Intention visualization warning cap based on brain-computer interface

Abstract

The invention provides an intention visualization warning cap based on a brain-computer interface, and relates to the technical field of electronic monitoring, and the intention visualization warning cap comprises the following steps: collecting original electroencephalogram signals of the scalp of a user in a non-invasive manner through a plurality of dry electrodes or semi-dry electrodes arranged at preset positions of a warning cap body lining; amplifying the original electroencephalogram signal to obtain an amplified original electroencephalogram signal, and filtering the amplified original electroencephalogram signal to obtain a preprocessed simulated electroencephalogram signal; converting the preprocessed analog electroencephalogram signal into an electroencephalogram signal in a digital form; the electroencephalogram signals in the digital form are calculated and analyzed, time domain, frequency domain or time frequency features related to the intention of the user are extracted, and feature vectors used for classification are generated. According to the invention, the stability, the adaptability and the maintainability of the whole equipment are improved.

Description

Technical Field

[0001] This invention relates to the field of brain-computer interface technology, and in particular to an intention visualization warning cap based on a brain-computer interface. Background Technology

[0002] For patients who are disabled due to severe neurological diseases, their cognitive functions may be intact, but they have lost their language and motor skills and are unable to communicate effectively with the outside world. This makes it difficult for caregivers to detect their basic needs in a timely manner, which seriously affects their quality of life and dignity and brings great challenges to nursing work.

[0003] Currently, high-precision brain-computer interface (BCI) systems mainly acquire and analyze electroencephalogram (EEG) signals to achieve functions such as character spelling, cursor control, or robotic arm manipulation. These systems are typically complex, costly, and require professional personnel for installation, calibration, and signal interpretation. Furthermore, their applications are often limited to laboratories or specific rehabilitation environments, making them difficult to popularize in daily care. Their output methods are mostly screen displays or mechanical movements, which are not intuitive or discreet enough to meet the needs of caregivers who require quick, long-distance understanding of the patient's condition without professional background.

[0004] Traditional patient call devices (such as manual buttons and pull cords) rely entirely on the patient's residual motor function and are unusable for paralyzed patients. Current technology lacks a daily assistive device that can be worn naturally around the clock and translate the patient's inner thoughts into visual signals that caregivers can intuitively and instantly understand. Integrating the five core components of a non-invasive brain-computer interface based on electroencephalography (EEG)—signal acquisition, preprocessing, feature extraction, classification, and control output—into a wearable device represents a crucial direction for the practical application and widespread adoption of BCI technology, with excellent application prospects. Different neurological diseases exhibit unique characteristics in their EEG manifestations, but through advanced signal processing and pattern recognition algorithms, it is still possible to extract stable patterns related to specific basic intentions from complex EEG signals and convert them into clear control commands. This forms the theoretical basis of this invention. This invention designs a universal wearable brain-computer interface device whose core algorithm platform possesses adaptability and scalability, not limited to a single disease. Through self-learning and adaptation of the algorithm to different users' neural signals, it achieves reliable recognition of diverse intentions. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide an intention visualization warning cap based on brain-computer interface, which improves the overall stability, adaptability and maintainability of the device.

[0006] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows: Firstly, an intention visualization warning cap based on a brain-computer interface includes: The signal acquisition module is used to non-invasively acquire raw electroencephalogram (EEG) signals from the user's scalp through multiple dry or semi-dry electrodes arranged in preset positions on the inner lining of the warning cap. The signal preprocessing module is used to amplify the original EEG signal to obtain the amplified original EEG signal, and then filter the amplified original EEG signal to obtain the preprocessed analog EEG signal. The signal conversion module is used to convert preprocessed analog EEG signals into digital EEG signals. The feature extraction module is used to perform calculations and analysis on digital EEG signals, extract time-domain, frequency-domain, or time-frequency features related to user intent, and generate feature vectors for classification. The intent recognition module is used to input feature vectors into a pre-trained intent recognition algorithm model for pattern matching and classification, so as to obtain a category instruction that represents the specific user intent. The control drive module is used to parse the intent category instruction into the corresponding display control command according to the preset mapping relationship; and drive the display module integrated on the external side of the warning cap to generate a visual light signal or pattern signal corresponding to the intent through the display control command.

[0007] Furthermore, by using multiple dry or semi-dry electrodes positioned at predetermined locations within the lining of the warning cap, raw electroencephalogram (EEG) signals from the user's scalp are non-invasively collected, including: Electrodes are placed in two preset areas corresponding to the international 10-20 EEG electrode placement system, including the frontal, central, or parietal regions, to cover the main scalp locations where EEG activity related to the user's intentions is generated. After ensuring stable contact between the electrodes and the user's scalp, raw EEG signals are collected continuously and at set intervals to obtain raw data reflecting brain electrical activity in real time. During the acquisition process, the potential difference between each electrode and the reference electrode is recorded simultaneously to obtain raw EEG signals at the microvolt level.

[0008] Furthermore, the original EEG signal is amplified to obtain an amplified original EEG signal, and then filtered to obtain a preprocessed analog EEG signal, including: The raw EEG signal is input into an instrumentation amplifier with high input impedance and high common-mode rejection ratio for differential amplification. The weak EEG signal is amplified to a voltage range suitable for processing with an appropriate gain, thus obtaining the amplified EEG signal. The amplified EEG signal was bandpass filtered to retain the effective frequency band of the EEG signal, while filtering out noise components such as power frequency interference, electromyography artifacts and electrooculography artifacts, to obtain a preliminary filtered signal. The preliminary filtered signal is further input into a notch filter to eliminate power frequency interference at a specific frequency, so as to obtain a pure preprocessed analog EEG signal.

[0009] Furthermore, the preprocessed analog EEG signals are converted into digital EEG signals, including: The preprocessed analog EEG signal is input into the analog-to-digital converter unit and sampled using a sufficiently high sampling rate that meets the requirements of the EEG signal sampling theorem in order to completely preserve the time-domain information of the signal and obtain the sampled discrete-time signal. Discrete-time sampled signals are used as input data, and analog-to-digital converters with sufficiently high resolution are used to quantize the sampled signals, converting the continuous amplitude of the analog signal into discrete digital quantities, thereby generating discrete-time and discrete-amplitude digital EEG signals. The digital EEG signal is processed to remove DC bias, eliminating the DC bias introduced by the electrode amplifier circuit. The DC-biased digital signal is then buffered and packaged according to design requirements to form a digital EEG data stream.

[0010] Furthermore, computational analysis is performed on the digital EEG signals to extract time-domain, frequency-domain, or time-frequency features related to the user's intent, and feature vectors for classification are generated, including: The digital EEG signal is segmented by sliding window, dividing the continuous EEG data into fixed-length analysis periods, with each period serving as a basic analysis unit; The digital EEG signals within the analysis unit are used to perform feature calculations and extract various feature parameters that reflect the user's intentions. These feature parameters include time-domain statistical features, power spectrum features of specific frequency bands, time-frequency distribution features, and nonlinear dynamic features. By combining and normalizing multiple feature parameters, a high-dimensional feature vector is obtained.

[0011] Furthermore, the feature vector is input into a pre-trained intent recognition algorithm model for pattern matching and classification to obtain a category instruction representing a specific user intent, including: The intention recognition algorithm model is loaded into the embedded microprocessor to run the feature vector. The intention recognition algorithm model is a classifier built based on the pattern recognition algorithm. It internally stores classification decision parameters learned from a large number of samples. After loading, the model has the ability to perform classification calculations on the current input feature vector. Based on the loaded intent recognition algorithm model, pattern matching calculation is performed on the feature vectors, and the similarity between various preset intent patterns is calculated. The intent patterns include EEG patterns associated with steady-state visual evoked potentials, event-related potentials P300, motor imagery, or specific cognitive tasks, so as to obtain a set of similarity values ​​that reflect the degree of matching between the feature vectors and various intent patterns. Based on the similarity value, the intent category with the highest similarity is selected as the current recognition result, resulting in a category instruction output that represents the specific user intent.

[0012] Furthermore, based on a preset mapping relationship, the intent category instruction is parsed into corresponding display control commands, including: The intent-display map is pre-stored in the microprocessor. The map defines the correspondence between each intent category instruction and the corresponding display effect parameters, including color, brightness, blink frequency, pattern type, or text content. Based on the intent category instruction, the corresponding display effect parameters are retrieved from the mapping table, and a display control command containing the parameters is generated; The display control commands are encapsulated according to the communication protocol required by the display module and sent to the display driver circuit.

