Child concentration training system and method based on low-cost brain-computer interface

Abstract

The invention relates to the technical field of brain-computer interfaces, and discloses a child concentration training system and method based on a low-cost brain-computer interface, and the system comprises an electroencephalogram signal subsystem, a feature transmission subsystem and a concentration training subsystem which are in communication connection in sequence. The method corresponds to the system. According to the method, based on coarse-grained electroencephalogram signals collected by low-cost brain-computer interface equipment, fine-grained brain-robot interaction and extensible multi-robot cooperation are achieved, high precision is achieved with low computing resources, multiple users control multiple robots to complete specified actions for concentration training, and the training efficiency is improved. The defects and deficiencies of an existing concentration training system are overcome; the equipment is low in cost, simple in composition, convenient and comfortable to wear, easy to popularize and civil, and capable of realizing rapid, efficient and human-centered brain-computer interaction.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of brain-computer interfaces, and particularly relates to a child concentration training system and method based on a low-cost brain-computer interface. BACKGROUND

[0002] In recent years, with the development of electroencephalogram technology, brain-computer interface technology is widely applied to the processing of various tasks. The brain-computer interface has important significance in medical rehabilitation, game entertainment, human-computer interaction and education training and other fields. A promising application is to detect the concentration and relaxation state of the brain of a child through a brain-computer interface, so as to train the concentration of the child. At the same time, the existing high-fidelity brain-computer interface is usually expensive, which brings great challenges to the popularization of civilian brain-computer systems, and cheaper devices can only collect coarse-grained signals, which hinders the practical application of the training concentration device due to the lack of time resolution and accuracy. At the same time, the existing methods for training the concentration of children are mostly dull and boring, and a system can only train one target, which may have a negative impact on the training effect. Therefore, the existing methods are difficult to meet the requirements of low cost, high accuracy and multi-target use at the same time. SUMMARY

[0003] The application aims to provide a child concentration training system and method based on a low-cost brain-computer interface, so as to solve the technical problem that the attention training based on the brain-computer interface in the prior art is difficult to meet the requirements of low cost, high accuracy and multi-target use at the same time.

[0004] To achieve the above-mentioned purpose, the application provides a child concentration training system based on a low-cost brain-computer interface, which comprises an electroencephalogram subsystem, a feature transmission subsystem and a concentration training subsystem which are sequentially connected in communication. The electroencephalogram subsystem is configured to collect and input original electroencephalogram signals, process the original electroencephalogram signals based on a preset signal and processing algorithm, and transmit the processing result to the feature transmission subsystem; wherein a plurality of BrainLink mental force headbands are used for collecting the original electroencephalogram signals, and the trainee wears the BrainLink mental force headband; The feature transmission subsystem is configured to be provided with a transmitter module and a wireless channel module, the transmitter module is used for feature extraction and coding of the processing result, the wireless channel module is used for placing the coding result in an additive white Gaussian noise channel environment for transmission, and the end-to-end transmission performance is optimized through joint cross-entropy and mutual information loss function; The concentration training subsystem is configured to convert the encoding result into a command through a preset channel decoder and a semantic decoder, specify the robot and its action based on the command, and execute the command only when the trainee is in a preset attention state, thereby performing concentration training.

[0005] Preferably, the original EEG signal is processed based on a preset signal and processing algorithm, including: Down-sampling the original EEG signal; Introducing Gaussian noise to enhance the signal; at the same time, simulating data interference caused by noise or sensor disconnection.

[0006] Preferably, when the transmitter module extracts features and encodes the processing result, EEG data is collected over time, including: Dynamic EEG feature encoder is used for feature extraction, and the dynamic EEG feature encoder also serves as a semantic encoder; for the EEG signal collected by BrainLink, the encoder selects a processing path optimized for low-dimensional high-noise data, and introduces an intermittent mask mechanism to simulate incomplete signals by adding random masks in the signal; At the same time, a sparse attention mechanism is used to capture long-range dependencies in long-time sequence inputs; and a model compression technique is used to reduce storage occupancy and computational overhead.

[0007] Preferably, the joint cross-entropy and mutual information loss function The mathematical expression is: The end-to-end transmission performance optimization based on the joint cross-entropy and mutual information loss function includes: In the first stage, mutual information is maximized To improve channel capacity and data rate; in the second stage, cross-entropy is used to constrain the consistency of channel encoding and decoding; in the third stage, cross-entropy is used again to ensure the faithful recovery of semantic information in the encoding and decoding process; wherein, And are preset weight parameters.

