A single-channel brain wave training method and system based on multi-person cooperation

Abstract

A single-channel brain wave training method and system based on multi-person cooperation, which comprises obtaining single-channel brain wave data of multiple participants; for each participant, performing a meditation index calculation step: extracting a frequency spectrum feature vector within a preset time window from the single-channel brain wave data corresponding to each participant, and inputting it into a pre-trained deep learning model to obtain a meditation evaluation value of the corresponding participant; calculating the time domain stability index of the single-channel brain wave data within the preset time window, and calculating the meditation index of the corresponding participant; continuously monitoring the meditation index of each participant, and determining the state category currently occupied by each participant according to the meditation index of each participant and its historical data; based on the meditation indexes of all non-eliminated participants, determining a group meditation index score or a group state level; according to the group meditation index score, generating a brightness control instruction of a lamp; or, according to the group state level, generating a color value control instruction of the lamp.

Description

Technical Field

[0001] This invention relates to the field of multi-person collaborative EEG training technology, and in particular to a single-channel EEG training method and system based on multi-person collaboration. Background Technology

[0002] Single-channel EEG acquisition technology has been applied to individual relaxation and concentration training. By analyzing the power of frequency bands such as alpha waves and gamma waves, the cognitive state of users can be quantitatively assessed.

[0003] However, existing EEG neurofeedback training systems are all designed for single-person, independent use, with feedback mechanisms adjusting only to the individual's state. On the one hand, the lack of multi-person collaborative training mechanisms makes it difficult for users to persist in the long term due to a lack of social interaction and group motivation; on the other hand, the inability to integrate and analyze the EEG states of multiple individuals makes it difficult to generate a dynamic feedback signal that reflects the overall performance of the group, resulting in a lack of a sense of collective purpose and team collaboration experience during the training process.

[0004] In summary, existing technologies lack a brainwave training method that can support multi-person collaboration and provide dynamic feedback. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention provides a single-channel EEG training method based on multi-person collaboration, comprising the following steps: Acquire single-channel EEG data from multiple participants; For each participant, the following steps are performed to calculate the meditation index: A spectral feature vector within a preset time window is extracted from the single-channel EEG data corresponding to each participant and input into a pre-trained deep learning model to obtain the meditation assessment value for the corresponding participant; the temporal stability index of the single-channel EEG data within the preset time window is calculated, and based on the single-channel EEG data, the temporal stability index, and the meditation assessment value, the meditation index for the corresponding participant is calculated; wherein, the spectral feature vector includes at least the relative power of Alpha waves, the relative power of Gamma waves, and the Alpha / Beta power ratio; Continuously monitor each participant's meditation index, and determine each participant's current state category based on each participant's meditation index and its historical data; Based on the meditation index of all non-eliminated participants, determine the group meditation index score or group state level; Based on the group meditation index score, a brightness control command for the luminaire is generated; or, a color value control command for the luminaire is generated based on the group state level.

[0006] Optionally, after determining the current state category of each participant, the following may also be included: Based on the state category, a corresponding voice prompt instruction is generated; wherein, when the Zen index of any participant is continuously less than the elimination threshold within a first preset time period, an elimination warning voice instruction is triggered and sent to the corresponding participant; if the Zen index of the participant is still less than the elimination threshold after at least two elimination warning voice instruction prompts, the participant is eliminated.

[0007] Optionally, a spectral feature vector within a preset time window is extracted from the single-channel EEG data corresponding to each participant and input into a pre-trained deep learning model to obtain the meditation assessment value of the corresponding participant, including: Extract the spectral feature vector within a preset time window from the single-channel EEG data corresponding to each participant; The spectral feature vector is input into a pre-trained deep learning model; the deep learning model extracts the local time-frequency features of the spectral feature vector through three sequentially connected one-dimensional convolutional layers, and inputs them into two long short-term memory network layers to capture temporal dependencies, and then maps the spectral feature vector into a Zen evaluation value through a fully connected layer and outputs it.

[0008] Optionally, the temporal stability index of the single-channel EEG data within a preset time window is calculated, and based on the single-channel EEG data, the temporal stability index, and the meditation assessment value, the meditation index of the corresponding participant is calculated, including: Calculate the coefficient of variation of the single-channel EEG data within a preset time window, and use the reciprocal of the coefficient of variation as the temporal stability index of the single-channel EEG data; The meditation index of the corresponding participant is calculated based on the following preset formula: ;in, Let be the meditation index of the j-th participant; Let be the meditation assessment value of the j-th participant; The relative power of the Alpha wave, The weighting coefficient for the relative power of the Alpha wave; The relative power of the Gamma wave, The weighting coefficient for the relative power of the Gamma wave; The Alpha / Beta power ratio is... The weighting coefficient for the Alpha / Beta power ratio; The time-domain stability index is... The weighting coefficients for the time-domain stability index; The function represents a normalized mapping based on the individual historical baseline data of the j-th participant.

