instructions entered by the user through the SSVEP speller, it converts these instructions into natural language request instructions and sends them to the API interface of the Wenxin Yiyan model.

[0122] The Wenxin Yiyan model API supports natural language understanding (NLU) and natural language generation (NLG), so the natural language request instruction will be processed by the Wenxin Yiyan model's large-scale corpus and pre-trained model for in-depth semantic analysis and reasoning.

[0123] In specific implementation, the Wenxin Yiyanda model communicates with the system through HTTP requests. The system sends the input user instructions to the Wenxin Yiyanda model API interface by using HTTP POST requests.

[0124] The request contains the user's input text, such as "I want to control my home appliances", and possible context information. The Wenxin Yiyan model will use its language model to parse the request text, determine the user's intention, and recognize that this is a home appliance control request.

[0125] To achieve this, the Wenxin Yiyan model parses the text through multiple levels of semantic understanding networks and uses a pre-trained knowledge base to infer appropriate device control scenarios.

[0126] The semantic understanding module of the Wenxin Yiyan model can automatically process these natural language inputs, extract the operation commands that users need to execute, and generate device control options.

[0127] In this process, the semantic reasoning and context understanding capabilities of the Wenxin Yiyan model play an important role. The Wenxin Yiyan model not only understands the user's current input, but also combines contextual information, such as the user's past operation history or device status information, to generate more personalized operation suggestions.

[0128] For example, if the user has frequently controlled the air conditioner before, the Wenxin Yiyanda model will give priority to pushing air conditioner-related control options to the user.

[0129] After semantic understanding and reasoning are completed, the Wenxin Yiyan model will generate a JSON response with a list of control devices and operation options, and return it to the system as feedback information.

[0130] When the system receives the device control information returned by the Wenxin Yiyan model, it will parse the returned device data and generate a new SSVEP control interface based on the requirements of the SSVEP control interface.

[0131] The JSON response data of the feedback information returned by the Wenxin Yiyan model usually includes the following parts: 1) Device name: The name of each controllable device, such as "light", "air conditioner", "TV", etc.

[0132] 2) Device functions: The executable operations corresponding to each device, such as "on / off", "adjust temperature", "adjust brightness", etc.

[0133] 3) Device status: The current status information of the device, such as "enabled", "disabled", etc., is used to dynamically update the interface.

[0134] Based on this data, the user interface module will automatically generate a new SSVEP control interface, in which each device and control option will be mapped to a flashing square in the SSVEP control interface.

[0135] Each flashing square corresponds to a device or control function and flashes at different frequencies. The system uses feedback data from the Wenxin Yiyan model to ensure that the control options of each device are accurately presented, while ensuring that users can perform specific device operations by looking at the corresponding flashing frequency.

[0136] In addition, the Wenxin Yiyan model also plays a key role in the intelligent generation of equipment lists.

[0137] When the Home Assistant module automatically discovers devices through the local area network, the Wenxin Yiyan model will perform intelligent reasoning based on the type and status of these devices and the user's historical operation data.

[0138] For example, if the system detects three devices: air conditioner, light and TV, the Wenxin Yiyanda model will infer the control interface configuration that best suits the user based on the characteristics of these devices.

[0139] An air conditioner may have multiple operation options such as "on / off" and "temperature adjustment", while a light may only have an "on / off" option. Using the feedback information of the large model based on the complexity of the device operation, the user interaction interface module can automatically generate the most appropriate interface layout, thereby reducing the user's operating burden.

[0140] In this process, the Wenxin Yiyan model not only serves as a language model, it also plays a role in organizing, classifying and reasoning device information.

[0141] This reasoning capability enables the user interface module to automatically generate the most suitable SSVEP control interface when facing complex equipment control requirements, and dynamically update the control options according to the real-time equipment status (such as equipment online, offline, etc.). Through this deep integration with Wenxin Yiyan, the system realizes intelligent and dynamic control interface generation, improving the efficiency of equipment control and user experience.

[0142] Finally, the multi-language support of the Wenxinyiyan model also makes it possible for the international expansion of the system. Users can not only input instructions in Chinese, but Wenxinyiyan also supports semantic understanding and generation capabilities in multiple languages. This means that users can control the device in multiple languages ​​(such as English, French, etc.), and the system will generate the corresponding SSVEP control interface based on different language inputs. This provides a wider range of adaptability and flexibility for the global application of the system.