[0008] Preferably, the attention state includes a non-transition state and a transition state. The non-transition state includes sustained concentration and sustained relaxation. The transition state includes concentration-relaxation and relaxation-concentration.

[0009] Preferably, the command is executed only when the trainee is in a preset attention state, specifically: each attention state corresponds to different actions of different robots.

[0010] As preferred, when the attention state is in the transition state, a preset intervention strategy is cut in based on the command, the intervention strategy being used for early intervention between the transition state and the non-transition state.

[0011] To achieve the above object, the application further provides a child concentration training method based on a low-cost brain-computer interface, applied to the child concentration training system based on the low-cost brain-computer interface as described above, and the method comprises the following steps: The original electroencephalogram signal is collected and input, the original electroencephalogram signal is processed based on a preset signal and processing algorithm, and the processing result is transmitted to the feature transmission subsystem; wherein the original electroencephalogram signal is collected by using multiple BrainLink mental force headbands, and the trainee wears the BrainLink mental force headband; The transmitter module and the wireless channel module are provided, the transmitter module is used for feature extraction and coding of the processing result, the wireless channel module is used for transmission of the coding result in an additive white Gaussian noise channel environment, and the end-to-end transmission performance is optimized by joint cross-entropy and mutual information loss function; The coding result is converted into a command by using a preset channel decoder and a semantic decoder, the robot and its action are specified based on the command, and the command is executed only when the trainee is in a preset attention state, so as to perform the concentration training.

[0012] To achieve the above object, the application further provides a child concentration training computer device based on a low-cost brain-computer interface, comprising at least one processor, at least one memory and a data bus; The processor and the memory complete mutual communication through the data bus; The memory stores program instructions executable by the processor, and the processor calls the program instructions to execute the child concentration training method based on the low-cost brain-computer interface as described above.

[0013] To achieve the above object, the application further provides a storage medium having a computer program stored thereon, the computer program being executed by the processor to implement the child concentration training method based on the low-cost brain-computer interface as described above.

[0014] Beneficial effects: compared with the prior art, the child concentration training system and method based on a low-cost brain-computer interface provided in the application only needs to collect coarse-grained electroencephalogram signals by using a low-cost brain-computer interface device, but supports fine-grained brain-robot interaction and scalable multi-robot cooperation, and satisfactory precision is achieved with low computing resources, the specified action is completed by using multi-user control multi-robot for concentration training, and the shortcomings and deficiencies of the prior art concentration training system are overcome; the system has low device cost, simple composition, convenient and comfortable wearing, is easy to popularize for civilian use, can realize fast, efficient and human-centered brain-computer interaction, the whole process does not need intervention, has low time cost and high automation degree, a non-invasive brain-computer interface device is used, no adverse effects are caused on the human body, dry electrodes are used in the device, and the device can be repeatedly used, so that resource waste is avoided and the cost is further reduced. BRIEF DESCRIPTION OF DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or the prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0016] Figure 1 The structural block diagram of the child concentration training system based on a low-cost brain-computer interface provided in the embodiments of the present application is shown in the figure. Figure 2 The feature transmission flowchart based on the feature transmission subsystem provided in the embodiments of the present application is shown in the figure. Figure 3 The time sequence coordinate graph of four different attention states provided in the embodiments of the present application is shown in the figure, wherein (a) is sustained relaxation, (b) is relaxation-concentration, (c) is concentration-relaxation, and (d) is sustained concentration. Figure 4 The flowchart of the child concentration training method based on a low-cost brain-computer interface provided in the embodiments of the present application is shown in the figure.

[0017] The implementation, functional features and advantages of the present application will be further described with reference to the embodiments and the drawings. DETAILED DESCRIPTION

[0018] The technical solutions in the embodiments of the present application will be clearly and completely described below. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0019] In this document, the term "comprising" is intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0020] Brain-computer interface (BCI) technology is a key technology widely used in fields such as medicine, education, and entertainment. As a bridge technology integrating biological intelligence and artificial intelligence, BCI is driving disruptive changes in these fields. A method most similar to this embodiment is a children's attention training system based on multimodal neural fusion. This training system requires the child to wear a composite head-mounted device integrating 32-channel EEG electrodes and an eye-tracking module. It synchronously collects beta-wave oscillation signals from the prefrontal cortex and pupil diameter changes, and uses a support vector machine (SVM) algorithm to establish an attention intensity prediction model. During training, the system dynamically adjusts the difficulty of cognitive tasks based on real-time calculated attention indices and provides neural feedback through a VR headset. However, this solution has significant drawbacks: high-density EEG devices require professional personnel to perform electrode impedance calibration, which takes more than 15 minutes per calibration, and the VR interaction module results in a single system cost of up to $23,000; multimodal data fusion relies on laboratory-grade GPU computing platforms (such as NVIDIA A100), which has high deployment and maintenance complexity; more importantly, beta wave signals are susceptible to electromyography artifacts (such as micro-movements of children's heads), which severely restricts the universal application in home and school settings.