[0009] Optionally, based on the meditation index of all non-eliminated participants, a group meditation index score is determined, including: A weight is calculated for the Zen index of each non-eliminated participant. The weight is calculated based on the non-eliminated participant quantity balance factor, the individual historical performance factor, and the current volatility suppression factor. The non-eliminated participant quantity balance factor is the reciprocal of the current number of non-eliminated participants. The individual historical performance factor is the ratio of the rolling mean of the Zen index of the non-eliminated participant to the rolling mean of the Zen index of all non-eliminated participants in the first historical period. The current volatility suppression factor is determined based on the standard deviation of the Zen index of each non-eliminated participant in the second historical period. The group meditation index score is calculated by summing the products of the meditation index and its corresponding weight for all non-eliminated participants using a dynamic weighted average algorithm.

[0010] Optionally, based on the group meditation index score, a brightness control command for the luminaire is generated, including: The group meditation index score is compared with a first preset brightness threshold and a second preset brightness threshold to determine the corresponding target brightness level; wherein, when the group meditation index score is less than the first preset brightness threshold, the target brightness level is a dim brightness level; when the group meditation index score is between the first preset brightness threshold and the second preset brightness threshold, the target brightness level corresponds to a normal brightness level; when the group meditation index score is greater than the second preset brightness threshold, the target brightness level corresponds to a bright brightness level. Based on the group meditation index score and the target brightness level, a brightness control command for the luminaire is generated.

[0011] Optionally, based on the meditation index of all non-eliminated participants, the group state level is determined, including: The individual state level of each non-eliminated participant is determined based on their meditation index; wherein the individual state level includes at least a first state level, a second state level, a third state level, and a fourth state level, and each state level corresponds to a different degree of individual cognitive state. The group status level is determined based on the individual status level with the highest proportion among all non-eliminated participants.

[0012] Optionally, the color value control instruction for the lamp is generated according to the group status level. Specifically, the target color value corresponding to the group status level is obtained based on a preset color value mapping table, and the color value control instruction for the lamp is generated based on the target color value.

[0013] Optionally, the method further includes: Obtain the preset training target threshold; The group meditation index score is compared with a preset training target threshold to determine whether the training target has been achieved.

[0014] Corresponding to the aforementioned multi-person collaborative single-channel EEG training method, the present invention provides a multi-person collaborative single-channel EEG training system, comprising: The EEG data acquisition module is used to acquire single-channel EEG data from multiple participants. The meditation index calculation module is used to perform the meditation index calculation steps for each participant: extracting the spectral feature vector within a preset time window from the single-channel EEG data corresponding to each participant, and inputting it into a pre-trained deep learning model to obtain the meditation assessment value of the corresponding participant; calculating the temporal stability index of the single-channel EEG data within the preset time window, and calculating the meditation index of the corresponding participant based on the single-channel EEG data, the temporal stability index, and the meditation assessment value; wherein, the spectral feature vector includes at least the relative power of Alpha waves, the relative power of Gamma waves, and the Alpha / Beta power ratio; The monitoring module is used to continuously monitor the meditation index of each participant and determine the current state category of each participant based on the meditation index and its historical data. The group status assessment module is used to determine the group meditation index score or group status level based on the meditation index of all non-eliminated participants. The lighting control instruction generation module is used to generate brightness control instructions for the lighting fixtures based on the group meditation index score; or, to generate color value control instructions for the lighting fixtures based on the group state level.

[0015] Compared with the prior art, the present invention has the following beneficial effects: (1) By acquiring single-channel EEG data of multiple participants and extracting spectral feature vectors within a preset time window, the synchronous acquisition and feature processing of multiple participants' EEG states were realized; by calculating the meditation index of the corresponding participants, the quantitative assessment of individual cognitive states was realized; by determining the group meditation index score or group state level based on the meditation index of all non-eliminated participants, and generating brightness control instructions or color value control instructions for lamps accordingly, the individual EEG states of multiple participants were integrated into a dynamic feedback signal that can reflect the overall performance of the group, and the changes in light brightness or color created a visual atmosphere for group collaborative training.

[0016] (2) Before eliminating the corresponding participant, trigger the elimination warning voice command first, and only execute the elimination when the command is issued twice and the Zen index is still less than the elimination threshold. This allows the participant to receive clear improvement prompts and buffer opportunities before being eliminated, which maintains the seriousness of the group training rules and avoids the negative experience caused by the sudden elimination of the corresponding participant.

[0017] (3) The local features of the spectral feature vector in the time-frequency domain are extracted through three one-dimensional convolutional layers, the temporal dependencies in the data are captured through two long short-term memory network layers, and the features are mapped to the Zen index and output through a fully connected layer. The synergistic effect of the above network structure enables the deep learning model to simultaneously utilize the local feature extraction capability of the convolutional neural network and the temporal modeling capability of the long short-term memory network, thereby improving the mapping accuracy and model robustness from single-channel EEG spectral features to individual cognitive state indices.

[0018] (4) The temporal fluctuation characteristics of the EEG signal are quantified by the inverse of the coefficient of variation of single-channel EEG data within a preset time window, and the temporal stability index is introduced as a component of the meditation index. The Normalize function is combined with the normalization mapping based on the individual's historical baseline data to eliminate the scoring bias caused by the difference in the physiological baseline of different participants, so that the meditation index across individuals is comparable. Finally, the meditation assessment value and the normalized weighted fusion value are verified and fused, taking into account the feature extraction capability of deep learning and the interpretability of physiological indicators, thereby improving the robustness and individual adaptability of the meditation state assessment.