[0013] Furthermore, the display control command drives the integrated display module on the external side of the warning cap to generate a visual light signal or pattern signal corresponding to the intent, including: The display module's driving circuit receives display control commands and parses the data content in the commands to extract display effect parameters, including color, brightness, flicker frequency, or pattern type. The display effect parameters are used as the control basis. The drive circuit generates corresponding electrical signals according to the parameters, and controls the LED array, LED strip or miniature display screen integrated on the outside of the cap to display according to the specified color, brightness, flashing mode or pattern. Based on the display effect, the display module continues to present the visual signal, maintaining a visual cue state corresponding to the user's current intent, until a new intent category instruction is recognized and a new display control command is generated, driving the display module to update the display content.

[0014] In a second aspect, a computing device includes: One or more processors; A storage device for storing one or more programs, which, when executed by one or more processors, cause the one or more processors to perform the program.

[0015] Thirdly, a computer-readable storage medium storing a program that, when executed by a processor, performs the aforementioned actions.

[0016] The above-described solution of the present invention has at least the following beneficial effects: Non-invasive EEG signal acquisition is achieved through the arrangement of semi-dry electrodes according to the international 10-20 system. Refined signal preprocessing is performed via high input impedance differential amplification, bandpass filtering combined with notch filtering, analog-to-digital conversion according to the sampling theorem to eliminate DC offset, multi-dimensional features are extracted through sliding window segmentation to generate normalized feature vectors, and these feature vectors are input into a pre-trained pattern recognition algorithm model for multi-intent pattern similarity matching and classification. Based on a preset intent-display mapping table, these commands are parsed into display control commands, driving the display module to generate visual signals. This series of techniques effectively overcomes the shortcomings of existing brain-computer interface devices, such as non-standard signal acquisition. Poor preprocessing noise reduction makes it difficult to extract effective EEG features; single intention recognition feature analysis and fuzzy matching and classification logic result in low recognition accuracy and slow response; and the lack of precise mapping between intention commands and visualization display leads to a mismatch between the displayed signal and the user's true intention, making it difficult for caregivers to quickly and intuitively identify the intention. At the same time, the poor connection between signal processing and intention conversion links causes insufficient overall system stability and practicality. Therefore, we have achieved standardized and refined processing of the entire process of EEG signal acquisition and display driving, which improves the purity of the original EEG signal, the accuracy of feature extraction, and the precision and response efficiency of user intention recognition.

[0017] It achieves precise mapping and rapid conversion between intent category commands and visual light and pattern signals, enabling the display module to output visual signals corresponding to the user's true intent in real time. This allows nursing staff to intuitively and quickly identify user needs without the need for specialized equipment. At the same time, it makes the signal processing and intent conversion system of the warning cap more standardized, stable, and operable, improving the overall practical value and scalability of the device. It also builds an efficient and reliable silent communication channel for disabled patients who have lost their ability to speak and move independently. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of an intent visualization warning cap based on a brain-computer interface provided in an embodiment of the present invention.

[0019] Figure 2 This is a schematic diagram illustrating the process of classifying and identifying the state of a support rod based on vibration and tilt feature parameters to obtain classification results in a brain-computer interface-based intention visualization warning cap according to an embodiment of the present invention. Detailed Implementation

[0020] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0021] like Figure 1 As shown, an embodiment of the present invention proposes an intent visualization warning cap based on a brain-computer interface, comprising: The signal acquisition module is used to non-invasively acquire raw electroencephalogram (EEG) signals from the user's scalp through multiple dry or semi-dry electrodes arranged in preset positions on the inner lining of the warning cap. The signal preprocessing module is used to amplify the original EEG signal to obtain the amplified original EEG signal, and then filter the amplified original EEG signal to obtain the preprocessed analog EEG signal. The signal conversion module is used to convert preprocessed analog EEG signals into digital EEG signals. The feature extraction module is used to perform calculations and analysis on digital EEG signals, extract time-domain, frequency-domain, or time-frequency features related to user intent, and generate feature vectors for classification. The intent recognition module is used to input feature vectors into a pre-trained intent recognition algorithm model for pattern matching and classification, so as to obtain a category instruction that represents the specific user intent. The control drive module is used to parse the intent category instruction into the corresponding display control command according to the preset mapping relationship; and drive the display module integrated on the external side of the warning cap to generate a visual light signal or pattern signal corresponding to the intent through the display control command.

[0022] In this embodiment of the invention, the intention recognition and visual warning of disabled patients are broken down into six standardized modular processes: signal acquisition, preprocessing, conversion, feature extraction, intention recognition, and control drive. Raw EEG signals are non-invasively acquired using semi-dry electrodes, amplified and filtered for analog EEG signal preprocessing, and digitized via analog-to-digital conversion. Multi-dimensional features related to user intentions are extracted and classification feature vectors are generated. These feature vectors are input into a pre-trained algorithm model to complete intention pattern matching and classification. Finally, intention commands are parsed according to a preset mapping relationship, and the display module is driven to generate corresponding visual light or pattern signals. This approach effectively overcomes the limitations of existing brain-computer interface devices, such as disjointed signal processing, susceptibility to noise interference leading to low intention recognition accuracy, and the inability to visualize intention commands. The lack of a standardized process for converting EEG signals makes it difficult to quickly and accurately identify the inner intentions of disabled patients. This also hinders nurses from intuitively and immediately understanding patient needs without professional background. This solution addresses the technical challenges of achieving a seamless and standardized process from EEG signal acquisition to intention visualization alerts. This improves the purity of EEG signal processing and the accuracy and efficiency of user intention recognition. It enables precise and rapid conversion of patient thoughts and intentions into intuitive, visual signals, allowing nurses to instantly identify patients' specific basic intentions through the visual signals of the alert cap without the need for specialized instruments or knowledge. This builds an efficient and reliable silent communication bridge between disabled patients who have lost language and motor skills and nurses. Simultaneously, the modular process design enhances the overall stability, adaptability, and maintainability of the equipment.

[0023] In a preferred embodiment of the present invention, raw electroencephalogram (EEG) signals from the user's scalp are non-invasively collected by using multiple dry or semi-dry electrodes arranged at predetermined positions within the lining of the warning cap, including: Electrodes are placed in two preset areas corresponding to the international 10-20 EEG electrode placement system. These areas include the forehead, central region, or parietal region, to cover the main scalp locations where EEG activity related to user intent is generated. Specifically, this involves determining the physical position of the warning cap liner in contact with the forehead, central region, or parietal region of the human scalp, according to the positioning specifications of the international 10-20 EEG electrode placement system. These specifications are based on specific anatomical landmarks of the human scalp and determine the precise locations of each EEG acquisition area through proportional division, ensuring that the electrode placement corresponds to the core areas of EEG activity. Subsequently, based on the user intent recognition requirements, two preset areas are selected from the forehead, central region, and parietal region as electrode placement areas. The selection principle is to cover the main scalp locations where EEG activity related to basic intents such as pain, seeking help, and toileting is generated, ensuring that the collected EEG signals can reflect the electrophysiological changes of the corresponding brain thought activities. Next, multiple dry bioelectric signal electrode contacts are fixed at corresponding positions in the two pre-defined areas of the cap liner. The number of electrodes is configured according to the complexity of the intent recognition. The simplified configuration can place one electrode in each of the two areas, while the standard configuration can place two or more electrodes in each area. All electrodes are electrically connected to the signal processing and control module inside the cap through flexible printed circuits to ensure stable transmission of the signals collected by the electrodes. At the same time, a reference electrode is placed at a fixed reference position on the scalp. The reference electrode works in conjunction with the acquisition electrode to calculate the potential difference during the acquisition process. The reference electrode also adopts a dry structure, is fixed to the cap liner, and completes the electrical connection with the signal processing module.