[0021] Existing methods for training children's attention mainly include systems based on invasive brain-computer interface implantation, systems based on non-invasive EEG headbands with wet electrodes, and systems based on electronic devices. The shortcomings of these methods will be analyzed in this paper.

[0022] (1) Disadvantages of invasive brain-computer interface systems Invasive brain-computer interfaces (BCIs) directly acquire neural signals by implanting cortical electrodes, offering high precision. However, this method has significant drawbacks: First, it requires craniotomy to implant the electrode array, posing a risk of infection. Clinical data shows a postoperative infection rate of approximately 3.7%, and the development of the child's skull may lead to electrode displacement, causing irreversible damage to the child's health. Second, the system relies on laboratory-grade neural signal processing equipment, with a single set costing over $500,000, and requires weekly impedance calibration by a professional neuroscientist, making it difficult to popularize in home or school settings.

[0023] (2) Disadvantages of the wet electrode-based non-invasive EEG headband system The mainstream consumer-grade wet electrode EEG headband currently reduces the contact impedance between the electrode and the scalp through conductive gel, which can improve the signal-to-noise ratio of the brain electrical signal and monitor the intensity of beta waves to assess concentration. However, it has significant drawbacks in practical application: first, the conductive gel maintenance is complex. Wet electrodes need to be regularly replenished with electrolytic gel to maintain stable impedance. In the context of children's training, frequent head movements can accelerate gel loss, requiring interruptions for reapplication, disrupting training continuity. Second, long-term stability defects. After four weeks of continuous use, the silver chloride plating of the wet electrode is prone to oxidation and corrosion, requiring regular replacement of the electrode (single cost of about $200), and the growth of children's head circumference (1.2-1.5 cm per year) may cause the electrode array to lose contact with the scalp.

[0024] (3) Based on the shortcomings of the electronic device system Some high-end solutions use systems based on smartwatches, smartphones, tablets, and other electronic devices to train children's concentration, but they have the following drawbacks: first, these modules increase the overall system cost and increase power consumption, which contradicts the original intention of low cost; second, electronic devices are addictive, and long-term use of electronic devices can have adverse effects on children's physical and mental development, especially for children with ADHD, which may further exacerbate symptoms; third, concentration in the virtual world of electronic games may not transfer to reality.

[0025] In summary, to address the pain points of existing technologies, the present embodiment discloses a low-cost brain-computer interface-based children's concentration training system and method. In short, the present embodiment has the characteristics of low device cost, strong portability, comfortable wearing, simple composition, high accuracy, etc. It uses a non-invasive dry electrode EEG headband to collect brain electrical signals, preventing adverse health effects on younger users, while avoiding the need for frequent replacement of wet electrodes, avoiding resource waste, and reducing the cost of concentration training.

[0026] The low-cost brain-computer interface-based child concentration training system disclosed in the embodiment is briefly summarized as follows: the embodiment is a system for collecting electroencephalogram signals by using a low-cost brain-computer interface device to evaluate relaxation and concentration, so as to train concentration. The system uses a low-cost Brainlink mental force headband (only about 700 yuan per device) to replace a device with a price of tens of thousands of yuan or even hundreds of thousands of yuan, thereby reducing the hardware cost. The basic principle of realizing brain concentration and relaxation evaluation is the difference in neural oscillation frequency and spatial distribution of the brain in different cognitive states. The beta wave power of the prefrontal lobe is significantly improved when concentrating, and the amplitude of the occipital alpha wave in the closed-eye relaxation state is 3-5 times that in the open-eye state. Specifically, first, the low-cost brain-computer interface device is used to synchronously collect electroencephalogram signals of multiple users, and a Gaussian noise layer is added to the original signals to simulate real environment interference. The power frequency noise and electromyographic artifacts are removed through band-pass filtering (such as 0.5-40 Hz). Then, combined with the multi-head attention and residual structure, the semantic features of the low signal-to-noise ratio signals are retained, and joint feature enhancement and semantic coding compression are performed. Finally, the decoder is deployed at the robot end to restore the electroencephalogram signal instructions in real time, and the control of the robot is realized. The different attention states of the user are classified, each category corresponds to an action of the robot, and the user controls the robot to make the specified action through the electroencephalogram device, so as to train the concentration.

