An electroencephalogram state recognition method for intelligent cockpit relaxation scene

Abstract

The application discloses an electroencephalogram state recognition method for a smart cabin relaxation scene, and relates to the technical field of electromyographic sensing, and comprises the following steps: in a driving state, first electroencephalogram data is collected for quality judgment, second electroencephalogram data meeting state recognition requirements is obtained and is preprocessed to obtain third electroencephalogram data; electroencephalogram features of the third electroencephalogram data are extracted within a preset time window and electroencephalogram state recognition is performed; according to electroencephalogram state recognition results corresponding to continuous time windows, stability is judged and state recognition confidence is determined; according to the stability judgment result and the state recognition confidence, a recognition result that can be called by a smart cabin is output. The application can obtain a more stable, reliable and applicable recognition result for a vehicle-mounted relaxation scene, and provides a complete electroencephalogram state recognition technical path for a smart cabin relaxation scene.

Description

Technical Field

[0001] This invention relates to the field of electromyography (EMG) sensing technology, and more specifically, to a method for EEG state recognition in a soothing scenario in a smart cockpit. Background Technology

[0002] In the fields of electromyography (EMG) sensing technology, including EEG signal processing, smart cockpits, human-computer interaction, physiological state recognition, in-vehicle health management, and computer program control, the rapid development of smart car and smart cockpit technologies means that current cockpit relaxation functions rely heavily on manual triggering or preset programs. This lacks precise perception of the real-time physiological state of drivers and passengers, making it impossible to autonomously determine user states such as fatigue, tension, drowsiness, and relaxation. Furthermore, it is difficult to dynamically adapt relaxation solutions to changes in state, resulting in low trigger accuracy, insufficient personalization, and poor adaptability. While EEG signals can directly reflect physiological and emotional states such as attention, relaxation level, and fatigue, these signals are weak and easily interfered with. In addition, the in-vehicle environment is affected by factors such as vehicle vibration, head movement, wearing deviations, and electromagnetic noise, leading to poor stability and large instantaneous fluctuations in raw EEG data. Directly using this data for cockpit control can easily result in misjudgments of state and frequent device switching.

[0003] Existing EEG recognition technology is mostly applied to single fields such as medical monitoring and fatigue warning, and is not adapted to the needs of in-vehicle comfort scenarios. It lacks an integrated processing solution for EEG data quality control, smoothing and noise reduction, feature extraction, state determination and confidence verification in the in-vehicle environment. It cannot stably output reliable results of the user's physical and mental state, and it is difficult to meet the actual application needs of autonomous and intelligent comfort control in smart cockpits. Summary of the Invention

[0004] To address the aforementioned issues, the present invention aims to provide a brainwave state recognition method for smart cockpit relaxation scenarios, applied in the field of electromyography sensing technology. This method addresses the technical problems of existing smart cockpit relaxation scenarios, such as the lack of a stable and reliable brainwave state recognition mechanism, the susceptibility of raw brainwave data or instantaneous parameters to noise, wearing status, individual differences, and short-term fluctuations, making it difficult to directly convert them into status labels that can be called by the vehicle system.

[0005] To achieve the above technical objectives, this application provides a method for EEG state recognition in smart cockpit relaxation scenarios, applied in the field of electromyography sensing technology, comprising the following steps: In driving mode, the first EEG data is collected for quality judgment, in order to obtain the second EEG data that meets the state recognition requirements and perform preprocessing to obtain the third EEG data. Extract the EEG features of the third EEG data within a preset time window and perform EEG state recognition; Based on the EEG state recognition results corresponding to the continuous time window, stability is judged and the confidence level of state recognition is determined; Based on the stability judgment result and the state recognition confidence level, the recognition result that can be called by the intelligent cockpit is output.

[0006] Preferably, when collecting the first EEG data, the first EEG data is collected using at least one connection method, wherein the connection method includes wired serial port / USB direct connection, Bluetooth wireless connection, WiFi local area network connection, vehicle bus protocol connection, NFC near field pairing connection, and 5G / mobile communication remote connection.