[0024] After ensuring stable contact between the electrodes and the user's scalp, raw EEG signals are acquired continuously and at set intervals to obtain raw data reflecting brain electrical activity in real time. Specifically, this involves: placing an alarm cap on the user, ensuring the cap's soft, breathable, and elastic material conforms to the user's head contour; adjusting the cap's position to ensure all positioned acquisition and reference electrodes maintain tight and stable contact with the user's scalp without loosening or shifting, guaranteeing contact stability for signal acquisition; then activating the alarm cap's power supply and signal acquisition modules, putting the signal acquisition module into operation; simultaneously activating both continuous and set-period acquisition modes, with both modes running concurrently. In the acquisition mode, the electrodes continuously capture EEG signals from the scalp surface at a preset sampling frequency without time intervals. In the set-period acquisition mode, the EEG signals are acquired in segments according to a preset time period. The duration of the period is set according to clinical EEG acquisition standards and the real-time requirements of intention recognition, generally 0.5 to 2 seconds per acquisition cycle. Within each cycle, a signal is quickly acquired and temporarily stored. Through the combination of the two acquisition modes, raw data that accurately reflects brain electrical activity is acquired in real time. The acquired raw data is immediately transmitted to the temporary storage unit of the signal acquisition module to avoid data loss and provide a complete raw data source for signal processing.

[0025] During the acquisition process, the potential difference between each electrode and the reference electrode is recorded synchronously to obtain raw EEG signals at the microvolt level. Specifically, during the EEG signal acquisition process, the potential detection unit of the signal acquisition module is activated synchronously to detect and record the potential changes between each acquisition electrode and the reference electrode in real time. The potential detection unit captures the potential values ​​of the acquisition electrode and the reference electrode in real time through the circuit. Using the potential difference calculation formula, the potential difference between the acquisition electrode and the reference electrode is equal to the real-time potential value of the acquisition electrode minus the real-time potential value of the reference electrode. The potential data of each set of acquisition electrodes and reference electrodes are calculated in real time to obtain the potential difference value corresponding to each acquisition electrode. Since the potential signal generated by brain electrical activity is a weak electrical signal, the calculated potential difference value is in the microvolt level, with a value range of 10-100 microvolts, which is consistent with the physiological characteristics of scalp surface EEG signals. The microvolt-level potential difference values ​​corresponding to all acquisition electrodes are integrated and encoded to form a continuous raw EEG signal at the microvolt level that reflects brain electrical activity. Then, the raw EEG signal is transmitted in real time to the signal amplification and filtering unit of the signal processing and control module to enter the signal preprocessing stage.

[0026] In this embodiment of the invention, electrodes are arranged according to the international 10-20 EEG electrode placement system standard to accurately cover the main scalp areas of brain activity. This overcomes the problems of arbitrary electrode placement and poor signal acquisition targeting in traditional EEG acquisition methods, ensuring that the acquired raw EEG signals are correlated with the user's intent, providing an effective and accurate signal foundation for subsequent intent recognition. By ensuring stable contact between the electrodes and the scalp and employing a dual-mode acquisition method with continuous and set periods, the problems of EEG signal acquisition being easily affected by contact status, incomplete data acquisition, or insufficient real-time performance are overcome. This achieves all-weather, uninterrupted real-time acquisition of raw brain activity data, avoiding... The loss of key signals is eliminated; by synchronously calculating the potential difference between the acquisition electrode and the reference electrode to generate microvolt-level raw EEG signals, the problem of no benchmark and no numerical reference of directly acquired EEG signals is overcome. It accurately captures the weak EEG signals generated by the brain and completely preserves the original characteristics of EEG signals related to the user's intentions. This lays a reliable data foundation for signal amplification, filtering, feature extraction and other processing links. At the same time, the non-invasive acquisition method of dry electrodes does not require professional operation and is suitable for patients to wear for a long time. It solves the problems of complex operation and limited use scenarios of traditional high-precision brain-computer interface acquisition devices, and improves the practicality and popularity of the device.

[0027] In a preferred embodiment of the present invention, the original EEG signal is amplified to obtain an amplified original EEG signal, and the amplified original EEG signal is filtered to obtain a preprocessed analog EEG signal, including: The raw EEG signal is input to an instrumentation amplifier with high input impedance and high common-mode rejection ratio for differential amplification. The weak EEG signal is then amplified to a suitable voltage range for processing with an appropriate gain, resulting in an amplified EEG signal. Specifically, the microvolt-level raw EEG signal transmitted from the signal acquisition module is directly connected to the signal input terminal of the instrumentation amplifier within the signal processing and control module. This instrumentation amplifier is a dedicated signal amplification device with high input impedance and high common-mode rejection ratio. Its input impedance is configured at the gigaohm level, and its common-mode rejection ratio is adjusted to 80 dB or higher, effectively preventing signal attenuation of the weak EEG signal during the input amplification stage. Simultaneously, it strongly suppresses external common-mode noise mixed in during the acquisition process, ensuring that the amplification process only targets the valid EEG signal. Based on the voltage adaptation requirements of the signal processing stage, a fixed gain parameter is set for the instrumentation amplifier. The selection of the gain parameter is based on the core standard of accurately amplifying the microvolt-level raw EEG signal to the millivolt level. The amplification calculation of the EEG signal follows the basic voltage amplification formula: amplified EEG signal voltage value = raw EEG signal voltage value × instrumentation amplifier gain.

[0028] The voltage value of the raw EEG signal is in the physiological range of 10 microvolts to 100 microvolts. If the voltage value of the raw EEG signal at a certain moment is 50 microvolts, and the gain of the instrumentation amplifier is set to 1000 times, the voltage value of the amplified EEG signal at that moment can be calculated to be 50 microvolts multiplied by 1000, which equals 50,000 microvolts, or 50 millivolts. This voltage value is within the standard adaptation voltage range of subsequent analog-to-digital conversion, feature extraction and other stages. The instrumentation amplifier amplifies the positive and negative polarity signals of the raw EEG signal simultaneously through differential amplification. While completing the signal amplification, it further removes residual common-mode noise, and finally obtains an amplified EEG signal that retains only the characteristics of brain electrical activity potential changes. This signal is directly transmitted to the filtering stage for processing.

[0029] The amplified EEG signal undergoes bandpass filtering to retain its effective frequency band while removing noise components such as power line interference, electromyography (EMG) artifacts, and electrooculography (EOG) artifacts, resulting in a preliminary filtered signal. Specifically, the amplified EEG signal is seamlessly connected to the input of the bandpass filter within the signal amplification and filtering unit. Based on the physiological characteristics of human brain electrical activity, a fixed effective passband frequency is set for the bandpass filter. The lower limit of the passband frequency is set to 0.5 Hz, and the upper limit is set to 30 Hz. This frequency band is considered the effective frequency band for EEG signals that accurately reflect brain activity; signals exceeding this frequency band are excluded. To eliminate noise, the bandpass filter possesses precise frequency selection and signal blocking characteristics, allowing only valid EEG signals within the 0.5 Hz to 30 Hz passband to pass completely. Simultaneously, it selectively blocks and attenuates various noise components outside the passband. Low-frequency drift noise below 0.5 Hz and EEG artifacts are directly blocked by the filter, preventing them from entering the signal path. EMG artifacts above 30 Hz are attenuated by the filter, with an attenuation sufficient to eliminate their interference with the valid signal. The bandpass filter can also initially filter out some high-frequency power frequency interference clutter, reducing the signal strength of power frequency interference. After the amplified EEG signal undergoes frequency selection and noise reduction processing by the bandpass filter, most of the mixed power frequency interference, EMG artifacts, and EMG artifacts are effectively removed, retaining only the valid EEG signal components that reflect the user's intent. The resulting preliminary filtered signal is still an analog signal, fully preserving the original time and frequency domain characteristics of the valid EEG signal, and is directly transmitted to the next notch filter stage for deep noise reduction.

[0030] The preliminary filtered signal is further input into a notch filter to eliminate power frequency interference at a specific frequency, thereby obtaining a clean pre-processed analog EEG signal. Specifically, this involves connecting the preliminary filtered signal to the input of a notch filter within the signal amplification and filtering unit. This notch filter is a narrowband stop filter, capable of deep attenuation of a specific single-frequency signal. Based on the power frequency standard of civilian power grids, the stopband center frequency of the notch filter is precisely set to 50 Hz, and the signal attenuation amplitude is adjusted to 40 dB or higher to ensure complete attenuation and elimination of the 50 Hz power frequency interference signal. The remaining 50 Hz specific frequency power frequency interference in the preliminary filtered signal is then eliminated. Frequency interference is deeply suppressed when passing through the notch filter. The voltage value of this type of interference signal is attenuated to a level that has no interference capability. The effective EEG signal in the preliminary filtered signal, because its frequency is in the range of 0.5 Hz to 30 Hz, can pass through the notch filter without attenuation. After the targeted deep noise reduction processing of the notch filter, the residual 50 Hz power frequency interference in the preliminary filtered signal is completely eliminated. The signal obtained at this time is a pure analog EEG signal with no obvious noise interference and complete preservation of effective features. This signal is the preprocessed analog EEG signal, which will be directly sent to the analog-to-digital conversion unit of the signal processing and control module for the next step of analog-to-digital signal processing.