[0027] The low-cost brain-computer interface-based child concentration training system disclosed in the embodiment is described in detail.

[0028] Reference Figure 1 , Figure 1 The structure block diagram of the low-cost brain-computer interface-based child concentration training system provided by the embodiment of the application is shown in the figure.

[0029] As Figure 1 shown, the embodiment discloses a low-cost brain-computer interface-based child concentration training system, which comprises an electroencephalogram signal subsystem 10, a feature transmission subsystem 20 and a concentration training subsystem 30 which are sequentially and communicatively connected.

[0030] Briefly from the timing, the low-cost brain-computer interface-based child concentration training system in the actual application of the embodiment: in the first part, multiple users wear brain-computer interface devices, and collect the concentration information of the electroencephalogram signals thereof; in the second part, the data is sent to the server for feature extraction, and then the extracted features are transmitted through a wireless channel; in the third stage, the transmitted electroencephalogram signals are decoded, and the robot makes actions corresponding to the instructions; through the training of the concentration of the user, the cooperative work between different robots is achieved.

[0031] The electroencephalogram signal subsystem 10 is configured to collect and input raw electroencephalogram signals, process the raw electroencephalogram signals based on a preset signal and processing algorithm, and transmit the processing result to a feature transmission subsystem; wherein a plurality of BrainLink mental force headbands are used for collecting the raw electroencephalogram signals, and the trainee wears the BrainLink mental force headband.

[0032] Specifically, the raw electroencephalogram signals are processed based on a preset signal and processing algorithm, including: The raw electroencephalogram signals are down-sampled and processed; Gaussian noise is introduced to enhance the signals; at the same time, data interference caused by noise or disconnection of the sensor is simulated.

[0033] In the specific application of the embodiment, the electroencephalogram signal collection subsystem uses a plurality of BrainLink mental force headbands as data collection devices, and the BrainLink mental force headband is composed of a core control unit, a headband and a metal dry electrode. The three-channel metal dry electrode array is arranged in the inner side of the headband, and the impedance thereof is lower than 10 kΩ. The headband can directly contact the scalp to collect electroencephalogram signals without using conductive gel, and has optimized sensitivity in the low frequency and high frequency bands. The core control unit integrates a main control chip, an analog-to-digital converter, a Bluetooth communication module and a motion sensor. The main control chip is based on the ARM Cortex-M4F architecture, is responsible for coordinating the data flow between the analog-to-digital conversion, Bluetooth transmission and the motion sensor, and realizes real-time processing of multiple tasks. The analog-to-digital converter converts the analog electroencephalogram signals into digital signals at a fixed sampling rate, ensuring high-precision data collection. The Bluetooth module is responsible for transmitting electroencephalogram signal feature data to the upper computer; and the motion sensor detects head movement by integrating a gyroscope, generates a noise reference signal, and identifies the wearing and falling off state of the device.

[0034] Before the signal enters the core model, the system performs key preprocessing and data enhancement steps to improve the quality of the data and the robustness of the model. First, down-sampling: the raw electroencephalogram signals are down-sampled and processed to reduce the data dimension while retaining the effective information, and to reduce the subsequent computational overhead. Second: Gaussian noise is introduced to enhance the signals. This operation aims to simulate signal interference in the real collection environment, thereby expanding the training data set and improving the generalization ability and robustness of the model in a noisy environment. At the same time, the data interference caused by noise or disconnection of the sensor is simulated to enhance the processing ability of the model for incomplete data. After the collected raw electroencephalogram signals are processed, they are transmitted in a serialized and concurrent manner through a Bluetooth serial port to an edge mobile server or an edge portable server in real time, so as to enter the next processing stage.

[0035] The feature transmission subsystem 20 is configured to be provided with a transmitter module for feature extraction and encoding of the processing result and a wireless channel module for placing the encoded result in an additive white Gaussian noise channel environment for transmission and optimizing end-to-end transmission performance through a joint cross-entropy and mutual information loss function.

[0036] Specifically, when the transmitter module performs feature extraction and encoding on the processing result, the electroencephalogram data is collected over time, including: The dynamic electroencephalogram feature encoder is used for feature extraction, and the dynamic electroencephalogram feature encoder also serves as a semantic encoder; for the electroencephalogram signals collected by BrainLink, the encoder selects a processing path optimized for low-dimensional high-noise data, and introduces an intermittent mask mechanism to simulate incomplete signals by adding random masks in the signals. At the same time, a sparse attention mechanism is used to capture long-range dependencies in long-time sequence inputs; and a model compression technique is used to reduce storage occupation and computational overhead.