[0143] In summary, the brain-computer interface interaction system based on the generative large model proposed in the present invention realizes natural language understanding, device intelligent reasoning and interface dynamic generation by combining the Wenxinyiyan large model, which significantly improves the intelligence level of the system and user experience. The system uses the Wenxinyiyan large scale to realize the intelligence of the entire process from user input to device control interface generation, so that the system can automatically adapt to different devices and scenarios, thereby providing a more efficient and personalized smart home control solution.

[0144] Figure 4 The working flow chart of the robot / drone control system according to one embodiment of the present invention is disclosed. Figure 4 In the illustrated embodiment, the brain-computer interface interaction system based on the generative big model proposed in the present invention can not only realize the control of smart home devices through the Wenxinyiyan big model, but can also be applied to the complex task control of robotic arms and the control of drones.

[0145] The embedding of the Wenxin Yiyan model enables the system to process complex natural language instructions and convert them into device control commands suitable for execution, thereby completing complex operating tasks.

[0146] The following are the specific technical implementation details of the system in terms of robotic arm and drone control.

[0147] The brain-computer interface interaction system based on the generative large model proposed in the present invention is used as a robotic arm control system, wherein the device control module includes a robotic arm control module: The robot arm control module is used to obtain detailed information of the robot arm and send it to the large model end, and perform corresponding control operations on the robot arm according to the device control instructions selected by the user; The large model end generates feedback information matching the current task scenario based on the detailed information of the robotic arm combined with natural language request instructions, and the feedback information includes movement commands, posture adjustments, grasping instructions and placement instructions.

[0148] In terms of complex task control of the robotic arm, users input instructions through the SSVEP speller, such as "I want to operate the robotic arm", or more specific task instructions such as "grab the object and place it in the specified location."

[0149] When the user inputs these natural language control instructions, the brain-computer interface interaction system proposed in the present invention sends these instructions to the Wenxin Yiyanda model through the API. The natural language understanding module of the Wenxin Yiyanda model can parse the complex sentences input by the user and extract the operation object (robotic arm) and task goal (grasp the object and place it).

[0150] Furthermore, the Wenxin Yiyan model can not only understand basic operation commands, but also infer more detailed operation steps based on the context of the task and the current scenario.

[0151] For example, for complex tasks such as "grasping an object", the large model will generate subtasks such as moving the robotic arm over the object, adjusting the robotic arm posture, grasping the object, and performing placement operations.

[0152] After receiving the instruction generated by Wenxinyiyan, the brain-computer interface interaction system proposed in the present invention breaks it down into multiple specific control actions and executes them step by step through the robotic arm control module.

[0153] For example, the JSON response data generated by Wenxinyiyan contains the following: 1) Movement command: such as "move 5cm in the positive direction of the X axis".

[0154] 2) Posture adjustment: such as "rotate the manipulator 45 degrees".

[0155] 3) Crawl instructions: such as "execute crawl".

[0156] 4) Placement instructions: such as "move to the target location and place it".

[0157] These control commands dynamically generate flashing blocks through the SSVEP control interface, and users can select specific operation steps by looking at different blocks. The instructions generated by the Wenxin Yiyan model can adapt to tasks of different complexity. The brain-computer interface interaction system proposed in this invention dynamically adjusts the layout and control options of the SSVEP control interface according to task requirements to ensure that users can intuitively operate the robotic arm when performing complex tasks.

[0158] For example, for simple tasks such as "move", the system will display basic directional control options, while for complex tasks such as "grab and place", the system will display more subtasks and operation steps to help users complete the task step by step.

[0159] The brain-computer interface interaction system based on the generative large model proposed in the present invention is used as a drone control system, wherein the device control module includes a drone control module: The drone control module is used to obtain detailed information of the drone and send it to the large model end, and perform corresponding control operations on the drone according to the device control instructions selected by the user; The large model side generates feedback information matching the current mission scenario based on the detailed information of the drone combined with the natural language request instruction, and the feedback information includes flight path planning, speed adjustment and mission execution. The detailed information can be the current environment and mission requirements.

[0160] In terms of drone control, the system also makes use of Wenxin Yiyan's semantic understanding and task reasoning capabilities.

[0161] Drone control usually involves complex operations such as flight path planning, speed adjustment, mission execution (such as taking pictures or transporting items), etc. Users can input instructions through the SSVEP speller, such as "let the drone fly to a specified location and take pictures" or "plan a flight path."

[0162] The Wenxin Yiyanda model can perform semantic analysis on these instructions, understand the user's task requirements, and generate corresponding control steps. Similar to the control of the robotic arm, the Wenxin Yiyanda model will automatically disassemble complex drone tasks and generate specific control commands such as flight path, speed adjustment, and task execution.