[0031] In this embodiment of the invention, a high-input-impedance, high-common-mode-rejection-ratio instrumentation amplifier is used to perform quantitative gain differential amplification of the microvolt-level raw EEG signal. Combined with fixed-frequency bandpass filtering, multiple types of noise are initially and accurately filtered out. Then, a dedicated power frequency notch filter completely eliminates residual 50 Hz power frequency interference. This effectively overcomes the technical problems of existing brain-computer interface devices, such as the tendency to amplify noise synchronously when amplifying weak EEG signals, and the lack of layered design in the filtering stage leading to incomplete noise removal. These problems result in the effective features of the pre-processed EEG signal being masked by noise, affecting the accuracy of subsequent intention recognition. Simultaneously, this invention solves the technical pain points of traditional EEG signal preprocessing equipment, which is complex in structure, requires professional personnel to adjust parameters, and is difficult to integrate into wearable devices. This process achieves precise amplification and layering of the raw EEG signal, along with targeted noise reduction, ultimately yielding a clean pre-processed analog EEG signal that retains complete and effective EEG characteristics. This provides a high-quality, highly reliable signal foundation for subsequent analog-to-digital conversion, feature extraction, and intent recognition, improving the accuracy and timeliness of subsequent user intent recognition. Simultaneously, the modular miniaturization and parameter-fixed design of the entire pre-processing workflow seamlessly integrates into the signal processing and control module of the warning cap, eliminating the need for professional operation and parameter tuning. This perfectly adapts to daily care scenarios in homes, communities, and general medical institutions, further enhancing the integration, practicality, and accessibility of the warning cap device, meeting the design and usage requirements of wearable medical assistive devices.

[0032] In a preferred embodiment of the present invention, converting the preprocessed analog EEG signal into a digital EEG signal includes: The preprocessed analog EEG signal is input to the analog-to-digital converter (ADC) unit and sampled at a sufficiently high sampling rate that meets the requirements of the EEG sampling theorem to fully preserve the time-domain information of the signal, resulting in a sampled discrete-time signal. Specifically, the amplified and filtered clean preprocessed analog EEG signal is directly sent to the ADC unit within the signal processing and control module. The sampling module of this unit samples the analog EEG signal, strictly adhering to the EEG sampling theorem. The effective highest frequency of the preprocessed analog EEG signal is determined to be 30 Hz. According to the sampling theorem, the sampling rate must be set to at least twice the highest frequency of the signal. Calculations show that the base sampling rate is 2 times... 30 Hz is equal to 60 Hz. To fully preserve the temporal details of the EEG signal, 250 Hz was actually selected as the sampling rate. This sampling rate is much higher than the basic requirement and can effectively avoid signal aliasing distortion. The sampling module of the analog-to-digital conversion unit collects the continuous preprocessed analog EEG signal at a fixed time interval of 250 Hz, that is, it completes the acquisition and recording of the signal voltage value every 4 milliseconds. It captures the real-time voltage amplitude of the analog EEG signal at each sampling moment in sequence, transforming the originally continuously changing time-domain analog signal into a discrete-time signal with voltage values ​​only at discrete time points. This discrete-time signal retains all the temporal characteristics of the original analog EEG signal, providing basic data for quantization processing.

[0033] Discrete-time sampled signals are used as input data, and a sufficiently high-resolution analog-to-digital converter (ADC) is used to quantize the sampled signals, converting the continuous amplitude of the analog signal into discrete digital quantities to generate discrete-time, discrete-amplitude digital EEG signals. Specifically, this involves: sending the discrete-time signal as input data to the quantization module of the ADC unit for quantization processing. The quantization module uses a high-resolution ADC, selecting 16-bit resolution as the quantization standard to ensure that the quantized digital signal can accurately reproduce the amplitude characteristics of the analog signal. The full-scale voltage range of the ADC is determined to be ±5 volts, and the total voltage range is calculated to be 5 volts plus 5 volts equal to 10 volts. The quantization step size is calculated based on the resolution, using the formula: total voltage range divided by... With a resolution of 2 bits, the quantization step size is calculated to be 10 volts divided by 2 to the power of 16, which is approximately 0.0001526 volts, or 0.1526 millivolts. The quantization module matches the voltage amplitude at each sampling moment in the discrete-time signal with the calculated quantization step size, converting the continuous voltage amplitude values ​​into corresponding discrete decimal digital values. Each voltage amplitude can be accurately mapped to a unique digital value, completing the amplitude conversion from analog signal to digital signal. After this process, the signal, which was originally only discrete in time, achieves dual discreteness in time and amplitude, ultimately generating a discrete-time, discrete-amplitude digital EEG signal, which is then promptly transmitted to the processing module of the analog-to-digital converter unit.

[0034] The digital EEG signal undergoes DC offset removal processing to eliminate the DC bias introduced by the electrode amplification circuit. Based on design requirements, the DC-off digital signal is buffered and packaged to form a digital EEG data stream. Specifically, this involves: removing the DC offset from the digital EEG signal, which is introduced by the hardware characteristics of the electrodes and amplification circuit and is unrelated to the user's brain activity, thus requiring complete elimination; selecting a continuous segment of digital EEG signal as an analysis sample, containing several consecutive sampling points; recording the digital voltage value of each sampling point as the voltage value of the first sampling point, the second sampling point, and so on up to the Nth sampling point; calculating the DC offset of this sample, which is calculated by dividing the sum of all sampling point voltage values ​​by the number of sampling points. That is, the DC offset equals the sum of the first sampling point voltage value plus the second sampling point voltage value plus... plus the Nth sampling point voltage value, then divided by N. After the DC offset is applied, the voltage value of each sampling point in the digital EEG signal is corrected. The correction formula is that the corrected sampling point voltage value is equal to the original sampling point voltage value minus the DC offset. The correction of all sampling points is completed in sequence to completely eliminate the DC bias introduced by the circuit, ensuring that the digital EEG signal only reflects the user's brain electrical activity characteristics. After the DC offset removal process is completed, according to the subsequent processing design requirements of the embedded microprocessor, the corrected digital EEG signal is sent to the buffer unit of the analog-to-digital conversion unit for temporary storage. Then, the digital signal is packaged according to a fixed data frame format. Each data frame contains key information such as the sampling timestamp, the number of EEG acquisition channels, and the sampling point data within that time period. The consecutive data packets are sequentially connected to form a standardized and structured digital EEG data stream. This data stream is directly transmitted to the embedded microprocessor to provide a standardized digital signal source for feature extraction and intent recognition.

[0035] In this embodiment of the invention, the analog EEG signal is sampled by setting an appropriate high sampling rate according to the sampling theorem, and then accurately quantized by a 16-bit high-resolution analog-to-digital converter. The DC offset is then eliminated by calculating the mean, and the data is packaged into a digital EEG data stream according to a standard format. This effectively overcomes the technical problems of existing brain-computer interface devices in the analog-to-digital conversion stage, such as signal aliasing and distortion due to insufficient sampling rate, amplitude restoration deviation due to low resolution, and hardware interference in subsequent feature extraction due to ineffective removal of DC offset. Furthermore, the lack of a standardized signal format makes it difficult for subsequent algorithm modules to recognize and process the signal. It also solves the problem that traditional high-precision EEG analog-to-digital conversion devices are large in size and consume high power, making them unsuitable for integration into wearable warning caps. Addressing key technical challenges, this implementation process achieves precise and distortion-free conversion from preprocessed analog EEG signals to digital EEG data streams. It fully preserves the time-domain and amplitude characteristics of EEG signals related to user intent, eliminates irrelevant interference introduced by hardware, and generates standardized digital EEG data streams that can seamlessly interface with embedded microprocessors, improving the accuracy and efficiency of feature extraction and intent recognition. Simultaneously, the entire analog-to-digital conversion process relies on miniaturized, low-power analog-to-digital conversion units, which can be seamlessly integrated into the signal processing and control module of the warning cap. This meets the size and power consumption requirements of wearable medical assistive devices, enhancing the integration, practicality, and portability of the warning cap device, making it suitable for daily care scenarios such as homes and general medical institutions.