[0019] (5) By introducing a factor to balance the number of participants who have not been eliminated, the bias in the group score caused by changes in the current number of participants is eliminated; by introducing a factor to reflect the long-term stable performance of participants rather than a single point of state, the weights are made to reflect the long-term stable performance of participants rather than a single point of state; by introducing a factor to suppress the current fluctuation amplitude, the interference of instantaneous abnormal fluctuations on the group score is reduced. The synergistic effect of the above three factors enables the group Zen Index score to maintain the fairness, stability and representativeness of the calculation results under the multiple influences of dynamic changes in the number of participants, differences in individual performance and random fluctuations in data.

[0020] (6) By comparing the group meditation index score with the first preset brightness threshold and the second preset brightness threshold and dividing it into three brightness levels, the abstract group cognitive state is mapped into discrete and intuitive changes in light brightness, which makes it easier for participants to quickly identify the overall state of the group through visual perception and enhances the immediacy and comprehensibility of the group feedback signal.

[0021] (7) By mapping the meditation index of non-eliminated participants to individual state levels (at least four levels), the continuous numerical scores are discretized into interpretable qualitative states, reducing the cognitive load of state interpretation; the group state level is determined based on the individual state level with the highest proportion, making the state consensus in group collaborative training visible.

[0022] (8) By using a preset color value mapping table, the group state level is converted into the target color value of the lamp. The change of light environment color is used as a real-time visual feedback carrier of the group's meditation state. The abstract EEG assessment results are transformed into intuitive sensory stimulation, reducing the threshold for participants to recognize the state.

[0023] (9) By obtaining the preset training target threshold and comparing it with the group meditation index score, the group training has a clear target endpoint and quantitative evaluation standard, which makes it easier to objectively judge whether the training has achieved the expected effect and provides target guidance for the group collaborative training process. Attached Figure Description

[0024] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this invention, illustrate exemplary embodiments of the invention and are used to explain the invention, but do not constitute an undue limitation of the invention. In the drawings: Figure 1 This is a simplified flowchart of an embodiment of the single-channel EEG training method based on multi-person collaboration of the present invention; Figure 2 This is a schematic diagram of light brightness levels in an embodiment of the single-channel EEG training method based on multi-person collaboration of the present invention. Figure 3 The present invention is a framework diagram of an embodiment of a multi-person collaborative single-channel EEG training system. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] like Figure 1 As shown, the present invention provides a single-channel EEG training method based on multi-person collaboration, which includes the following steps: Acquire single-channel EEG data from multiple participants; For each participant, the following steps are performed to calculate the meditation index: A spectral feature vector within a preset time window is extracted from the single-channel EEG data corresponding to each participant and input into a pre-trained deep learning model to obtain the meditation assessment value for the corresponding participant; the temporal stability index of the single-channel EEG data within the preset time window is calculated, and the meditation index for the corresponding participant is calculated based on the single-channel EEG data, the temporal stability index, and the meditation assessment value; wherein, the spectral feature vector includes at least the relative power of the Alpha wave, the relative power of the Gamma wave, and the Alpha / Beta power ratio; Continuously monitor each participant's meditation index, and determine each participant's current state category based on each participant's meditation index and its historical data; Based on the meditation index of all non-eliminated participants, determine the group meditation index score or group state level; The brightness control instructions for the luminaires are generated based on the group meditation index score; or, the color value control instructions for the luminaires are generated based on the group state level.

[0027] In this embodiment, before acquiring single-channel EEG data from multiple participants, a first communication link is established with each participant's EEG acquisition device (e.g., a worn EEG data acquisition module) via Bluetooth. The EEG acquisition device and the lamp have different Bluetooth connection priorities; the connection priority of the EEG acquisition device is higher than that of the lamp. After establishing the first communication link for each participant, a second communication link is established with the lamp via Bluetooth. During the data acquisition process, a heartbeat signal is sent to the connected device at a preset time interval (e.g., 5 seconds). If no heartbeat response is received for a preset number of consecutive times (e.g., 3 times), the device is determined to be disconnected, and a reconnection is automatically initiated.

[0028] Preferably, before acquiring single-channel EEG data from multiple participants, one minute of resting-state single-channel EEG data from all participants is collected as individual historical baseline data for calibrating the deep learning model.

[0029] This invention achieves synchronous acquisition and characterization of multiple participants' brainwave states by acquiring single-channel EEG data from multiple participants and extracting spectral feature vectors within a preset time window; it also achieves quantitative assessment of individual cognitive states by calculating the meditation index of corresponding participants; and it determines the group meditation index score or group state level based on the meditation indices of all non-eliminated participants, and generates brightness control instructions or color value control instructions for lighting fixtures accordingly. This integrates the individual EEG states of multiple participants into a dynamic feedback signal that reflects the overall performance of the group, and creates a visual atmosphere for collaborative group training by changing the brightness or color of the lights.

[0030] Furthermore, after determining the current state category of each participant, the process also includes: generating corresponding voice prompts based on the current state category of each participant; wherein, when the Zen index of any participant is continuously less than the elimination threshold for a first preset duration (preferably 3 minutes), an elimination warning voice command is triggered and sent to the corresponding participant; if the Zen index of the participant is still less than the elimination threshold after at least two elimination warning voice command prompts, then the participant is eliminated.