[0163] For example, when the user inputs "let the drone fly to a specified location and take photos", the Wenxin Yiyanda model will generate the following control steps based on the current environment and task requirements: 1) Flight path planning: such as "fly 100 meters north".

[0164] 2) Speed ​​adjustment: such as "set the flight speed to 5 m / s".

[0165] 3) Task execution: such as “take a photo after arriving at the specified location”.

[0166] These control steps are presented through the SSVEP control interface, and users can select flight paths, adjust speeds, or perform shooting tasks by looking at different control blocks. The system generates control instructions through the Wenxin Yiyanda model to ensure the accuracy and intelligence of the drone's mission.

[0167] For example, when the instructions entered by the user are incomplete or ambiguous, the Wenxin Yiyanda model can automatically complete the task details through contextual reasoning, such as intelligently selecting the best path in flight path planning, or adjusting flight parameters according to environmental conditions during mission execution.

[0168] In addition, the real-time performance of the Wenxin Yiyan large model in UAV missions is also fully guaranteed. Since UAV flight missions require continuous real-time control operations, the device information acquisition unit of the device control module will monitor and obtain the flight status of the UAV (such as speed, attitude, position, etc.) in real time, and the device control module will transmit this information to the large model end.

[0169] The Wenxin Yiyan big model can dynamically generate adaptive feedback information based on the real-time state change information collected during flight. For example, when the drone encounters obstacles or environmental changes during flight, the device information collection unit updates the obstacle information and environmental parameters to the big model in real time. The big model quickly generates corresponding countermeasures as real-time feedback information through the semantic understanding and reasoning of the Wenxin Yiyan big model. The real-time feedback information includes obstacle avoidance path planning or flight altitude adjustment instructions.

[0170] The user interface module uses real-time feedback information to display in the form of dynamic blocks through the SSVEP control interface. Users can quickly select the corresponding response operation by looking at the flashing blocks, thereby achieving real-time intervention and control of the drone to ensure its safe flight and complete the mission as expected.

[0171] In summary, the application of the Wenxin Yiyan large model in this system not only supports complex task control of equipment such as robotic arms and drones, but also improves the real-time and accuracy of control instructions through deep collaboration between the equipment control module and the large model. Combined with the dynamically updated SSVEP control interface and the intelligent semantic reasoning function of the Wenxin Yiyan large model, the system significantly improves the efficiency, stability and user experience of task execution, and can better adapt to a variety of task scenarios and equipment types.

[0172] The brain-computer interface interaction system proposed in the present invention can dynamically generate SSVEP control interfaces that adapt to different task scenarios, and automatically adjust the interface layout and control options according to task requirements, thereby providing users with a simple and efficient operation method. Through the deep integration of the Wenxin Yiyan model, the system has demonstrated great flexibility and intelligent advantages in the fields of robotic arms and drone control, allowing users to easily perform complex operation tasks, and improving the feasibility and scalability of brain-computer interface technology in practical applications.

[0173] The present invention proposes a brain-computer interface interaction system based on a generative large model, which is deeply integrated with the large model API. The Chinese sentence instructions are input through the SSVEP speller, and the Wenxinyiyan large model API is called to generate an SSVEP control interface adapted to various scenarios according to the sentences input by the user. Relying on the intelligent reasoning ability of the large model, the flashing mode and interactive layout of the SSVEP control interface can be automatically designed, and can be customized according to the user's usage habits, scenarios and device characteristics, which improves the flexibility of the system.

[0174] The brain-computer interface interaction system based on generative large models proposed in this invention is not only an upgrade of the existing brain-computer interface technology, but also a high-level exploration, dedicated to promoting brain-computer interface technology to a wider, intelligent and personalized application field: 1) Cross-domain intelligent control center: Through the combination with the big model, the system, as a cross-domain intelligent control center, can play an important role in smart home, industrial automation and personal assistant devices. The system's dynamic SSVEP control interface generation and intelligent task reasoning provide users with a convenient and efficient interaction method; 2) Inclusive technology applications: With the enhancement of multi-language support, it can adapt to the needs of global users, reduce language barriers, and help more people improve their quality of life through brain-computer interface technology. In addition, the scalability of the system enables it to quickly adapt to new devices and emerging technologies in the future, and has strong application prospects; 3) Personalized interactive experience: Based on the user's usage data and preferences, this system can optimize the generated SSVEP paradigm to achieve truly personalized interaction. Whether it is home control, robotic arm task execution, or drone flight, the system can provide the most suitable operation interface for users; 4) Promote the popularization and intelligent development of brain-computer interface technology: Through the innovative design of this system, brain-computer interface technology is no longer limited to simple interactive operations, but is moving towards more complex task execution and intelligent reasoning. This system not only improves the efficiency of existing brain-computer interface technology, but also provides a foundation for future intelligent applications, laying an important foundation for the popularization and intelligent development of brain-computer interface technology.