[0036] like Figure 2 As shown, in another preferred embodiment of the present invention, computational analysis is performed on the digital form of the electroencephalogram (EEG) signal to extract time-domain, frequency-domain, or time-frequency features related to the user's intent, and a feature vector for classification is generated, including: A sliding window segmentation method is used to segment the digital EEG signal, dividing the continuous EEG data into fixed-length analysis periods. Each period serves as a basic analysis unit. Specifically, the feature extraction module performs sliding window segmentation on the continuous digital EEG data stream. The segmentation parameters are set to balance the preservation of the temporal features of the EEG signal with the real-time requirements of intent recognition. First, the fixed length of the sliding window is determined to be 2 seconds. Combined with the 250 Hz sampling rate set in the previous analog-to-digital conversion stage, the number of sampling points contained in a single window is calculated to be 2 seconds multiplied by 250 Hz, which equals 500 sampling points. At the same time, the window sliding step size is set to 0.5 seconds, and the corresponding number of sampling points is 0.5 seconds multiplied by 250 Hz, which equals 125 sampling points. After setting the parameters, the continuous digital EEG data stream is initially segmented using a fixed-length window to obtain the first analysis unit containing 500 sampling points. Subsequently, the window slides continuously along the time axis at a set step size of 125 sampling points, and the resulting analysis units are segmented sequentially. There is an overlap of 375 sampling points between adjacent analysis units. Through this sliding segmentation method, the continuous and uninterrupted digital EEG data stream is divided into multiple independent analysis periods of fixed length. Each period serves as a basic analysis unit, ensuring the integrity of the EEG signal within each analysis unit. By overlapping windows and continuously sliding, the original data stream is processed without omission, while controlling the amount of data in a single window, thus balancing the efficiency and real-time performance of feature calculation.

[0037] The system performs feature calculations on the digital EEG signals within the analysis unit, extracting various feature parameters reflecting user intent. These parameters include time-domain statistical features, power spectrum features of specific frequency bands, time-frequency distribution features, and nonlinear dynamic features. Specifically, the feature extraction module performs multi-dimensional feature calculations on the 500 sampling points of the digital EEG signals within each basic analysis unit, sequentially extracting time-domain statistical features, power spectrum features of specific frequency bands, time-frequency distribution features, and nonlinear dynamic features. The time-domain statistical feature calculation directly performs statistical calculations on the voltage values ​​of the raw digital EEG signals within the analysis unit, extracting the mean, variance, root mean square, and other parameters. The peak value has four core characteristics. The mean value is calculated as follows: Time-domain mean = Sum of voltage values ​​at all sampling points within the analysis unit ÷ Number of sampling points; the peak value is the maximum absolute value among all voltage values ​​at sampling points within the analysis unit, obtained through a traversal and filtering process. The power spectrum characteristics of a specific frequency band are calculated as follows: First, a Fast Fourier Transform is performed on the digital EEG signal within the analysis unit to convert the time-domain signal into a frequency-domain signal, obtaining the power spectrum distribution across the entire frequency band. Then, based on the physiological characteristics of the EEG signal, four characteristic frequency bands (δ, θ, α, β) related to the user's intent are selected, and the integral value of the power spectrum within each frequency band is calculated. That is, the power spectrum characteristic value of a certain frequency band = the sum of voltage values ​​at all sampling points within that frequency band. The sum of power spectral values ​​corresponding to each frequency point yields the power spectral characteristic parameters for the four frequency bands. Time-frequency distribution characteristic calculation: A short-time Fourier transform is performed on the digital EEG signal within the analysis unit to obtain the joint distribution characteristics of the signal in the time and frequency domains, i.e., the time-frequency distribution map. Two core features are extracted from this distribution map: the mean amplitude of the time-frequency distribution, which is the sum of the amplitudes corresponding to all time-frequency points in the distribution map divided by the total number of points; and the maximum amplitude of the time-frequency distribution, obtained by traversing the distribution map and filtering, serves as the time-frequency distribution characteristic parameter. Nonlinear dynamic characteristic calculation: The Hjorth parameter, adapted to the computing capabilities of wearable devices, is extracted. As core nonlinear dynamic characteristics, the following parameters are used: Activity = variance of digital EEG signal within the analysis unit, directly using the previously calculated time-domain variance value; Mobility = standard deviation of first-order difference of digital EEG signal ÷ standard deviation of original signal, firstly perform first-order difference processing on the original signal, calculate the standard deviation of the difference signal, and then divide it with the standard deviation of the original signal; Complexity = standard deviation of second-order difference of digital EEG signal ÷ standard deviation of first-order difference, perform difference processing again on the first-order difference signal to obtain the second-order difference signal, calculate its standard deviation, and then divide it with the standard deviation of the first-order difference signal to finally obtain the three nonlinear dynamic characteristic parameters.

[0038] Multiple feature parameters are combined and normalized to obtain a high-dimensional feature vector. Specifically, this involves combining the extracted time-domain statistical features, specific frequency band power spectrum features, time-frequency distribution features, and nonlinear dynamic features in a predetermined fixed order to form an original high-dimensional feature set containing 13 feature dimensions, i.e., the original feature vector. This vector retains all-dimensional features of the EEG signal related to the user's intent. Subsequently, the original feature vector is normalized using a min-max normalization algorithm to eliminate the dimensional differences between different feature parameters, avoiding the impact of a single feature's excessively large numerical range on the accuracy of subsequent intent recognition. First, the feature parameters of the same dimension of all analysis units collected by the warning cap are traversed, and the global maximum and global minimum values ​​of each feature dimension are determined. Then, each feature parameter in the original feature vector is calculated independently. The normalization calculation formula is: normalized feature value = original feature value - global minimum value of the dimension ÷ global maximum value of the dimension - global minimum value of the dimension. After calculation, the value of each feature dimension is mapped to a unified range of 0-1. After the normalization of all feature parameters is completed, the 13 normalized feature parameters are combined in the original order to finally obtain a standardized, dimensionless high-dimensional feature vector. This feature vector will be directly sent to the intent recognition module of the embedded microprocessor as the core input data for intent pattern matching and classification.

[0039] In this embodiment of the invention, the digital EEG signal is continuously segmented by a sliding window with reasonable parameters. Multi-dimensional feature parameters are extracted from the time domain, frequency domain, time-frequency domain, and nonlinear dynamics. Then, the feature parameters are standardized and combined through min-max normalization to generate a high-dimensional feature vector. This effectively overcomes the technical problems of existing brain-computer interface devices, such as the single feature extraction dimension, inability to fully capture the feature information related to the user's intention in the EEG signal, the difference in the dimensions of the feature parameters, the serious impact of direct combination on the accuracy of subsequent intention classification, the lack of standardized design of the feature extraction process, and the high computational complexity that makes it difficult to adapt to the low-power embedded processors of wearable devices. This implementation process preserves the temporal characteristics and integrity of EEG signals through sliding window segmentation, and comprehensively captures EEG features related to user intent through multi-dimensional feature extraction, allowing the feature vector to fully reflect the electrical signal changes corresponding to brain activity. Normalization eliminates dimensional interference, and the generated standardized high-dimensional feature vectors perfectly match the classification requirements of the intent recognition algorithm model, improving the accuracy and robustness of intent recognition. At the same time, the computational steps of the entire feature extraction process have been simplified and standardized, with computational load adapted to the operating capabilities of low-power embedded microprocessors, and can be seamlessly integrated into the signal processing and control module of the warning cap. It balances the comprehensiveness of feature extraction with the real-time performance and portability of the device, laying a high-quality feature data foundation for personalized intent recognition for different disabled patients.

[0040] In a preferred embodiment of the present invention, the feature vector is input into a pre-trained intent recognition algorithm model for pattern matching and classification to obtain a category instruction representing a specific user intent, including: The intention recognition algorithm model is loaded into the embedded microprocessor and runs. The intention recognition algorithm model is a classifier built based on pattern recognition algorithm. It internally stores classification decision parameters learned from a large number of samples. After loading, the model has the ability to classify the current input feature vector. Specifically, the embedded microprocessor has a pre-embedded and stored intention recognition algorithm model trained on a large number of EEG sample data. This model is a dedicated classifier built based on pattern recognition algorithm. It internally stores classification decision parameters learned from a large number of EEG samples of different types and different groups of people. These parameters include core data such as feature benchmark thresholds, feature weight matrices, and classification boundary values ​​corresponding to various intention patterns. It can accurately match the EEG feature patterns corresponding to basic intentions such as pain, seeking help, using the toilet, confirmation, and cancellation. After the standardized high-dimensional feature vector is transmitted to the intent recognition module, the system parses the feature vector according to the model's preset data format, extracts feature parameters dimension by dimension, and loads them into the input layer of the intent recognition algorithm model, completing the full loading of the feature vector. After loading, the algorithm model is activated immediately, and the classifier calls all the internally stored classification decision parameters to build a matching calculation link between the feature vector and the preset intent pattern. At this time, the model has all the ability to perform intent classification calculation on the currently input feature vector and enters the matching calculation state.