[0031] Before eliminating a participant, this invention triggers an elimination warning voice command, and only executes the elimination if the command is issued twice and the Zen index is still less than the elimination threshold. This allows participants to receive clear improvement prompts and buffer opportunities before being eliminated, thus maintaining the seriousness of the group training rules and avoiding the negative experience caused by the sudden elimination of the participant.

[0032] In this embodiment, the elimination threshold is set according to a first preset percentage (preferably 70%) of the average group meditation index.

[0033] This invention sets the elimination threshold as a first preset percentage of the average group Zen index, enabling the elimination standard to adaptively adjust with fluctuations in the overall group performance. This avoids the unfairness caused by a fixed threshold in different training stages or with different group compositions, and ensures that the elimination mechanism dynamically matches the real-time state of the group.

[0034] Preferably, based on each participant's meditation index and its historical data, the current state category of each participant is determined, and corresponding voice prompts are generated, including: Based on each participant's meditation index and its historical data, when any participant's meditation index is greater than or equal to the first threshold (preferably 80 points) within a second preset duration (preferably 30 seconds), the participant's current state category is determined to be the excellent performance category, and a first voice prompt instruction is generated. When the average value of any participant's meditation index within a second preset time period increases compared to the average value within an earlier second preset time period, and the increase is greater than or equal to the second threshold (preferably 15 points), the participant's current state category is determined to be the "significant progress" category, and a second voice prompt instruction is generated. When any participant's meditation index is less than the second preset percentage (preferably 85%) of the current group's average meditation index within a second preset time period, the current state category of the non-eliminated participant is determined to be the poor performance category, and a third voice prompt instruction is generated. When the meditation index of any participant is less than the third preset percentage (preferably 70%) of the average meditation index of the current group within the third preset time period (preferably 60s), the current state category of the non-eliminated participant is determined to be the influencing group category, and a fourth voice prompt instruction is generated.

[0035] This invention combines the meditation index of each non-eliminated participant with its historical data to classify them into four status categories: excellent performance, significant improvement, poor performance, and influencing the group. Correspondingly, it generates differentiated voice prompts, achieving refined identification and classification feedback of individual status. This allows participants in different statuses to receive guidance information that matches their current performance and trends, improving the relevance of individual feedback.

[0036] For example, the first voice prompt includes words of praise (e.g., "So-and-so is great, in excellent condition!"); the second voice prompt includes words of encouragement (e.g., "So-and-so has made progress, keep it up!"); the third voice prompt includes encouraging prompts (e.g., "So-and-so, keep it up, relax and breathe!"); and the fourth voice prompt includes gentle reminders (e.g., "So-and-so, please note, everyone needs you!").

[0037] This invention maps praise, encouragement, motivational prompts, and gentle reminders to four state categories, enabling the voice feedback to accurately match the participant's state category at the semantic level. This enhances the emotional adaptability of the feedback and helps participants in different states generate differentiated psychological motivation.

[0038] In this embodiment, a spectral feature vector within a preset time window is extracted from the single-channel EEG data corresponding to each participant and input into a pre-trained deep learning model to obtain the meditation assessment value of the corresponding participant, including: Extract the spectral feature vector within a preset time window (preferably 10 consecutive seconds) from the single-channel EEG data corresponding to each participant; The spectral feature vector is input into a pre-trained deep learning model. The deep learning model extracts the local time-frequency features of the spectral feature vector through three sequentially connected one-dimensional convolutional layers (with kernel sizes of 3, 3, and 5, filter numbers of 32, 64, and 128, activation function of ReLU, followed by a batch normalization layer and a max pooling layer after each layer). These features are then input into two long short-term memory network layers (with 128 and 64 hidden units, and a dropout rate of 0.3) to capture temporal dependencies. Finally, a fully connected layer maps the spectral feature vector to Zentrum evaluation values ​​(ranging from 0 to 100) and outputs them.

[0039] This invention extracts local time-frequency features from the spectral feature vector using three one-dimensional convolutional layers, captures temporal dependencies in the data using two long short-term memory (LSM) network layers, and then maps the features to a meditation index using a fully connected layer for output. The synergistic effect of these network structures allows the deep learning model to simultaneously utilize the local feature extraction capabilities of convolutional neural networks and the temporal modeling capabilities of LSM networks, improving the mapping accuracy and robustness of the model from single-channel EEG spectral features to an individual's cognitive state index.

[0040] In this embodiment, the training dataset for the deep learning model uses single-channel EEG data from no fewer than 500 subjects, each for at least 30 minutes, covering various cognitive states such as resting state, meditative state, focused state, and distracted state. Training labels are manually assigned by three or more experienced meditation instructors based on the subjects' behavioral performance and subjective reports, with labels ranging from 0 to 100 points representing a meditation assessment score.