[0175] Although the above methods are illustrated and described as a series of actions for simplicity of explanation, it should be understood and appreciated that these methods are not limited by the order of the actions, because according to one or more embodiments, some actions may occur in a different order and / or concurrently with other actions from those illustrated and described herein or not illustrated and described herein but understandable to those skilled in the art.

[0176] As shown in this application and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0177] In the description of the present invention, it should be noted that the directions or positional relationships indicated by terms such as “upper”, “lower”, “front”, “back”, “left”, “right”, “vertical”, “inside” and “outside” are based on the directions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore cannot be understood as a limitation on the present invention.

[0178] In the present invention, unless otherwise clearly specified and limited, a first feature being "above" or "below" a second feature may include that the first and second features are in direct contact, or may include that the first and second features are not in direct contact but are in contact through another feature between them. Moreover, a first feature being "above", "above" and "above" a second feature includes that the first feature is directly above and obliquely above the second feature, or simply indicates that the first feature is higher in level than the second feature. A first feature being "below", "below" and "below" a second feature includes that the first feature is directly below and obliquely below the second feature, or simply indicates that the first feature is lower in level than the second feature.

[0179] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be directly connected, or indirectly connected through an intermediate medium, or it can be the connection inside two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0180] Those skilled in the art will appreciate that information, signals, and data may be represented using any of a variety of different technologies and techniques. For example, data, instructions, commands, information, signals, bits, symbols, and chips cited throughout the above description may be represented by voltage, current, electromagnetic waves, magnetic fields or magnetic particles, optical fields or optical particles, or any combination thereof.

[0181] Those skilled in the art will further appreciate that the various illustrative logic blocks, modules, circuits, and algorithm steps described in conjunction with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or a combination of the two. To clearly illustrate this interchangeability of hardware and software, various illustrative components, blocks, modules, circuits, and steps are generally described above in terms of their functionality. Whether such functionality is implemented as hardware or software depends on the specific application and the design constraints imposed on the overall system. The technician may implement the described functionality in different ways for each specific application, but such implementation decisions should not be interpreted as resulting in a departure from the scope of the present invention.

[0182] The various illustrative logic modules and circuits described in conjunction with the embodiments disclosed herein may be implemented or executed with a general purpose processor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, or any combination thereof designed to perform the functions described herein. A general purpose processor may be a microprocessor, but in an alternative, the processor may be any conventional processor, controller, microcontroller, or state machine. The processor may also be implemented as a combination of computing devices, such as a combination of a DSP and a microprocessor, a plurality of microprocessors, one or more microprocessors in cooperation with a DSP core, or any other such configuration.

[0183] The above embodiments are provided for persons familiar with the art to implement or use the present invention. Personnel familiar with the art can make various modifications or changes to the above embodiments without departing from the inventive concept of the present invention. Therefore, the protection scope of the present invention is not limited to the above embodiments, but should be the maximum scope of the innovative features mentioned in the claims.

Claims

1. A brain-computer interface interaction system based on a generative large model, characterized in that: Including user interface module, EEG equipment, data transmission module, data processing and analysis module, large model terminal and equipment control module: The user interaction interface module is used to obtain the user's control intention using steady-state visual evoked potential technology, and dynamically generate a steady-state visual evoked potential control interface based on the feedback information of the large model end for the user to select to generate corresponding device control instructions; The electroencephalogram device is used to collect the electroencephalogram signal of the user; The data transmission module is used to transmit the EEG signals collected by the EEG device to the data processing and analysis module; The data processing and analysis module processes and analyzes the EEG signal, and then sends the decoded natural language request instruction to the large model end, and sends the decoded device control instruction to the device control module; The large model end uses a generative large model to perform semantic understanding and reasoning recognition on the input natural language request instructions, generates feedback information of the current task scenario according to user needs, and sends it to the user interaction interface module via the data processing and analysis module; The device control module performs corresponding control operations on the device according to the device control instruction selected by the user.

2. The brain-computer interface interaction system based on the generative large model according to claim 1 is characterized in that: The large model end adopts the Wenxinyiyan large model.