[0007] Preferably, when collecting the first EEG data, state assessment parameters, EEG frequency band energy, and basic raw data are collected as the first EEG data. The state assessment parameters include attention parameters and relaxation parameters; the EEG frequency band energy includes Delta frequency band energy, Theta frequency band energy, Alpha frequency band energy, Beta frequency band energy, and Gamma frequency band energy; and the basic raw data includes raw EEG signals.

[0008] Preferably, when judging the quality of the first EEG data, the current EEG data is judged to meet the state recognition requirements based on signal quality parameters, electrode contact status, data missing rate, number of abnormal peaks, short-term fluctuation amplitude or preset effective range, and the first EEG data that does not meet the state recognition requirements is marked as invalid.

[0009] Preferably, when preprocessing the second EEG data, the second EEG data is preprocessed by at least one of the following methods: noise reduction, filtering, smoothing, outlier removal, normalization, timestamp alignment, data compensation, or missing data processing.

[0010] Preferably, when extracting EEG features, based on the third EEG data, at least one of the following features is extracted within a preset time window: average attention parameter, average relaxation parameter, average energy of each EEG frequency band, Alpha / Beta ratio, Theta / Beta ratio, Alpha / Theta ratio, EEG fluctuation amplitude, rate of change, stability index, peak distribution, and energy change trend. The time window within the preset time window can be dynamically adjusted according to changes in user state and the needs of the cabin relaxation scenario.

[0011] Preferably, when identifying the EEG state, a finite state machine for the EEG state is constructed, and the legal transition logic between each state is defined; the rationality of the transition of the current candidate state is constrained by combining the historical state of the preceding time window, and the finite state machine is used for state constraint; instantaneous transition results that do not conform to the physiological change law are filtered out, and compliant candidate states are locked to obtain the EEG state.

[0012] Preferably, when performing stability judgment, the recognition result of the current time window is compared with the recognition result of the previous time window or multiple consecutive time windows to determine whether the state is continuously stable. Specifically, when judging whether the state is continuously stable, if the state label remains consistent in multiple consecutive time windows, or the change in state level is less than a preset threshold, the state is determined to be stable; if the state changes frequently, a valid state label is not output temporarily, or a result pending confirmation is output.

[0013] Preferably, when determining the confidence level, the confidence level of the state recognition result is determined by sequentially performing multi-dimensional indicator quantitative scoring, weighted fusion by weight allocation, correction and deduction of abnormal factors, interval mapping and rating, and dynamic time series verification.

[0014] Preferably, when outputting the recognition results that can be called by the smart cockpit, the status label, status level, confidence level, duration, change trend and validity mark are used as the recognition results.

[0015] The present invention discloses the following technical effects: This invention reduces the impact of unstable wear, vehicle vibration, head movements, poor contact, or transient noise on recognition results by performing signal quality assessment, noise reduction, filtering, smoothing, and outlier removal on EEG data.

[0016] This invention extracts features related to states such as tension, fatigue, relaxation, and attention fluctuations from attention parameters, relaxation parameters, and different EEG frequency bands by using time-domain, frequency-domain, or time-frequency feature extraction, thereby solving the problem that raw EEG data is difficult to use directly for cockpit control.

[0017] This invention improves the reliability of state recognition results by judging the stability of a continuous time window based on the persistence and trend of the state over a period of time.

[0018] When the signal quality is substandard, the status characteristics are not obvious, or the recognition results are unstable, this invention may not output a valid control trigger result, or may output results such as "invalid state", "state to be confirmed", or "insufficient confidence", thereby reducing the possibility of cockpit equipment malfunction.

[0019] This invention can provide stable input for intelligent cockpit rest mode, soothing scenes, music matching, and multimodal linkage control.

[0020] This invention can be deployed in vehicle infotainment systems, mobile apps, intermediate gateways, edge computing devices, or other computing devices connected to smart cockpits, offering strong deployment flexibility.

[0021] This invention provides a complete EEG state recognition technology path for relaxation scenarios in smart cockpits through a process of "signal quality judgment, feature extraction, state recognition, stability judgment, confidence assessment, and result output". Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a schematic diagram of the method described in this invention.

[0024] Figure 2 This is a schematic diagram of the module structure of the EEG state recognition device described in this invention.