[0041] The training steps and parameters for a deep learning model are as follows: Step 1: Data preprocessing—bandpass filtering (0.5-50Hz) and independent component analysis (ICA) are performed on single-channel EEG data to remove artifacts from electrooculography and electromyography, and sliding window slicing is performed with a 10-second window and a 5-second step size. Step 2: Feature Extraction – Calculate a 12-dimensional spectral feature vector for each window; the 12-dimensional spectral feature vector includes, in sequence: (1) Delta band power, used to reflect slow waves and deep relaxation activities; (2) Theta band power, used to reflect relaxation, inner focus and meditation-related activities; (3) Low-Alpha band power, used to reflect initial relaxation activities in a conscious state; (4) High-Alpha band power, used to reflect relatively stable calm and relaxation activities; (5) Low-Beta band power, used to reflect general focus and thinking activities; (6) High-Beta band power, used to reflect higher alertness or tension activities; (7) Low-Gamma band power, used to reflect higher-order information processing activities; (8) Mid-Gamma band power, used to reflect strong high-frequency information integration activities; (9) Alpha wave relative power, which is the power of the Low-Alpha and High-Alpha bands. The sum of power is the proportion of the total power of the above eight frequency bands, used to characterize the overall relaxation level; (10) Gamma wave relative power is the proportion of the sum of power of the Low-Gamma and Mid-Gamma frequency bands to the total power of the above eight frequency bands, used to characterize high-frequency cognitive integration activities; (11) Alpha / Beta power ratio is the ratio of the total Alpha power to the total Beta power, where the total Alpha power is the sum of the power of the Low-Alpha and High-Alpha frequency bands, and the total Beta power is the sum of the power of the Low-Beta and High-Beta frequency bands, used to characterize the relative relationship between relaxation activities and alert activities; (12) Theta wave relative power is the proportion of the power of the Theta frequency band to the total power of the above eight frequency bands, used to characterize inner focus and meditation tendency; It should be noted that the above 12-dimensional spectral feature vectors are used as input for training the deep learning model, and a single dimension is not directly equivalent to the meditative state; Step 3: Data partitioning – Divide the training set, validation set, and test set into a 7:1.5:1.5 ratio, and use stratified sampling to ensure a balanced distribution of each cognitive state category; Step 4: Model Training – Using the Adam optimizer, the initial learning rate is 0.001, the learning rate decay strategy is to decrease to 0.5 times the original rate every 20 epochs, the batch size is 64, the maximum training epochs are 200 epochs, an early stopping strategy (patience=15) is adopted, and the loss function is mean squared error (MSE). Step 5: Model Validation and Deployment – ​​Evaluate the model performance on the test set, requiring a root mean square error (RMSE) ≤ 8 points and a Pearson correlation coefficient ≥ 0.85. Once the requirements are met, deploy the model to the AI ​​server for real-time inference.

[0042] The temporal stability index of single-channel EEG data within a preset time window was calculated. Based on the single-channel EEG data, the temporal stability index, and the meditation assessment value, the meditation index of the corresponding participant was calculated, including: Calculate the coefficient of variation of single-channel EEG data within a preset time window, and use the reciprocal of the coefficient of variation as the temporal stability index of single-channel EEG data; The meditation index of the corresponding participant is calculated based on the following preset formula: ;in, Let be the meditation index of the j-th participant; Let be the meditation assessment value of the j-th participant; The relative power of the Alpha wave, The weighting coefficient for the relative power of the Alpha wave; The relative power of the Gamma wave, The weighting coefficient for the relative power of the Gamma wave; The Alpha / Beta power ratio is... The weighting coefficient for the Alpha / Beta power ratio; The time-domain stability index is... The weighting coefficients for the time-domain stability index; The function represents a normalized mapping based on the individual historical baseline data of the j-th participant.

[0043] This invention quantifies the temporal fluctuation characteristics of EEG signals by using the inverse of the coefficient of variation of single-channel EEG data within a preset time window, and introduces a temporal stability index as a component of the meditation index. It combines this with a Normalize function to perform normalization mapping based on individual historical baseline data, eliminating scoring biases caused by differences in physiological baselines among different participants, thus making cross-individual meditation indices comparable. Finally, through verification and fusion of meditation assessment values ​​and normalized weighted fusion values, it balances the feature extraction capabilities of deep learning with the interpretability of physiological indicators, improving the robustness and individual adaptability of meditation state assessment.

[0044] In this embodiment, the group meditation index score is determined based on the meditation index of all non-eliminated participants, including: The Zen index of each non-eliminated participant is calculated with a corresponding weight, which is calculated based on the balance factor of the number of non-eliminated participants, the individual historical performance factor, and the current volatility suppression factor. The equilibrium factor for the number of participants not yet eliminated is the reciprocal of the current number of participants not yet eliminated, F1 = F1 represents the balance factor of the number of participants not eliminated, and N represents the current number of participants not eliminated. The individual historical performance factor is the ratio of the rolling mean of the Zen index of the non-eliminated participant to the rolling mean of the Zen index of all non-eliminated participants within the first historical period, F2= F2 represents the individual's historical performance factor. This represents the rolling average of the Zen index of the i-th non-eliminated participant within the first historical time period (preferably 300s). It represents the rolling average of the Zen index of all non-eliminated participants within the first historical period (preferably 300s); The current volatility suppression factor is determined based on the standard deviation of the Zen index of each surviving participant during the second historical period, F3 = F3 represents the current volatility suppression factor. It represents the standard deviation of the meditation index of the i-th non-eliminated participant within the second historical period (preferably 30s); The group meditation index score is calculated by summing the products of the meditation index of all non-eliminated participants and their corresponding weights using a dynamic weighted average algorithm. The specific calculation formula is as follows: , This represents the group meditation index score. This represents the Zen index of the i-th non-eliminated participant (the specific calculation formula can be found above). The calculation formula uses i and j only to distinguish between non-eliminated participants and all participants. This represents the calculated weight of the i-th non-eliminated participant. .