3. The brain-computer interface interaction system based on the generative large model according to claim 1 or claim 2, characterized in that: The user interface module is a steady-state visual evoked potential speller: The steady-state visual evoked potential speller dynamically generates a steady-state visual evoked potential control interface according to feedback information from the large model end.

4. The brain-computer interface interaction system based on the generative large model according to claim 1 or claim 2, characterized in that: The data processing and analysis module is a host computer or a cloud server.

5. The brain-computer interface interaction system based on the generative large model according to claim 1 or claim 2, characterized in that: The data processing and analysis module uses a task discriminant component analysis algorithm to decode and identify the collected electroencephalogram signals and generate natural language request instructions.

6. The brain-computer interface interaction system based on generative large model according to claim 5 is characterized in that: The task discriminant component analysis algorithm further comprises the following steps: Combining the original EEG signal with the time-delayed EEG signal to construct enhanced training data and test data; Project the augmented training and test data into the subspace spanned by the reference signal; A two-dimensional linear discriminant analysis was performed on the training trials to find the projection direction to discriminate trials across all categories; Use Fisher's criterion to derive the projection direction; By using the idempotence of the projection matrix, the process of solving the projection direction is simplified to obtain the final projection matrix. Using the obtained projection matrix, the correlation coefficient between the single test data and the matrix generated by the universal filter is calculated to identify the frequency components of the signal; The target frequency was determined by analyzing the maximum value of the correlation coefficient, thereby identifying the stimulation frequency of the steady-state visual evoked potential signal.

7. The brain-computer interface interaction system based on the generative large model according to claim 1 or claim 2, characterized in that: The user interaction interface module determines the number of devices by analyzing the feedback information of the large model end, generates a corresponding number of steady-state visual evoked potential flashing blocks based on the number of devices, and assigns a different flashing frequency to each steady-state visual evoked potential flashing block.

8. The brain-computer interface interaction system based on the generative large model according to claim 1 or claim 2, characterized in that: The user interaction interface module determines the specific control task of the device through the feedback information of the large model end, and generates a corresponding dynamic steady-state visual evoked potential control interface according to the specific control task.

9. The brain-computer interface interaction system based on the generative large model according to claim 1 or claim 2, characterized in that: The user interaction interface module is used to generate text information for controlling the device and dynamically adjust the flashing frequency, color and position layout of the flashing blocks, and automatically calculate and determine the optimal layout of the steady-state visual evoked potential control interface so that each flashing block is located at the optimal position within the user's field of view.

10. The brain-computer interface interaction system based on a generative large model according to claim 1 or claim 2, characterized in that: The device control module also includes a device information collection unit: The device information acquisition unit is used to obtain device status information in real time; The device control module is connected to the large model end and sends the real-time acquired device status information to the large model end; The large model side generates corresponding real-time feedback information based on the device status information acquired in real time; The user interaction interface module dynamically adjusts the steady-state visual evoked potential control interface based on real-time feedback information from the large model end.

11. The brain-computer interface interaction system based on a generative large model according to claim 1 or claim 2, characterized in that: The large model side supports voice input control instructions in multiple languages.

12. The brain-computer interface interaction system based on a generative large model according to claim 1 or claim 2, characterized in that: The device control module includes a home device control module: The home device control module is used to obtain detailed information of connectable home devices and send it to the large model end, and perform corresponding control operations on the home devices according to the device control instructions selected by the user; The large model side generates feedback information matching the current task scenario based on the detailed information of the device and the historical operation data of the user, and the feedback information includes the device name, device function and device status.

13. The brain-computer interface interaction system based on a generative large model according to claim 1 or claim 2, characterized in that: The equipment control module includes a robotic arm control module: The robot arm control module is used to obtain detailed information of the robot arm and send it to the large model end, and perform corresponding control operations on the robot arm according to the device control instructions selected by the user; The large model end generates feedback information matching the current task scenario based on the detailed information of the robotic arm combined with natural language request instructions, and the feedback information includes movement commands, posture adjustments, grasping instructions and placement instructions.

14. The brain-computer interface interaction system based on the generative large model according to claim 1 or claim 2, characterized in that: The device control module includes a drone control module: The drone control module is used to obtain detailed information of the drone and send it to the large model end, and perform corresponding control operations on the drone according to the device control instructions selected by the user; The large model side generates feedback information matching the current mission scenario based on the detailed information of the drone and natural language request instructions. The feedback information includes flight path planning, speed adjustment, and mission execution.

Images