[0037] Specifically, the joint cross-entropy and mutual information loss function has a mathematical expression as follows: The optimization of end-to-end transmission performance based on the joint cross-entropy and mutual information loss function includes: In the first stage, the mutual information is maximized to improve channel capacity and data rate; in the second stage, the cross-entropy is used to constrain the consistency of channel encoding and decoding; and in the third stage, the cross-entropy is used again to ensure the faithful recovery of semantic information in the encoding and decoding process; wherein, and are preset weight parameters.

[0038] Referring to Figure 2 , Figure 2 is a feature transmission flowchart based on the feature transmission subsystem provided by the embodiments of the present application.

[0039] In actual application, the feature transmission subsystem can be divided into a transmitter module and a wireless channel module. The flow of the feature transmission subsystem is as shown in Figure 2 .

[0040] For the transmitter module, the transmitter module is responsible for feature extraction and encoding of the brain electrical signals collected by the BrainLink device, and can be deployed on an edge mobile server or a resource-limited edge portable server. The transmitter module receives the raw brain electrical signals from the BrainLink, which has the characteristics of low dimension and high noise. Since the brain electrical data collected at a single time point is not very reliable due to the noise in the collection process, we collect brain electrical data over time to more accurately represent brain activity. The signal is first pre-processed, including downsampling and adding Gaussian noise for data enhancement. Then, a dynamic electroencephalogram feature encoder (also serving as a semantic encoder) is used for feature extraction. According to the characteristics of the brain electrical signals collected by the BrainLink, the encoder selects a processing path optimized for low-dimensional high-noise data, and introduces an intermittent mask mechanism to simulate incomplete signals by adding random masks to the signals, thereby enhancing the model's ability to process incomplete data. At the same time, in order to solve the problem of high time and space complexity of the self-attention mechanism when processing high-dimensional data, the embodiment adopts a sparse attention mechanism, which effectively captures long-range dependencies in long time series input while reducing computation and storage requirements. In order to realize lightweight deployment on the edge portable server, the system uses model compression techniques including weight pruning, filter pruning, and weight sharing and half-precision quantization, which significantly reduces the storage occupancy and computational overhead of the model.

[0041] For the wireless channel module, the system transmits the encoded brain electrical semantic features in an additive white Gaussian noise (AWGN) channel environment. To evaluate and optimize the end-to-end transmission performance under this specific channel condition, the system is trained using the joint cross-entropy and mutual information loss function as described above, which is optimized from three levels. After multiple iterations, the system can effectively reduce the semantic difference between the encoded features and the decoded features, significantly improve the transmission robustness and semantic fidelity in a noisy environment, and thus overcome the information loss problem caused by low signal-to-noise ratio in traditional communication.

[0042] The concentration training subsystem 30 is configured to convert the encoding result into a command through a preset channel decoder and a semantic decoder, specify a robot and its action based on the command, and execute the command only when the trainee is in a preset attention state, thereby performing concentration training.

[0043] Specifically, the attention state includes a non-transition state and a transition state. The non-transition state includes sustained concentration and sustained relaxation. The transition state includes concentration-relaxation and relaxation-concentration.

[0044] Specifically, the command is only executed when the trainee is in a preset attention state, specifically: each attention state corresponds to different actions of different robots.

[0045] Specifically, when the attention state is in the transition state, the preset intervention strategy is cut in based on the command, and the intervention strategy is used for early intervention between the transition state and the non-transition state.

[0046] In the specific application of the embodiment, the concentration training subsystem starts when the transmitted electroencephalogram signal features arrive at the receiver module, and the communication device is the main part of the module, which can be embedded in each robot or an independent device, including a decoder and a command code mapping module. In the concentration training subsystem, the features are sent to each semantic communication device. Subsequently, the channel decoder and the semantic decoder convert the features into commands using code mapping. These commands specify the robots and their actions. Then, the robots perform the specified actions according to these commands, and through training the user's concentration, the robots complete the tasks specified in the specified attention state. Unlike the traditional two-class method of concentration-relaxation, we divide the user's attention state into four states: sustained concentration, concentration-relaxation, relaxation-concentration, and sustained relaxation, each of which corresponds to different actions of different robots. The innovation of this methodology has the following significant advantages over the traditional two-class method: The attention state classification of the embodiment better fits the dynamics and continuity of cognitive states, improving the ecological validity of state recognition. The traditional two-class method forcibly simplifies the complex attention dynamic process into two discrete and mutually exclusive static categories, which ignores the inherent volatility and continuity of cognitive states. By introducing the two transition states of "concentration-relaxation" and "relaxation-concentration", this framework can more accurately capture and describe the dynamic trajectory of the user's attention as it transitions between high and low levels. This not only better aligns with the neurophysiological basis of human cognition, but also greatly improves the ecological validity of the state recognition model in real-world scenarios, making the interpretation of the user's internal state more realistic and nuanced.