[0025] Figure 3 This is a schematic diagram illustrating the relationship between the output and application of the EEG state recognition results described in this invention. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0027] like Figure 1 As shown, the present invention provides a method for EEG state recognition in smart cockpit relaxation scenarios, which is applied to the field of electromyography sensing technology. The method is used to take EEG data obtained by external or portable EEG acquisition devices, and after data quality judgment, preprocessing, feature extraction, state recognition, stability judgment and confidence assessment, output a state label or state level that can be called by smart cockpit relaxation scenarios.

[0028] Step S100: Acquire EEG data.

[0029] In one embodiment, the EEG data in step S100 can be collected by an external EEG acquisition device, a portable EEG acquisition terminal, a wearable EEG device, or a third-party EEG device connected to the vehicle system.

[0030] For example, the connection methods include wired serial port / USB direct connection, i.e., local wired transmission of external EEG devices; Bluetooth wireless connection, i.e., short-range communication of portable and wearable devices; WiFi local area network connection, i.e., data backhaul of terminal devices; vehicle bus protocol connection, i.e., third-party devices interfacing with the vehicle's CAN / LIN bus; NFC near-field pairing connection, i.e., fast authentication and binding of wearable devices; and 5G / mobile communication remote connection, i.e., data upload and aggregation of remote terminals.

[0031] For example, EEG data includes state assessment parameters, EEG frequency band energy, and basic raw data. The state assessment parameters include attention parameters and relaxation parameters; the EEG frequency band energy includes Delta frequency band energy, Theta frequency band energy, Alpha frequency band energy, Beta frequency band energy, and Gamma frequency band energy; and the basic raw data includes raw EEG signals.

[0032] Step S200 involves assessing the signal quality of the acquired EEG data.

[0033] In step S200 of one embodiment, the current EEG data is determined to meet the state recognition requirements based on the signal quality parameters output by the EEG acquisition device, electrode contact status, data missing rate, number of abnormal peaks, short-term fluctuation amplitude, or preset effective range.

[0034] For example, the state recognition requirements include signal quality parameter thresholds, electrode contact state requirements, data missing rate limits, abnormal peak control, short-term fluctuation amplitude range, frequency band energy compliance, and temporal continuity requirements. Among them, the signal quality parameter thresholds include signal-to-noise ratio, impedance value, and waveform regularity meeting the standards, with no serious power frequency, electromyography, or electrooculography interference; the electrode contact state requirements include all acquisition electrodes being properly attached, with no detachment, loosening, short circuits, or open circuits; the data missing rate limit includes the proportion of missing valid sampled data being lower than a preset allowable threshold; abnormal peak control includes the number of sudden spikes exceeding the physiological EEG amplitude, which must not exceed the limit; the short-term fluctuation amplitude range includes the short-term oscillation amplitude of the EEG waveform, which must fall within a preset effective physiological range; frequency band energy compliance includes the energy distribution of each EEG frequency band (α / β / θ / δ) conforming to the characteristics of normal human EEG; and the temporal continuity requirements include no gaps, jumps, or frame drops in the data sampling time sequence.

[0035] For example, by designing the above-mentioned state recognition requirements, the recognition accuracy is ensured by eliminating inferior and interfering data and avoiding misjudgments of brain states. By unifying data admission standards, the recognition conclusions are more repeatable, improving the stability of the results. Bad data can be screened out in advance, reducing computing power and resource consumption and avoiding invalid calculations. The data is guaranteed to conform to the objective characteristics of human brain electrophysiology and fit the physiological reality. It can quickly detect loose electrodes, poor contact, and hardware abnormalities, and meet the data acceptance specifications for experiments, testing, and clinical applications, as well as reduce the negative impact of environmental and behavioral interference on the algorithm.

[0036] For example, if the data quality does not meet the requirements during the judgment, the subsequent status recognition step will not be entered, and prompts such as "invalid signal" or "device not worn stably" will be output.

[0037] Step S300 involves preprocessing the EEG data after quality assessment.

[0038] In step S300 of one embodiment, preprocessing includes denoising, filtering, smoothing, outlier removal, normalization, timestamp alignment, data compensation, or missing data processing. The system can employ moving average, median filtering, low-pass filtering, band-pass filtering, or other digital signal processing methods to reduce the impact of instantaneous noise and abnormal jumps on the state recognition results.