[0041] Based on the loaded intent recognition algorithm model, pattern matching calculations are performed on the feature vectors. The similarity between various preset intent patterns is calculated, including EEG patterns associated with steady-state visual evoked potentials (SVPs), event-related potentials (P300), motor imagery, or specific cognitive tasks. This yields a set of similarity values ​​reflecting the degree of matching between the feature vectors and various intent patterns. Specifically, after the algorithm model is activated, pattern matching calculations are immediately performed on the loaded feature vectors. The core of this calculation is to determine the similarity between the feature vector and all preset intent patterns within the model. All preset intent patterns are associated with steady-state visual evoked potentials (SVPs). The model employs various EEG signal paradigms, including potentials, event-related potentials (P300), motor imagery, and specific cognitive tasks. Each intention pattern corresponds to a unique basic intention type. The model retrieves standard feature vectors from its internal storage for each intention pattern. These standard feature vectors are benchmark vectors representing the core features of the intention, obtained through extensive training on a large number of samples, and have the same dimension as the input high-dimensional feature vector. A cosine similarity algorithm is then used for matching calculations. This algorithm accurately measures the similarity between two vectors of the same dimension. The calculation process involves calculating the dot product between the input feature vector and the standard feature vector of a certain intention pattern, i.e., the dot product value = ... The input feature vector's first-dimensional parameter × the standard feature vector's first-dimensional parameter + the input feature vector's second-dimensional parameter × the standard feature vector's second-dimensional parameter + ... + the input feature vector's Nth-dimensional parameter × the standard feature vector's Nth-dimensional parameter; calculate the magnitude of the input feature vector and the magnitude of the standard feature vector of the intent pattern, respectively. The vector magnitude is the square root of the sum of the squares of the first-dimensional parameter, the squares of the second-dimensional parameter, ..., the squares of the Nth-dimensional parameter; the third step is to calculate the cosine similarity value between the input feature vector and the intent pattern. The cosine similarity value = dot product ÷ (input feature vector magnitude × standard feature vector magnitude), and this value ranges from -1. The closer the value is to 1, the higher the similarity between the two. According to the calculation process, the cosine similarity is calculated between the input feature vector and the standard feature vector of all preset intention patterns in the model. After all calculations are completed, a set of similarity values ​​corresponding one-to-one with the preset intention patterns are obtained. This set of values ​​fully reflects the matching degree between the current input feature vector and various preset intention patterns.

[0042] Based on the similarity value, the intent category with the highest similarity is selected as the current recognition result, resulting in a category instruction output representing the specific user intent. Specifically, after completing all similarity calculations, the intent recognition module performs a global traversal and numerical comparison of the obtained set of similarity values, filtering out the largest similarity value. The preset intent mode corresponding to this maximum value is the intent type with the highest matching degree of the current feature vector, and it is determined as the final result of this intent recognition. The intent recognition module retrieves the internal preset intent-instruction mapping table, which configures a unique numerical code or character identifier for each basic intent type as a category instruction. For example, pain or emergency corresponds to instruction 001, needing help corresponds to instruction 002, needing to use the toilet corresponds to instruction 003, confirmation corresponds to instruction 004, cancellation corresponds to instruction 005, etc. Based on the determined final intent type, the module retrieves the corresponding unique category instruction from the mapping table, completes the generation and output of the instruction, and this category instruction will be directly sent to the subsequent control drive module as the core basis for driving the visualization display.

[0043] In this embodiment of the invention, model activation is achieved by loading standardized feature vectors into a pre-trained pattern recognition classifier. A cosine similarity algorithm is used to accurately calculate the matching degree between the feature vectors and various preset intent patterns. Then, the intent type is determined by comparing similarity values, and a unique category instruction is generated. This effectively overcomes the technical problems of existing brain-computer interface device intent recognition algorithms, such as poor model-feature vector compatibility, low recognition accuracy due to a single pattern matching calculation method, lack of clear intent classification criteria leading to misjudgments, and the lack of standardized instruction forms for the generated intent information, making it difficult to interface with the display driver module. It also solves the problems of high computational complexity, slow response speed, and inability to adapt to low-power embedded microprocessors in wearable devices using traditional intent recognition algorithms. This implementation relies on a pre-trained algorithm model and standardized... The matching calculation method improves the accuracy and robustness of intent recognition and reduces the false positive rate. Intent classification is completed through explicit similarity value comparison rules, making the recognition results more objective and standardized. The generated standardized category commands can seamlessly interface with the control drive module, ensuring the efficiency and accuracy of intent information transmission and improving the overall response speed from feature extraction to intent command generation. A complete intent recognition can be completed within 2-4 seconds, meeting the real-time response needs in daily care. Simultaneously, the algorithm for the entire recognition process is lightweight, with computational load adapted to the operating capabilities of embedded microprocessors, allowing seamless integration into the signal processing and control module of the warning cap. This balances recognition accuracy with device portability and real-time performance, laying a precise and efficient command foundation for visual warning of intent.

[0044] In a preferred embodiment of the present invention, the intent category instruction is parsed into a corresponding display control command according to a preset mapping relationship, including: The intent-display mapping table, pre-stored in the microprocessor, is queried. This table defines the correspondence between each intent category instruction and its corresponding display effect parameters, including color, brightness, blink frequency, pattern type, and text content. Specifically, the embedded microprocessor's storage unit pre-stores a standardized intent-display mapping table. This structured table establishes a one-to-one correspondence between each unique intent category instruction and its corresponding display effect parameters. The mapping table is designed entirely based on the basic intent expression needs of disabled patients and the visual recognition needs of caregivers. The intent category instructions are standardized instructions in digital encoding form, while the display effect parameters include all parameters driving the display modules, such as color, brightness, blink frequency, pattern type, and text content. The core parameters of the operation are as follows: each intent category command corresponds to a complete and unique set of display effect parameters. For example, the command code for pain or emergency corresponds to red, 100% brightness, and a flashing frequency of 2 times or per second; the command code for needing help corresponds to yellow, 100% brightness, and a flashing frequency of 1 time or per second; the command code for needing to use the toilet corresponds to blue, 100% brightness, and a constant light mode; the command code for yes or confirmation corresponds to green, 80% brightness, and a constant light mode for 2 seconds; and the command code for no intent corresponds to white, 10% brightness, and a constant light mode. When the intent category command is transmitted to the control driver module, the module immediately retrieves the intent-display mapping table in the storage unit and loads it into the calculation cache area to complete the preliminary preparation for command and parameter matching, ensuring that the mapping table can be retrieved and called in real time.

[0045] Based on the intent category instruction, the corresponding display effect parameters are retrieved from the mapping table, and a display control command containing the parameters is generated. Specifically, the control driver module performs a precise search on the intent-display mapping table loaded in the cache, using the input intent category instruction as the search keyword, and performs a global match in the instruction column of the mapping table to locate the parameter row corresponding to the instruction. Then, all display effect parameters in that row are extracted to form a complete set of display effect parameters that match the current user intent. After the parameters are extracted, the control driver module integrates the parameter set with basic information such as instruction identifier and execution identifier according to the preset command format to generate a display control command. The instruction identifier is used to mark the intent category instruction corresponding to the control command, and the execution identifier is used to trigger the execution action of the display module. The display control command is a structured numerical instruction that fully contains all the parameter information and execution instructions required by the display module to achieve a specific visualization effect. For example, after retrieving the intent category instruction of pain or emergency, the parameter set of red, 100% brightness, and 2 flashes per second or second is extracted. The integrated display control command contains the parameter set and the execution identifier, clearly instructing the display module to achieve the corresponding visual effect according to these parameters.