[0047] The attention state classification of the embodiment realizes the context awareness and forward-looking regulation of human-computer interaction. The existing robot response under the two-class method model is often reactive, i.e., only after detecting "concentration" or "relaxation" does it trigger the corresponding action, lacking prediction of trends. By recognizing the transition state, the robot system has the ability of forward-looking interaction. In one specific example of the embodiment, for the intervention strategy, when the "concentration-relaxation" transition state is recognized, the robot can actively intervene and provide a gentle reminder or introduce a short interactive task before the user's attention is completely dispersed, to prevent complete interruption of attention. This intervention strategy based on state trends upgrades from post-repair to pre-prevention and mid-guidance, significantly improving the intelligence level of interaction.

[0048] Referring to Figure 3 , Figure 3 The time sequence coordinate graphs of four different attention states provided by the embodiment of the application.

[0049] As Figure 3 shown in the figure: (a) shows the time sequence of different attention states in the sustained relaxation state in the non-transition state; (b) shows the time sequence of different attention states in the relaxation-focus state in the transition state; (c) shows the time sequence of different attention states in the focus-relaxation state in the transition state; (d) shows the time sequence of different attention states in the sustained focus state in the non-transition state; wherein the more concentrated the attention is, the higher the focus score is, and the more relaxed, the lower the score is, and the red dotted box represents that the attention changes between focus and relaxation in this interval. Based on this, the division of the four types of states is realized, thereby providing a data basis for the mapping of commands in different states.

[0050] Now, the low-cost brain-computer interface-based children's concentration training system disclosed in the embodiment will be verified in combination with a specific application.

[0051] In the multi-brain-multi-robot concentration training experimental scene, we conduct experiments by wearing electroencephalogram acquisition devices on three users. In this test, the three users wear different types of electroencephalogram acquisition devices, control and coordinate three types of robots to complete various tasks through the change of concentration. For example, user 1 moves the mechanical arm of the robot to grab the object and place it on the wheeled robot, user 2 controls the wheeled robot to transport the object to the side of the nearby humanoid robot, and user 3 controls the humanoid robot to pick up the object and transport it to a remote place. We set that this series of collaborative actions must be completed when the user is in a state of concentration, thereby training the user's concentration.

[0052] Based on the above, the low-cost brain-computer interface-based children's concentration training system disclosed in the embodiment has at least the following technical advantages: The system device is small in size, low in cost and portable; the deployment control is flexible, and the system still maintains stable recognition of the low-quality electroencephalogram signals generated by the low-cost brain-computer interface.

[0053] The end-to-end delay of the system is low, meeting the needs of multi-user training and multi-robot collaboration. And when the number of users and robots increases, the time consumption of the system only increases linearly, supporting large-scale deployment.

[0054] The system uses semantic feature transmission, has strong noise resistance, and overcomes the shortcomings of traditional brain-computer interface systems that are prone to noise interference in raw signal transmission.

[0055] The system realizes that multiple users control multiple robots to complete a collaborative task at the same time, solves the bottleneck of the traditional EEG system that only supports "single brain-single device", and can realize the concentration training of multiple users.

[0056] The system has high concentration training capability, can collect EEG of the user in real time, automatically completes the transmission, collection and processing of the EEG signal, is simple to operate, and is suitable for children.

[0057] The low-cost brain-computer interface-based concentration training system for children disclosed in the embodiment is summarized as follows. Although the above-mentioned multi-modal neural fusion-based concentration training system for children can train the concentration of the user, the scheme needs to use expensive equipment, the system cost is high, cannot be popularized for civilian use, and has high end-to-end delay; the system only divides the concentration into "high / low" two categories, limits the generation of personalized training schemes, the feedback mechanism is dull, the existing system mainly depends on simple scoring, lacks game design, and the long-term use willingness of children is low.