[0039] For example, in order to solve the problem of unstable EEG data quality, the present invention preprocesses EEG data by means of signal quality judgment, outlier removal, filtering, smoothing and other methods. This overcomes the fact that EEG data is easily affected by factors such as loose wearing, poor contact, user head movement, vehicle vibration, electrode noise and other factors during the acquisition process, and improves the reliability of subsequent state recognition.

[0040] Step S400 involves extracting EEG features from the preprocessed EEG data.

[0041] In step S400 of one embodiment, time-domain features, frequency-domain features, or time-frequency features are extracted within a preset time window.

[0042] For example, since the data output by the EEG acquisition device is usually a continuously changing numerical stream, including attention parameters, relaxation parameters and multiple frequency band energy parameters, the above data itself cannot be directly used as the control basis for the relaxation scenario of the smart cockpit. In order to solve the problem that the raw EEG data is difficult to use directly for state judgment, the present invention converts the EEG data into feature parameters that can be used for state recognition through time domain, frequency domain or time-frequency feature extraction.

[0043] For example, the features extracted within a preset time window include the average value of attention parameters, the average value of relaxation parameters, the average energy of each EEG frequency band, the Alpha / Beta ratio, the Theta / Beta ratio, the Alpha / Theta ratio, the amplitude of EEG fluctuations, the rate of change, stability indicators, peak distribution, or energy change trends.

[0044] For example, the time window within the preset time window can be 5 to 120 seconds, preferably 10 to 60 seconds, and can also be dynamically adjusted according to changes in user status and cabin comfort requirements.

[0045] Step S500: Perform EEG state recognition.

[0046] In step S500 of one implementation, the user's state is matched and determined by combining the extracted feature parameters with a combination of threshold rules, weighted scoring, state machine, and machine learning classifier. The user's state is distinguished into types such as tension, fatigue, relaxation, attention fluctuation, shallow rest, tendency to deep rest, and inability to determine the state, and the user's EEG state is initially identified.

[0047] For example, when performing preliminary identification of the user's EEG state, feature parameters are normalized. All extracted EEG features, such as mean attention, mean relaxation, energy of each frequency band, Alpha / Beta, Theta / Beta, Alpha / Theta ratio, fluctuation amplitude, and rate of change, are uniformly mapped to the standard dimension range of 0-1 or 0-100. Extreme abnormal feature values ​​are removed, and personalized calibration is performed according to the individual baseline to eliminate the benchmark offset caused by individual differences in users and wearing deviations.

[0048] For example, after performing feature parameter normalization, a pre-defined feature threshold rule base for each state is established and initially screened, and corresponding feature threshold ranges are configured for each state such as tension, fatigue, relaxation, attention fluctuation, light rest, and deep rest tendency.

[0049] For example, in a tense state, the Beta frequency band energy is high, the relaxation level is low, and attention fluctuates greatly; in a fatigued state, the Theta frequency band energy increases, the attention parameter decreases, and the Alpha energy decays; in a relaxed state, the Alpha frequency band energy is consistently high, the relaxation parameter is high, and the ratio of high to low frequencies is within a comfortable range. A first round of coarse classification is performed on the normalized features using a fixed threshold rule to filter out the candidate state set and exclude obviously mismatched state types.

[0050] For example, in a state of tension, the core characteristic threshold ranges are as follows: relaxation level parameter: 20–45; attention parameter: 65–90; Beta frequency band energy: 70–95; Alpha frequency band energy: 25–50; Alpha / Beta ratio: 0.2–0.6; Theta / Beta ratio: 0.15–0.4; short-term EEG fluctuation amplitude: 60–85. The key criteria for judgment can be high Beta, low relaxation level, low α / β ratio, and large waveform fluctuations, which are consistent with the characteristics of sympathetic nerve excitation and mental tension.

[0051] For example, in a state of fatigue, the core characteristic threshold ranges are as follows: attention parameter: 15–40; relaxation parameter: 40–65; Theta band energy: 65–90; Beta band energy: 20–45; Theta / Beta ratio: 0.7–1.5; Alpha band energy: 30–55; EEG characteristic change rate: 45–70. The key criteria for judgment can be a significant increase in Theta, a significant decrease in attention, a high θ / β ratio, sluggish thinking, and mental fatigue.