[0045] This invention eliminates the bias in group scores caused by changes in the current number of participants by introducing a balancing factor for the number of non-eliminated participants; it ensures that the weights reflect the long-term stable performance of participants rather than a single point in time by introducing an individual historical performance factor; and it reduces the interference of instantaneous abnormal fluctuations on the group score by introducing a current fluctuation amplitude suppression factor. The synergistic effect of these three factors ensures that the group Zen Index score maintains the fairness, stability, and representativeness of the calculation results even under the multiple influences of dynamic changes in the number of participants, differences in individual performance, and random data fluctuations.

[0046] In this embodiment, a brightness control command for the lamp is generated based on the group meditation index score, including: The group meditation index score is compared with the first preset brightness threshold (preferably 40) and the second preset brightness threshold (preferably 60) to determine the corresponding target brightness level; Please refer to Figure 2When the group meditation index score is less than the first preset brightness threshold, the target brightness level is a dim brightness level (visual effect: a faint light, almost off); when the group meditation index score is between the first preset brightness threshold and the second preset brightness threshold, the target brightness level corresponds to a normal brightness level (visual effect: normal light, soft but not bright); when the group meditation index score is greater than the second preset brightness threshold, the target brightness level corresponds to a bright brightness level (visual effect: radiant, bright and warm). Based on the group meditation index score and the target brightness level, brightness control instructions for the luminaires are generated.

[0047] Preferably, the target brightness value for the dim brightness level is in the range of [0, 40); the target brightness value for the general brightness level is in the range of [40, 60]; and the target brightness value for the bright brightness level is in the range of (60, 100). The specific target brightness value can be equivalent to the group meditation index score.

[0048] This invention compares the group meditation index score with a first preset brightness threshold and a second preset brightness threshold and divides it into three brightness levels, so that the abstract group cognitive state is mapped into discrete and intuitive changes in light brightness. This makes it easier for participants to quickly identify the overall state of the group through visual perception and enhances the immediacy and comprehensibility of the group feedback signal.

[0049] In this embodiment, the group state level is determined based on the meditation index of all non-eliminated participants, including: Each participant's individual state level is determined based on their meditation index, which is the level of concentration. Each state level includes at least four levels: Level 1, Level 2, Level 3, and Level 4, each corresponding to a different degree of cognitive state. For example, Level 1 is a scattered state (distracted thinking, numerous distracting thoughts, inability to relax); Level 2 is a generally alert state (occasionally relaxed, but frequently interrupted by distracting thoughts); Level 3 is a slightly relaxed state (relatively relaxed body and mind, significantly reduced distracting thoughts); and Level 4 is a deeply relaxed state (deep meditation state, high focus and inner integration). It should be noted that the mapping relationship between the meditation index and the individual state level can be preset according to actual needs. For example, the meditation index ranges are [0,25], [26,50], [51,75], and [76,100], corresponding to Levels 1, 2, 3, and 4, respectively.

[0050] The group status level is determined based on the individual status level with the highest proportion among all non-eliminated participants.

[0051] Preferably, the color value control instruction for the lamp is generated according to the group status level, specifically: based on a preset color value mapping table, the target color value corresponding to the group status level is obtained, and the color value control instruction for the lamp is generated based on the target color value.

[0052] In this embodiment, a preset color value mapping table stores the mapping relationship between each individual's state level and the target color value. For example, if the individual with the highest percentage of states is in a relaxed state, the corresponding target color value is warm orange (255, 100, 50); if the individual with the highest percentage of states is in a generally awake state, the corresponding target color value is cyan (100, 200, 150); if the individual with the highest percentage of states is in a slightly relaxed state, the corresponding target color value is sky blue (50, 150, 255); if the individual with the highest percentage of states is in a deeply relaxed state, the corresponding target color value is soft purple (80, 50, 180).

[0053] It should be noted that the brightness / color value switching of this invention adopts a 0.5s linear gradual change, prohibiting instantaneous jumps, and the PWM drive frequency is constant at 1200-1500Hz to achieve flicker-free and smooth output of RG0.

[0054] In this embodiment, the method further includes: Obtain preset training target thresholds; preferably, the preset training target thresholds can be used to distinguish different training difficulties. The first preset training target threshold is 60 points, which corresponds to the general training difficulty and is suitable for beginners, large-scale team building activities, and children's training. The second preset training target threshold is 70 points, which corresponds to the moderate training difficulty and is suitable for those with a certain meditation foundation and corporate mindfulness courses. The third preset training target threshold is 75 points, which corresponds to the strict training difficulty and is suitable for experienced meditators, professional cognitive training, and competition mode.