[0058] The embodiment designs a set of popular civilian concentration system for children based on the low-cost mental force headband. The system has the following advantages: (1) the system platform can realize low end-to-end delay; (2) the system takes Brainlink as the EEG collection device of the system, has low cost and portability, has high practical value and market prospect, and has wide application range; (3) the system solves the shortcoming that the traditional concentration training system can only train one user, realizes that multiple users control multiple robots to complete a collaborative task at the same time, and realizes the concentration training of multiple users; (4) the system is simple to operate, can automatically complete the EEG signal collection and processing process, has high concentration training efficiency, and has high universality.

[0059] It is proved that the low-cost brain-computer interface-based concentration training system for children is feasible through experiments. The experiments show that different concentration and relaxation states will emit different categories of EEG signals, and the robot will perform different actions. By establishing the mapping relationship between the EEG signal and the robot category and action, the multiple action controls of different robots are realized, so as to train the concentration of the user. The whole system has low equipment cost and high accuracy, and can realize real-time training of concentration. It should be noted that the embodiment can also be used in medical rehabilitation, game entertainment, intelligent factory and other application scenarios.

[0060] Reference Figure 4 , Figure 4 The flowchart of the low-cost brain-computer interface-based concentration training method for children provided in the embodiment is shown in the figure.

[0061] As Figure 4As shown, the embodiment also discloses a low-cost brain-computer interface-based child concentration training method, applied to the low-cost brain-computer interface-based child concentration training system as described above, and the method comprises: S10: Collect and input the original electroencephalogram signal, process the original electroencephalogram signal based on the preset signal and processing algorithm, and transmit the processing result to the feature transmission subsystem; wherein a plurality of BrainLink mental force headbands are used to collect the original electroencephalogram signal, and the trainee wears the BrainLink mental force headband; S20: The transmitter module and the wireless channel module are provided, the transmitter module is used for feature extraction and coding of the processing result, the wireless channel module is used for transmitting the coding result in the additive white Gaussian noise channel environment, and the end-to-end transmission performance is optimized through the joint cross-entropy and mutual information loss function; S30: The coding result is converted into a command through the preset channel decoder and semantic decoder, the command is used to specify the robot and its action, and the command is only executed when the trainee is in the preset attention state, so as to perform the concentration training.

[0062] The embodiment also discloses a low-cost brain-computer interface-based child concentration training computer device, comprising at least one processor, at least one memory and a data bus; The processor and the memory complete mutual communication through the data bus; The memory stores program instructions executable by the processor, and the processor calls the program instructions to execute the low-cost brain-computer interface-based child concentration training method as described above.

[0063] The embodiment also discloses a storage medium having a computer program stored thereon, and the computer program is executed by the processor to implement the low-cost brain-computer interface-based child concentration training method as described above.

[0064] It should be noted that the low-cost brain-computer interface-based child concentration training method, computer device and storage medium of the embodiment correspond to the low-cost brain-computer interface-based child concentration training system described above. Therefore, the contents not specifically described in the low-cost brain-computer interface-based child concentration training method, computer device and storage medium of the embodiment can be but are not limited to functional definitions, working principles and technical effects, and all can be referred to the description in the low-cost brain-computer interface-based child concentration training system, which will not be described herein.

[0065] In the embodiments provided by the present application, it should be understood that the embodiments described herein can be realized by hardware, software, firmware, middleware, codes or any proper combination thereof. For hardware implementation, the processor can be realized in one or more of the following components: an application specific integrated circuit (ASIC), a digital signal processor (DSP), a digital signal processing device (DSPD), a programmable logic device (PLD), a field programmable gate array (FPGA), a processor, a controller, a microcontroller, a microprocessor, other electronic units designed to perform the functions described herein, or a combination thereof. For software implementation, the procedures described herein can be implemented with a computer program that directs relevant hardware to complete the procedures. When implemented, the above program can be stored in a computer readable storage medium or transmitted as one or more instructions or codes on a computer readable storage medium. The computer readable storage medium includes computer storage media and communication media including any medium that facilitates the transfer of a computer program from one place to another. The storage medium can be any available medium that can be accessed by a computer. The computer readable storage medium can include but not limited to RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage medium or other magnetic storage devices, or any other medium that can be used to carry or store desired program codes in the form of instructions or data structures and can be accessed by a computer.

[0066] Finally, it should be noted that the above only describes the preferred embodiments of the present application and is not intended to limit the present application. Although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that the technical solutions recorded in the foregoing embodiments can be modified or some technical features can be replaced by equivalent ones, and any modification, equivalent replacement, improvement, etc. within the spirit and principle of the present application should be included in the protection scope of the present application.