[0052] For example, in a relaxed state, the core characteristic threshold ranges are as follows: relaxation level parameter: 70–95; attention parameter: 45–70; alpha band energy: 70–90; beta band energy: 30–55; alpha / beta ratio: 1.2–2.5; alpha / theta ratio: 1.0–2.0; EEG fluctuation amplitude: 15–35. The key criteria for judgment can be alpha dominance, high relaxation level, reasonable alpha / beta ratio, stable waveform, and physical and mental relaxation without tension or drowsiness.

[0053] For example, the core feature threshold range for attention fluctuation states is as follows: attention parameter: 35-75 (large fluctuations within the window); relaxation parameter: 35-60; Alpha band energy: 40-65; Beta band energy: 40-65; Alpha / Beta ratio: 0.5-1.1; EEG fluctuation amplitude: 50-75; short-term change rate of features: 60-90. The key judgment points can be that the energy of each band is in the middle without obvious dominance, attention fluctuates wildly, the waveform fluctuates violently, and it is impossible to focus stably.

[0054] For example, in a light resting state, the core characteristic threshold ranges are as follows: relaxation parameter: 60-85; attention parameter: 25-50; Alpha band energy: 55-75; Theta band energy: 50-70; Alpha / Theta ratio: 0.8-1.3; Beta band energy: 25-45; EEG stability index: 65-85. The key criteria for judgment can be that Alpha and Theta are relatively high, attention is reduced, relaxation is good, and the person is in a quiet, closed-eye, light resting state.

[0055] For example, the core characteristic threshold range for a deep rest tendency state is as follows: relaxation parameter: 75-95; attention parameter: 5-30; Theta band energy: 75-95; Delta band energy: 60-85; Alpha band energy: 20-45; Alpha / Theta ratio: 0.2-0.6; EEG fluctuation amplitude: 5-25. The key criteria for judgment can be that Theta and Delta are significantly increased, Alpha is decreased, attention is extremely low, and the state tends towards deep relaxation and drowsiness.

[0056] For example, based on the preset characteristics of each state, a unique weight is assigned to each type of EEG feature (e.g., Alpha energy and relaxation have higher weights, while transient fluctuation features have lower weights); the comprehensive score of each candidate state is calculated by accumulating the "feature normalization score × corresponding weight"; a scoring grading interval is set to initially divide the state into high matching, medium matching, and low matching levels, and the quantitative rating is completed.

[0057] For example, a finite state machine for EEG states is constructed, and the legal transition logic between each state is defined (e.g., the relaxed state will not instantly transition to severe fatigue, avoiding unreasonable sudden changes); combined with the historical states of the preceding time window, the rationality of the transition of the current candidate state is constrained, and the finite state machine is used for state constraints; instantaneous transition results that do not conform to the physiological change pattern are filtered out, and compliant candidate states are locked.

[0058] For example, a machine learning classifier is used for fine discrimination. The normalized multidimensional EEG feature vector is input into a pre-trained machine learning model (such as SVM, random forest, neural network, etc.). The model outputs the classification probability value of each state based on a large number of labeled user EEG samples. The baseline state with the highest probability is output as the judgment result of the machine model.

[0059] For example, the results of multiple models are fused for decision-making, which integrates the initial threshold screening results, weighted score ranking, state machine constraint results, and machine learning classification probability information; a voting mechanism, probability weighted fusion, and rule priority determination method are used to perform consistency verification on the output of multiple models; conflicting results are eliminated, and the unique target state with the highest matching degree and the most reasonable logic is retained.

[0060] Step S600: Perform a stability assessment of the continuous time window.

[0061] In step S600 of one embodiment, the identification result of the current time window is compared with the identification results of the previous time window or multiple consecutive time windows to determine whether the state remains stable.

[0062] For example, when determining whether a state is continuously stable, if the state label remains consistent within multiple consecutive time windows, or if the change in state level is less than a preset threshold, then the state is determined to be stable; if the state changes frequently, then a valid state label is not output temporarily, or a result pending confirmation is output.

[0063] For example, the present invention uses continuous time window stability judgment to output state results not only based on a single moment or a single data point, but also based on the state persistence and change trend over a period of time, thereby improving the reliability of state recognition results and reducing misjudgments caused by instantaneous fluctuations.