[0055] The group meditation index score is compared with the preset training target threshold to determine whether the training target has been achieved; when the group meditation index score is greater than or equal to the preset training target threshold, it is determined that the training target has been achieved.

[0056] This invention obtains a preset training target threshold and compares it with the group meditation index score, so that group training has a clear target endpoint and quantitative evaluation standard, which facilitates objective judgment on whether the training has achieved the expected effect and provides goal guidance for the group collaborative training process.

[0057] Preferably, the method further includes: continuously monitoring the meditation index of non-eliminated participants and ranking them.

[0058] This invention can be applied to the following scenarios: Meditation centers and Zen academies: When practicing in groups, the central lamp serves as a direct indicator of the depth of collective meditation, enhancing the atmosphere of group practice and mutual encouragement.

[0059] Corporate Mindfulness Training: Gamification mechanisms are used in team building activities or corporate mindfulness courses to enhance team cohesion and individual focus.

[0060] School mental health courses: Students engage in group relaxation exercises in the classroom, with lighting and a competitive element adding to the fun.

[0061] Rehabilitation medical institutions: multi-person cognitive training or stress reduction therapy, which improves rehabilitation effects and patient cooperation through group interaction.

[0062] Parent-child interaction training: Parents and children participate together to enhance their understanding and emotional connection.

[0063] Cultural exhibitions and experiential activities: These interactive exhibits attract visitors to participate in brainwave experiences and popularize brain science knowledge.

[0064] like Figure 3 As shown, the present invention also provides a single-channel EEG training system based on multi-person collaboration, which includes: The EEG data acquisition module 10 is used to acquire single-channel EEG data from multiple participants; preferably, it acquires EEG signals at a sampling frequency of 256Hz, outputs one frame of raw data every 100ms, and has a built-in pre-filter to filter out 50 / 60Hz power frequency electromagnetic interference. The meditation index calculation module 20 is used to perform the meditation index calculation steps for each participant: extracting the spectral feature vector within a preset time window from the single-channel EEG data corresponding to each participant, and inputting it into a pre-trained deep learning model to obtain the meditation assessment value of the corresponding participant; calculating the temporal stability index of the single-channel EEG data within the preset time window, and calculating the meditation index of the corresponding participant based on the single-channel EEG data, the temporal stability index, and the meditation assessment value; wherein, the spectral feature vector includes at least the relative power of Alpha waves, the relative power of Gamma waves, and the Alpha / Beta power ratio; The monitoring module 30 is used to continuously monitor the meditation index of each participant and determine the current state category of each participant based on the meditation index and its historical data. Group status assessment module 40 is used to determine the group meditation index score or group status level based on the meditation index of all non-eliminated participants; The lighting control instruction generation module 50 is used to generate brightness control instructions for lighting fixtures based on the group meditation index score; or, to generate color value control instructions for lighting fixtures based on the group status level.

[0065] Furthermore, the system also includes a voice prompt module, used to generate corresponding voice prompt instructions based on the state category.

[0066] It should be noted that the various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For system embodiments, since they are basically similar to method embodiments, the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments.

[0067] Furthermore, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover 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 one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0068] The foregoing description illustrates and describes preferred embodiments of the present invention. It should be understood that the present invention is not limited to the forms disclosed herein and should not be construed as excluding other embodiments. It can be used in various other combinations, modifications, and environments, and can be altered within the scope of the inventive concept by means of the foregoing teachings or techniques or knowledge in related fields. Any modifications and variations made by those skilled in the art that do not depart from the spirit and scope of the present invention should be within the protection scope of the appended claims.

Claims

1. A single-channel EEG training method based on multi-person collaboration, characterized in that, Includes the following steps: Acquire single-channel EEG data from multiple participants; For each participant, the meditation index calculation step is performed: extract the spectral feature vector within a preset time window from the single-channel EEG data corresponding to each participant, and input it into the pre-trained deep learning model to obtain the meditation assessment value of the corresponding participant. Calculate the temporal stability index of the single-channel EEG data within a preset time window, and calculate the meditation index of the corresponding participant based on the single-channel EEG data, the temporal stability index, and the meditation assessment value; wherein, the spectral feature vector includes at least the relative power of Alpha waves, the relative power of Gamma waves, and the Alpha / Beta power ratio; Continuously monitor each participant's meditation index, and determine each participant's current state category based on each participant's meditation index and its historical data; Based on the meditation index of all non-eliminated participants, determine the group meditation index score or group state level; Based on the group meditation index score, a brightness control command for the luminaire is generated; or, a color value control command for the luminaire is generated based on the group state level.

2. The single-channel EEG training method based on multi-person collaboration according to claim 1, characterized in that, After determining the current state category of each participant, the following is also included: Based on the state category, a corresponding voice prompt instruction is generated; wherein, when the Zen index of any participant is continuously less than the elimination threshold within a first preset time period, an elimination warning voice instruction is triggered and sent to the corresponding participant; if the Zen index of the participant is still less than the elimination threshold after at least two elimination warning voice instruction prompts, the participant is eliminated.