Claims

1. A low-cost brain-computer interface based children's concentration training system, characterized in that, The system comprises an electroencephalogram subsystem, a feature transmission subsystem and a concentration training subsystem connected in sequence; The electroencephalogram subsystem is configured to collect and input raw electroencephalogram signals, process the raw electroencephalogram signals based on preset signal and processing algorithms, and transmit the processing results to the feature transmission subsystem; wherein a plurality of BrainLink mental force headbands are used for collecting the raw electroencephalogram signals, and the trainees wear BrainLink mental force headbands; The feature transmission subsystem is configured to be provided with a transmitter module and a wireless channel module, the transmitter module is used for feature extraction and coding of the processing results, the wireless channel module is used for placing the coding results in an additive white Gaussian noise channel environment for transmission, and the end-to-end transmission performance is optimized through joint cross-entropy and mutual information loss functions; The concentration training subsystem is configured to convert the coding results into commands through a preset channel decoder and a semantic decoder, specify robots and their actions based on the commands, and execute the commands only when the trainees are in a preset attention state, so as to perform concentration training.

2. The low-cost brain-computer interface based concentration training system for children according to claim 1, wherein, The processing of the raw electroencephalogram signals based on the preset signal and processing algorithms comprises: Down-sampling processing of the raw electroencephalogram signals; Introducing Gaussian noise to enhance the signal; at the same time, simulating data interference caused by noise or sensor disconnection.

3. The low-cost brain-computer interface based concentration training system for children of claim 1, wherein, When the transmitter module extracts and encodes the processing results, the electroencephalogram data is collected over time, including: Using a dynamic electroencephalogram feature encoder for feature extraction, and the dynamic electroencephalogram feature encoder also serves as a semantic encoder; for the electroencephalogram signals collected by BrainLink, the encoder selects a processing path optimized for low-dimensional high-noise data, and introduces an intermittent mask mechanism to simulate incomplete signals by adding random masks in the signals; At the same time, a sparse attention mechanism is used to capture long-range dependencies in long-time sequence inputs; and a model compression technique is used to reduce storage occupancy and computational overhead.

4. The low-cost brain-computer interface based concentration training system for children of claim 1, wherein, The joint cross-entropy and mutual information loss function The mathematical expression is: The optimization of the end-to-end transmission performance based on the joint cross-entropy and mutual information loss functions comprises: The first stage maximizes mutual information to improve channel capacity and data rate; the second stage utilizes cross-entropy to constrain consistency of channel encoding and decoding; the third stage again uses cross-entropy to ensure faithful recovery of semantic information in the encoding and decoding process; wherein, and are preset weight parameters. 5.The low-cost brain-computer interface based children concentration training system according to claim 1, wherein, The attention state includes a non-transition state and a transition state; The non-transition state includes sustained concentration and sustained relaxation; The transition state includes concentration-relaxation and relaxation-concentration.

6. The low-cost brain-computer interface based concentration training system for children according to claim 5, wherein, The command is executed only when the trainee is in a preset attention state, specifically: each attention state corresponds to different actions of different robots.

7. The low-cost brain-computer interface based concentration training system for children according to claim 6, wherein, When the attention state is in the transition state, a preset intervention strategy is cut in based on the command, and the intervention strategy is used for early intervention between the transition state and the non-transition state.

8. A low-cost brain-computer interface based children's concentration training method applied to the low-cost brain-computer interface based children's concentration training system of any one of claims 1 to 7, characterized in that, The method comprises: Collecting and inputting raw electroencephalogram signals, processing the raw electroencephalogram signals based on preset signal and processing algorithms, and transmitting the processing results to the feature transmission subsystem; wherein a plurality of BrainLink mental force headbands are used for collecting the raw electroencephalogram signals, and the trainees wear BrainLink mental force headbands; A transmitter module is arranged to perform feature extraction and encoding on the processing result, and a wireless channel module is arranged to place the encoded result in an additive white Gaussian noise channel environment for transmission, and to optimize end-to-end transmission performance through joint cross-entropy and mutual information loss functions. The encoded result is converted into a command through a preset channel decoder and a semantic decoder, the command is used to specify the robot and its actions, and the command is executed only when the trainee is in a preset attention state, so as to perform attention training.

9. A low-cost brain-computer interface based computer device for children's concentration training, characterized in that, The system comprises at least one processor, at least one memory and a data bus; The processor and the memory communicate with each other through the data bus; The memory stores program instructions executable by the processor, and the processor invokes the program instructions to execute the low-cost brain-computer interface-based attention training method for children according to claim 8.

10. A storage medium having stored thereon a computer program, characterized in that The computer program is executed by the processor to implement the low-cost brain-computer interface-based attention training method for children according to claim 8.

Images