[0046] The display control commands are encapsulated according to the communication protocol required by the display module and sent to the display driver circuit. Specifically, the control driver module retrieves a pre-stored communication protocol in the embedded microprocessor that is compatible with the warning cap intention visualization display module. This communication protocol is a standardized data transmission protocol that the display driver circuit can recognize, containing core conventions such as data frame format, verification rules, and transmission rate, ensuring that the encapsulated control commands can be accurately parsed by the display driver circuit. Subsequently, the module encapsulates the generated display control commands into data frames according to the requirements of this communication protocol, displaying the core data of the display control commands in the format of frame header, frame body, and frame trailer specified by the protocol. The data is split and reassembled. The frame header is a communication start identifier used to notify the display driver circuit to receive data. The frame body contains the core parameters and instruction information of the display control command. The frame tail is a checksum, which is obtained by calculating the sum of all data in the frame body, i.e., checksum = the first data bit in the frame body + the second data bit + ... + the Nth data bit. It is used to verify whether errors have occurred during data transmission. After encapsulation, the control driver module sends the encapsulated display control command data packet to the display driver circuit of the intent visualization display module through a dedicated electrical connection channel at the transmission rate preset by the protocol, ensuring that the command can be transmitted to the execution end without errors and in real time.

[0047] In this embodiment of the invention, the precise matching of instructions and display effect parameters is achieved by retrieving a pre-stored standardized intent-display mapping table. Based on the retrieved parameter set, a complete display control command is generated. The command is then encapsulated according to the adapted communication protocol and sent to the display driver circuit. This effectively overcomes the technical problems of existing brain-computer interface devices, such as the lack of fixed mapping rules for the conversion of intent instructions and display effects, the susceptibility to parameter matching errors, the lack of standardized format for control commands leading to inaccurate parsing by the display driver circuit, and the lack of a verification mechanism for command transmission, which can easily result in data loss or errors. Ultimately, this leads to a discrepancy between the visual display and the user's true intent, and nursing staff's inability to accurately identify the patient's needs. At the same time, it solves the technical pain points of traditional instruction parsing processes being cumbersome, slow in response speed, and unable to adapt to the real-time requirements of wearable devices.

[0048] This implementation relies on a standardized intent-display mapping table to achieve precise and unique matching between intent category commands and display effect parameters, fundamentally avoiding display deviations caused by parameter matching errors. The generated structured display control commands and the adapted communication protocol encapsulation ensure that the display driver circuit can quickly and accurately parse the command content, while the checksum setting guarantees the integrity and accuracy of command transmission. The entire parsing, generation, encapsulation, and transmission process is completed rapidly by the embedded microprocessor with extremely short response time, seamlessly connecting with the previous intent recognition stage, realizing real-time conversion from intent category commands to display control commands, laying a solid foundation for the immediacy and accuracy of visualization display. At the same time, the modular and standardized design of the entire process can be perfectly integrated with the signal processing and control module of the warning cap, supporting remote updates of the intent-display mapping table via wireless communication, realizing flexible adjustment of display effect parameters, and improving the adaptability and scalability of the device.

[0049] In a preferred embodiment of the present invention, a display module integrated externally to the warning cap is driven by a display control command to generate a visual light signal or pattern signal corresponding to the intent, including: Step 6.4: The display module's driving circuit receives the display control command and parses the data content in the command to extract display effect parameters including color, brightness, flicker frequency, or pattern type. Specifically, the display module's driving circuit is always in a real-time receiving state, receiving the encapsulated display control command data packet sent by the control driving module through a dedicated electrical connection channel. After receiving the data packet, the driving circuit first decapsulates the data packet according to the preset communication protocol, sequentially identifying and extracting the frame header, frame body, and frame tail checksum in the data packet. It first calculates the sum of all data in the frame body and verifies the consistency of the checksum to confirm that there is no loss or error in the data transmission process. Then, it extracts the core data content of the display control command from the frame body. Subsequently, the driving circuit parses the core data content field by field, accurately separating all display effect parameters such as color, brightness, flicker frequency, and pattern type according to the preset command data format. Each parameter is classified and stored according to type, forming a set of display effect parameters that can be directly called, ensuring that each parameter can be accurately identified and read by the execution unit of the driving circuit, thus preparing the parameters for driving the display component.

[0050] Step 6.5 uses the display effect parameters as the control basis. The driving circuit generates corresponding electrical signals based on the parameters to control the LED array, LED strip, or miniature display screen integrated on the outside of the cap to display according to the specified color, brightness, blinking mode, or pattern. Specifically, after completing the analysis of the display effect parameters, the driving circuit uses the categorized and stored set of display effect parameters as the core control basis. Through the internal signal conversion unit, it converts various digital display effect parameters into corresponding analog or digital electrical signals one by one. Different types of display effect parameters correspond to different electrical signal generation logics. Color parameter conversion: If the display component is an RGB LED array or LED strip, the driving circuit generates corresponding duty cycle electrical signals for the red, green, and blue light emission channels respectively based on the extracted color parameters. For example... Red corresponds to a high level in the red channel and a low level in the green-blue channel; blue corresponds to a high level in the blue channel and a low level in the red-green channel. Mixed colors generate electrical signals of different amplitudes for the corresponding channels according to the color ratio. If the display component is a micro-display, the color parameters will be converted into color driving electrical signals for the display pixels, defining the RGB color values ​​of the pixels. Brightness parameter conversion: The driving circuit calculates and generates the corresponding pulse width modulation electrical signal based on the extracted brightness percentage parameter. The duty cycle of the brightness electrical signal is positively correlated with the brightness percentage. The calculation formula is: Pulse width modulation electrical signal duty cycle = brightness percentage ÷ 100%. For example, when the brightness is 80%, the duty cycle of the pulse width modulation electrical signal is 0.8, and when the brightness is 100%, the duty cycle is 1. By adjusting the duty cycle of the electrical signal, the luminous power of the display component is controlled, achieving precise brightness adjustment.

[0051] Flashing frequency parameter conversion: The driving circuit calculates and generates corresponding periodic high and low level switching electrical signals based on the extracted flashing frequency parameters. The formula for calculating the flashing period is flashing period = 1 ÷ flashing frequency. For example, when the flashing frequency is 2 times per second, the flashing period is 0.5 seconds, and the driving circuit will generate periodic switching electrical signals with high and low levels lasting for 0.25 seconds each. If it is a constant-on mode, a continuous high-level electrical signal is generated without level switching. Pattern type parameter conversion: If the display component is a micro display screen, the driving circuit retrieves the pixel driving data of the corresponding pattern pre-stored internally based on the extracted pattern type parameters, and converts it into switching and color driving electrical signals for each pixel of the display screen, defining the display position and pixel characteristics of the pattern. After integrating all the generated electrical signals, the driving circuit transmits them to the LED array, LED strip, or micro display screen on the outside of the cap body through a dedicated driving channel. The electrical signals drive the light-emitting unit or pixel unit of the display component to work according to the parameter requirements, realizing the light display of the specified color, brightness, and flashing mode or the static or dynamic display of the specified pattern, completing the conversion from electrical signals to visual signals.

[0052] Step 6.6: Based on the display effect, the display module continues to present the visual signal, maintaining a visual prompt state corresponding to the user's current intention, until a new intention category instruction is recognized and a new display control command is generated, driving the display module to update the display content. Specifically, under the action of the electrical signal output by the drive circuit, the display component continuously presents the corresponding visual light signal or pattern signal according to the set display effect parameters, entering a visual prompt state matching the user's current intention. This state will be maintained until the intention recognition module recognizes the user's new intention and generates a new intention category instruction. After being parsed and encapsulated by the control drive module, the new display control command is sent to the drive circuit. After receiving the new display control command, the drive circuit immediately repeats the above operation, parsing the new display effect parameters and generating a corresponding new electrical signal. The new electrical signal will replace the original electrical signal, driving the display component to immediately switch the display effect, update the visual signal content, present a visual prompt corresponding to the user's new intention, and simultaneously terminate the display of the original visual signal, forming a closed-loop control where the visual signal is updated in real time according to the user's intention. If no new user intention is recognized, the display component will always continue to display according to the initial electrical signal requirements, ensuring that the nursing staff can continuously recognize the user's current intention state.

[0053] In this embodiment of the invention, the display control command is precisely decapsulated and parameter parsed through the display driving circuit. The display effect parameters are converted into suitable electrical signals and driven to drive the display component to achieve the specified visualization effect. At the same time, the visualization signal is continuously presented until a new intention command triggers an update. This effectively overcomes the technical problems of existing brain-computer interface devices, such as error-prone command parsing, lack of standard logic for parameter and electrical signal conversion leading to display effect deviation, slow display driving response, inability of visualization signals to match user intentions in real time, and lack of continuous display and automatic update mechanism, which prevents nursing staff from continuously obtaining patient intention information. It also solves the technical pain points of poor compatibility between traditional display modules and brain-computer interface systems, single display effect, and inability to meet the needs of intuitive recognition at long distances and without professional background.