[0064] Step S700: Perform a confidence assessment.

[0065] In step S700 of one embodiment, the confidence level of the state recognition result is calculated or determined based on factors such as EEG data quality, feature strength, consistency of classification results, stability of continuous time windows, and degree of matching of user historical data.

[0066] For example, when determining the confidence level, the following steps are performed sequentially: multi-dimensional indicator quantification scoring, weighted fusion with weight allocation, correction and deduction of abnormal factors, interval mapping grading, and dynamic time-series verification. Specifically, multi-dimensional indicator quantification scoring converts data quality, feature strength, classification result consistency, time window stability, and historical matching degree into scores ranging from 0 to 100; weighted fusion assigns weights based on the impact of each indicator on the identification and sums them to obtain a comprehensive base score; correction and deduction of abnormal factors deducts confidence scores according to rules based on negative factors such as data defects, sudden fluctuations, and matching deviations; interval mapping grading maps the corrected values ​​to high / medium / low confidence levels to determine the corresponding credibility; and dynamic time-series verification compares confidence fluctuations through continuous windows, performs secondary verification for significant jumps, and outputs the final confidence value.

[0067] For example, outputting a status label when the signal quality is poor, the status characteristics are not obvious, or the recognition result is unstable may lead to false triggering of the smart cockpit device. To solve the problem of lack of credibility judgment of status recognition results, the present invention calculates confidence or judges validity, and does not output control trigger result when the confidence is insufficient, or outputs result such as "invalid status" or "status pending confirmation".

[0068] For example, when identifying confidence level, if the confidence level reaches a preset threshold, the status label, status level, and confidence level are output; if the confidence level is lower than the preset threshold, the in-vehicle control is not triggered, or the result of "insufficient confidence level" is output.

[0069] Step S800: Output the recognition results that can be called by the smart cockpit.

[0070] In step S800 of one embodiment, the output identification result may include status label, status level, confidence level, duration, change trend and validity mark.

[0071] For example, in order to solve the problem that EEG state recognition results are difficult to be called by vehicle system, the present invention outputs the EEG state recognition results as standardized state labels, state levels or state confidence levels, so that they can be called by vehicle system, mobile terminal, intermediate gateway, edge computing device or smart cockpit control module, providing input for subsequent soothing scenarios, rest modes or multimodal linkage control.

[0072] For example, this invention offers strong deployment flexibility, allowing output results to be transmitted to in-vehicle systems, mobile apps, intermediate gateways, edge computing devices, or smart cockpit control modules. It does not require changes to the vehicle's hardware structure, nor does it require the EEG acquisition device to be built into the vehicle itself, making it suitable for use with third-party EEG acquisition devices. For example, the vehicle system can then configure subsequent rest modes, soothing scenes, music matching, multimodal linkage, or scene settings based on the output results.

[0073] For example, such as Figure 2 As shown, this application also provides an EEG state recognition device. The device includes a data acquisition module, a signal quality judgment module, a data preprocessing module, a feature extraction module, a state recognition module, a stability judgment module, a confidence assessment module, and a result output module. Each module can be installed in an in-vehicle system, a mobile app, an intermediate gateway, an edge computing device, or other computing devices connected to a smart cockpit.

[0074] For example, such as Figure 3 As shown, the EEG state recognition results output by this application can be invoked by the intelligent cockpit's soothing scenario. The vehicle system, rest mode module, music matching module, soothing scenario configuration module, or multimodal linkage control module can determine whether to activate the soothing scenario or adjust subsequent control strategies based on the state label, state level, confidence level, duration, and trend of change. When the recognition result is invalid, the state is unstable, or the confidence level is insufficient, the system will not trigger in-vehicle linkage control.

[0075] In summary, this invention provides a complete EEG state recognition technology path for intelligent cockpit relaxation scenarios through a process of "signal quality judgment—feature extraction—state recognition—stability judgment—confidence assessment—result output." Compared to methods that judge the state solely based on instantaneous EEG parameters or a single threshold, this application can obtain more stable, reliable, and suitable recognition results for in-vehicle relaxation scenarios.

[0076] It should be understood that although the steps in the flowcharts of this application's embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some of the steps in the figures may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be performed alternately or in turn with other steps or at least a portion of the sub-steps or stages of other steps.