3. The single-channel EEG training method based on multi-person collaboration according to claim 1, characterized in that, The spectral feature vector within a preset time window is extracted from the single-channel EEG data corresponding to each participant and input into a pre-trained deep learning model to obtain the meditation assessment value of the corresponding participant, including: Extract the spectral feature vector within a preset time window from the single-channel EEG data corresponding to each participant; The spectral feature vector is input into a pre-trained deep learning model; the deep learning model extracts the local time-frequency features of the spectral feature vector through three sequentially connected one-dimensional convolutional layers, and inputs them into two long short-term memory network layers to capture temporal dependencies, and then maps the spectral feature vector into a Zen evaluation value through a fully connected layer and outputs it.

4. The single-channel EEG training method based on multi-person collaboration according to claim 1, characterized in that, Calculate the temporal stability index of the single-channel EEG data within a preset time window, and based on the single-channel EEG data, the temporal stability index, and the meditation assessment value, calculate the meditation index of the corresponding participant, including: Calculate the coefficient of variation of the single-channel EEG data within a preset time window, and use the reciprocal of the coefficient of variation as the temporal stability index of the single-channel EEG data; The meditation index of the corresponding participant is calculated based on the following preset formula: ;in, Let be the meditation index of the j-th participant; Let be the meditation assessment value of the j-th participant; The relative power of the Alpha wave, The weighting coefficient for the relative power of the Alpha wave; The relative power of the Gamma wave, The weighting coefficient for the relative power of the Gamma wave; The Alpha / Beta power ratio is... The weighting coefficient for the Alpha / Beta power ratio; The time-domain stability index is... The weighting coefficients for the time-domain stability index; The function represents a normalized mapping based on the individual historical baseline data of the j-th participant.

5. The single-channel EEG training method based on multi-person collaboration according to claim 1, characterized in that, Based on the meditation index of all non-eliminated participants, a group meditation index score is determined, including: A weight is calculated for the Zen index of each non-eliminated participant. The weight is calculated based on the non-eliminated participant quantity balance factor, the individual historical performance factor, and the current volatility suppression factor. The non-eliminated participant quantity balance factor is the reciprocal of the current number of non-eliminated participants. The individual historical performance factor is the ratio of the rolling mean of the Zen index of the non-eliminated participant to the rolling mean of the Zen index of all non-eliminated participants in the first historical period. The current volatility suppression factor is determined based on the standard deviation of the Zen index of each non-eliminated participant in the second historical period. The group meditation index score is calculated by summing the products of the meditation index and its corresponding weight for all non-eliminated participants using a dynamic weighted average algorithm.

6. The single-channel EEG training method based on multi-person collaboration according to claim 1, characterized in that, Based on the group meditation index score, brightness control instructions for the lighting fixtures are generated, including: The group meditation index score is compared with a first preset brightness threshold and a second preset brightness threshold to determine the corresponding target brightness level; wherein, when the group meditation index score is less than the first preset brightness threshold, the target brightness level is a dim brightness level; when the group meditation index score is between the first preset brightness threshold and the second preset brightness threshold, the target brightness level corresponds to a normal brightness level; when the group meditation index score is greater than the second preset brightness threshold, the target brightness level corresponds to a bright brightness level. Based on the group meditation index score and the target brightness level, a brightness control command for the luminaire is generated.

7. The single-channel EEG training method based on multi-person collaboration according to claim 1, characterized in that, Based on the meditation index of all non-eliminated participants, the group state level is determined, including: The individual state level of each non-eliminated participant is determined based on their meditation index; wherein the individual state level includes at least a first state level, a second state level, a third state level, and a fourth state level, and each state level corresponds to a different degree of individual cognitive state. The group status level is determined based on the individual status level with the highest proportion among all non-eliminated participants.

8. The single-channel EEG training method based on multi-person collaboration according to claim 7, characterized in that, The color value control instruction for the lamp is generated according to the group status level. Specifically, the target color value corresponding to the group status level is obtained based on a preset color value mapping table, and the color value control instruction for the lamp is generated based on the target color value.

9. The single-channel EEG training method based on multi-person collaboration according to claim 1, characterized in that, Also includes: Obtain the preset training target threshold; The group meditation index score is compared with a preset training target threshold to determine whether the training target has been achieved.

10. A single-channel EEG training system based on multi-person collaboration, characterized in that, include: The EEG data acquisition module is used to acquire single-channel EEG data from multiple participants. The meditation index calculation module is used to perform the meditation index calculation steps for each participant: extract the spectral feature vector within a preset time window from the single-channel EEG data corresponding to each participant, and input it into the pre-trained deep learning model to obtain the meditation assessment value of the corresponding participant. Calculate the temporal stability index of the single-channel EEG data within a preset time window, and calculate the meditation index of the corresponding participant based on the single-channel EEG data, the temporal stability index, and the meditation assessment value; wherein, the spectral feature vector includes at least the relative power of Alpha waves, the relative power of Gamma waves, and the Alpha / Beta power ratio; The monitoring module is used to continuously monitor the meditation index of each participant and determine the current state category of each participant based on the meditation index and its historical data. The group status assessment module is used to determine the group meditation index score or group status level based on the meditation index of all non-eliminated participants. The lighting control instruction generation module is used to generate brightness control instructions for the lighting fixtures based on the group meditation index score; or, to generate color value control instructions for the lighting fixtures based on the group state level.

Images