[0054] This implementation process ensures accurate extraction of display parameters through protocol decomposition and precise parameter parsing, preventing display deviations from the outset. Standardized parameter-to-electrical signal conversion logic enables precise control of display effects such as color, brightness, and flicker frequency, ensuring a high degree of match between the visual signal and the user's intent. The real-time driving and continuous display mechanism of the display components guarantees that caregivers can intuitively and clearly identify patient intents from a distance without the aid of professional instruments, and can continuously monitor the patient's current status. The automatic update mechanism triggered by new intent commands achieves real-time synchronization between the visual signal and the user's intent, improving the timeliness and accuracy of intent-based warnings. Simultaneously, the display module uses low-power LED arrays, light strips, or micro-displays, perfectly compatible with the wearable design of the warning cap, balancing the eye-catching display effect with the low-power requirements of the device. This further enhances the practicality and wearability of the warning cap, allowing caregivers to quickly, accurately, and continuously perceive the basic intents of disabled patients, truly establishing an efficient and silent communication channel between patients and caregivers.

[0055] Embodiments of the present invention also provide a computing device, including: a processor and a memory storing a computer program, wherein the computer program, when executed by the processor, performs the actions described above. All implementations in the above embodiments are applicable to this embodiment and can achieve the same technical effects.

[0056] Embodiments of the present invention also provide a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the actions described above. All implementations in the above embodiments are applicable to this embodiment and can achieve the same technical effects.

[0057] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A brain-computer interface-based intention visualization warning cap, characterized in that, include: The signal acquisition module is used to non-invasively acquire raw electroencephalogram (EEG) signals from the user's scalp through multiple dry or semi-dry electrodes arranged in preset positions on the inner lining of the warning cap. The signal preprocessing module is used to amplify the original EEG signal to obtain the amplified original EEG signal, and then filter the amplified original EEG signal to obtain the preprocessed analog EEG signal. The signal conversion module is used to convert preprocessed analog EEG signals into digital EEG signals. The feature extraction module is used to perform calculations and analysis on digital EEG signals, extract time-domain, frequency-domain, or time-frequency features related to user intent, and generate feature vectors for classification. The intent recognition module is used to input feature vectors into a pre-trained intent recognition algorithm model for pattern matching and classification, so as to obtain a category instruction that represents the specific user intent. The control drive module is used to parse the intent category instruction into the corresponding display control command according to the preset mapping relationship; and drive the display module integrated on the external side of the warning cap to generate a visual light signal or pattern signal corresponding to the intent through the display control command.

2. The brain-computer interface-based intention visualization warning cap according to claim 1, characterized in that, Raw electroencephalogram (EEG) signals from the user's scalp are non-invasively collected using multiple dry or semi-dry electrodes positioned at predetermined locations within the lining of the warning cap, including: Electrodes are placed in two preset areas corresponding to the international 10-20 EEG electrode placement system, including the frontal, central, or parietal regions, to cover the main scalp locations where EEG activity related to the user's intentions is generated. After ensuring stable contact between the electrodes and the user's scalp, raw EEG signals are collected continuously and at set intervals to obtain raw data reflecting brain electrical activity in real time. During the acquisition process, the potential difference between each electrode and the reference electrode is recorded simultaneously to obtain raw EEG signals at the microvolt level.

3. The brain-computer interface-based intention visualization warning cap according to claim 2, characterized in that, The raw EEG signal is amplified to obtain an amplified raw EEG signal, and then filtered to obtain a preprocessed analog EEG signal, including: The raw EEG signal is input into an instrumentation amplifier with high input impedance and high common-mode rejection ratio for differential amplification. The weak EEG signal is amplified to a voltage range suitable for processing with an appropriate gain, thus obtaining the amplified EEG signal. The amplified EEG signal was bandpass filtered to retain the effective frequency band of the EEG signal, while filtering out noise components such as power frequency interference, electromyography artifacts and electrooculography artifacts, to obtain a preliminary filtered signal. The preliminary filtered signal is further input into a notch filter to eliminate power frequency interference at a specific frequency, so as to obtain a pure preprocessed analog EEG signal.

4. The brain-computer interface-based intention visualization warning cap according to claim 3, characterized in that, Converting preprocessed analog EEG signals into digital EEG signals includes: The preprocessed analog EEG signal is input into the analog-to-digital converter unit and sampled using a sufficiently high sampling rate that meets the requirements of the EEG signal sampling theorem in order to completely preserve the time-domain information of the signal and obtain the sampled discrete-time signal. Discrete-time sampled signals are used as input data, and analog-to-digital converters with sufficiently high resolution are used to quantize the sampled signals, converting the continuous amplitude of the analog signal into discrete digital quantities, thereby generating discrete-time and discrete-amplitude digital EEG signals. The digital EEG signal is processed to remove DC bias, eliminating the DC bias introduced by the electrode amplifier circuit. The DC-biased digital signal is then buffered and packaged according to design requirements to form a digital EEG data stream.

5. The brain-computer interface-based intention visualization warning cap according to claim 4, characterized in that, The system performs computational analysis on digital EEG signals to extract time-domain, frequency-domain, or time-frequency features related to user intent, and generates feature vectors for classification, including: The digital EEG signal is segmented by sliding window, dividing the continuous EEG data into fixed-length analysis periods, with each period serving as a basic analysis unit; The digital EEG signals within the analysis unit are used to perform feature calculations and extract various feature parameters that reflect the user's intentions. These feature parameters include time-domain statistical features, power spectrum features of specific frequency bands, time-frequency distribution features, and nonlinear dynamic features. By combining and normalizing multiple feature parameters, a high-dimensional feature vector is obtained.

6. The brain-computer interface-based intention visualization warning cap according to claim 5, characterized in that, The feature vector is input into a pre-trained intent recognition algorithm model for pattern matching and classification, resulting in a category instruction representing a specific user intent, including: The intention recognition algorithm model is loaded into the embedded microprocessor to run the feature vector. The intention recognition algorithm model is a classifier built based on the pattern recognition algorithm. It internally stores classification decision parameters learned from a large number of samples. After loading, the model has the ability to perform classification calculations on the current input feature vector. Based on the loaded intent recognition algorithm model, pattern matching calculation is performed on the feature vectors, and the similarity between various preset intent patterns is calculated. The intent patterns include EEG patterns associated with steady-state visual evoked potentials, event-related potentials P300, motor imagery, or specific cognitive tasks, so as to obtain a set of similarity values ​​that reflect the degree of matching between the feature vectors and various intent patterns. Based on the similarity value, the intent category with the highest similarity is selected as the current recognition result, resulting in a category instruction output that represents the specific user intent.

7. The brain-computer interface-based intention visualization warning cap according to claim 6, characterized in that, Based on a preset mapping relationship, intent category instructions are parsed into corresponding display control commands, including: The intent-display map is pre-stored in the microprocessor. The map defines the correspondence between each intent category instruction and the corresponding display effect parameters, including color, brightness, blink frequency, pattern type, or text content. Based on the intent category instruction, the corresponding display effect parameters are retrieved from the mapping table, and a display control command containing the parameters is generated; The display control commands are encapsulated according to the communication protocol required by the display module and sent to the display driver circuit.

8. The brain-computer interface-based intention visualization warning cap according to claim 7, characterized in that, The warning cap's external integrated display module generates visual light or pattern signals corresponding to the intent via display control commands, including: The display module's driving circuit receives display control commands and parses the data content in the commands to extract display effect parameters, including color, brightness, flicker frequency, or pattern type. The display effect parameters are used as the control basis. The drive circuit generates corresponding electrical signals according to the parameters, and controls the LED array, LED strip or miniature display screen integrated on the outside of the cap to display according to the specified color, brightness, flashing mode or pattern. Based on the display effect, the display module continues to present the visual signal, maintaining a visual cue state corresponding to the user's current intent, until a new intent category instruction is recognized and a new display control command is generated, driving the display module to update the display content.

9. A computing device, characterized in that, include: One or more processors; A storage device for storing one or more programs that, when executed by one or more processors, cause the one or more processors to perform any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a program that, when executed by a processor, performs any one of claims 1 to 8.

Images