[0077] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for EEG state recognition in a soothing scenario in a smart cockpit, characterized in that, Includes the following steps: In driving mode, the first EEG data is collected for quality judgment, in order to obtain the second EEG data that meets the state recognition requirements and perform preprocessing to obtain the third EEG data. Extract the EEG features of the third EEG data within a preset time window and perform EEG state recognition; Based on the EEG state recognition results corresponding to the continuous time window, stability is judged and the confidence level of state recognition is determined; Based on the stability judgment result and the state recognition confidence level, the recognition result that can be called by the intelligent cockpit is output.

2. The EEG state recognition method for a relaxation scenario in a smart cockpit according to claim 1, characterized in that: When collecting the first EEG data, the first EEG data is collected using at least one connection method, including wired serial port / USB direct connection, Bluetooth wireless connection, WiFi local area network connection, vehicle bus protocol connection, NFC near field pairing connection, and 5G / mobile communication remote connection.

3. The EEG state recognition method for soothing scenarios in intelligent cockpits according to claim 1, characterized in that: When collecting the first EEG data, state assessment parameters, EEG frequency band energy, and basic raw data are collected as the first EEG data. The state assessment parameters include attention parameters and relaxation parameters; the EEG frequency band energy includes Delta frequency band energy, Theta frequency band energy, Alpha frequency band energy, Beta frequency band energy, and Gamma frequency band energy; and the basic raw data includes raw EEG signals.

4. The EEG state recognition method for soothing scenarios in intelligent cockpits according to claim 1, characterized in that: When assessing the quality of the first EEG data, the system determines whether the current EEG data meets the state recognition requirements based on signal quality parameters, electrode contact status, data missing rate, number of abnormal peaks, short-term fluctuation amplitude, or preset effective range, and marks the first EEG data that does not meet the state recognition requirements as invalid.

5. The EEG state recognition method for soothing scenarios in intelligent cockpits according to claim 1, characterized in that: When preprocessing the second EEG data, at least one of the following methods is used: noise reduction, filtering, smoothing, outlier removal, normalization, timestamp alignment, data compensation, or missing data processing.

6. The EEG state recognition method for a relaxation scenario in a smart cockpit according to claim 1, characterized in that: When extracting EEG features, based on the third EEG data, at least one of the following features is extracted within a preset time window: average attention parameter, average relaxation parameter, average energy of each EEG frequency band, Alpha / Beta ratio, Theta / Beta ratio, Alpha / Theta ratio, EEG fluctuation amplitude, rate of change, stability index, peak distribution, and energy change trend. The time window within the preset time window is dynamically adjusted according to changes in user state and the needs of the cabin relaxation scenario.

7. The EEG state recognition method for soothing scenarios in intelligent cockpits according to claim 1, characterized in that: When identifying EEG states, a finite state machine for EEG states is constructed, and the legal transition logic between each state is defined. The rationality of the transition of the current candidate state is constrained by combining the historical states of the preceding time window, and the finite state machine is used for state constraints. Instantaneous transition results that do not conform to the physiological change law are filtered out, and compliant candidate states are locked to obtain the EEG state.

8. The EEG state recognition method for soothing scenarios in intelligent cockpits according to claim 1, characterized in that: When performing stability judgment, the recognition result of the current time window is compared with the recognition result of the previous time window or multiple consecutive time windows to determine whether the state is continuously stable. Specifically, when judging whether the state is continuously stable, if the state label remains consistent in multiple consecutive time windows, or the change in state level is less than a preset threshold, the state is determined to be stable; if the state changes frequently, a valid state label is not output temporarily, or a result pending confirmation is output.

9. A method for EEG state recognition in a soothing scenario in a smart cockpit according to any one of claims 1-8, characterized in that: When determining the confidence level, the following steps are performed in sequence: multi-dimensional indicator quantitative scoring, weighted fusion by weight allocation, correction and deduction of abnormal factors, interval mapping and grading, and dynamic time series verification to determine the confidence level of the state recognition result.

10. The EEG state recognition method for a soothing scenario in a smart cockpit according to claim 9, characterized in that: When outputting recognition results that can be called by the smart cockpit, the status label, status level, confidence level, duration, change trend, and validity mark are used as the recognition results.

Images