Intelligent nursing bed adaptive control system and method based on multi-mode electroencephalogram intention recognition

Abstract

The invention relates to the technical field of electroencephalogram control, and discloses an intelligent nursing bed self-adaptive control system and method based on multi-mode electroencephalogram intention recognition, and the system comprises a signal separation module, a state judgment module, an intention feature analysis module, a lateral feature extraction module, a hierarchical intention decision module and a control instruction generation module. Acquiring an occipital area steady-state visual evoked potential signal and a motion-related cortex potential signal based on the original electroencephalogram signal of the patient; analyzing the phase locking stability of the occipital area steady state visual evoked potential signal to judge an attention effective state; analyzing the motion-related cortex potential signal to obtain an energy change sequence; analyzing a contralateral dominating mode of the patient; performing hierarchical intention decision on the patient to obtain a preliminary motion intention of the patient; generating a final control instruction of the target equipment in combination with the physiological feedback signal of the patient; according to the invention, the efficiency of adaptive control of the intelligent nursing bed based on multi-mode electroencephalogram intention recognition can be improved.

Description

Technical Field

[0001] This invention relates to the field of brainwave control technology, and in particular to an adaptive control system and method for an intelligent nursing bed based on multimodal brainwave intention recognition. Background Technology

[0002] Existing technologies for EEG intention recognition control in intelligent nursing beds lack the ability to process and separate multi-dimensional features in parallel for the raw EEG signals from the patient's occipital region and motor cortex. This makes it difficult to effectively eliminate power frequency interference and physiological artifacts. Furthermore, they often rely on single-modality EEG signals for intention judgment, resulting in insufficient purity and comprehensiveness of signal extraction. Consequently, this affects the accuracy of attention state determination and the reliability of motor intention analysis.

[0003] Existing technologies lack systematic and precise implementation solutions for motor cortex lateralization feature extraction and hierarchical intention decision-making. They fail to deeply integrate and calculate effective attentional states, energy change sequences, and contralateral control patterns, and do not adequately incorporate real-time physiological feedback signals from patients for safety verification. This results in poor adaptability and insufficient safety of generated control commands, failing to match individual patient differences and dynamic physiological states. Therefore, improving the accuracy, adaptability, and safety of EEG intention recognition-based control in intelligent nursing beds has become an urgent problem to be solved. Summary of the Invention

[0004] This invention provides an adaptive control system and method for an intelligent nursing bed based on multimodal EEG intention recognition, in order to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, this invention provides an intelligent nursing bed adaptive control system based on multimodal EEG intention recognition, characterized in that the system includes a signal separation module, a state determination module, an intention feature parsing module, a lateralization feature extraction module, a hierarchical intention decision-making module, and a control command generation module, wherein:

[0006] The signal separation module is used to perform multi-dimensional feature parallel separation of the raw EEG signals of the patient's occipital brain region and motor cortex brain region to obtain the patient's occipital steady-state visual evoked potential signal and motor-related cortical potential signal.

[0007] The state determination module is used to perform phase-locked stability analysis on the steady-state visual evoked potential signal in the occipital region to determine the patient's effective attention state.

[0008] The intent feature parsing module is used to perform cooperative time-frequency analysis on the motion-related cortical potential signal when the attention effective state is effective, so as to obtain the energy change sequence of the motion-related cortical potential signal;

[0009] The lateralization feature extraction module is used to perform motor cortex chemometry analysis on the patient based on the spatial distribution of the energy change sequence to obtain the patient's contralateral innervation pattern.

[0010] The graded intention decision module is used to make graded intention decisions for the patient based on the effective attention state, the energy change sequence, and the contralateral dominance pattern, so as to obtain the patient's preliminary motor intention.

[0011] The control command generation module is used to generate the final control command for the target device based on the initial movement intention and the patient's real-time physiological feedback signal.

[0012] In a preferred embodiment, when the signal separation module performs multi-dimensional feature parallel separation of the raw EEG signals from the patient's occipital brain region and motor cortex brain region to obtain the patient's occipital steady-state visual evoked potential signal and motor-related cortical potential signal, it is specifically used for:

[0013] Simultaneously collect raw electroencephalogram (EEG) signals from the patient's occipital region and motor cortex.

[0014] By removing power frequency interference and physiological artifacts from the original EEG signal, a clean EEG signal is obtained from the original EEG signal.

[0015] Blind source separation was performed on the clean EEG signal to obtain the patient's independent signal source components;

[0016] Based on the scalp topology distribution map of the independent signal source components, the first type of component set of the occipital brain region and the second type of component set of the motor cortex brain region are screened out.

[0017] Based on the matching relationship between the peak power spectrum of the first type of component set and the preset visual stimulus frequency, the corresponding matching frequency signal components in the first type of component set are superimposed in the time domain to obtain the steady-state visual evoked potential signal of the patient's occipital region.

[0018] Event-related potential analysis is performed on the second type of component set, and waveform synthesis is performed on the analyzed motor preparation feature components to obtain the motor-related cortical potential signal of the patient.

[0019] In a preferred embodiment, when the state determination module performs phase-locked stability analysis on the steady-state visual evoked potential signal in the occipital region to determine the patient's effective attention state, it is specifically used for:

[0020] A sine-cosine reference signal with the same frequency as the preset visual stimulus is generated by a digital oscillator.

[0021] Calculate the instantaneous phase-locked value between the steady-state visual evoked potential signal in the pillow region and the sine-cosine reference signal;

[0022] The stability index of the instantaneous phase-locked value is obtained by performing in-window standard deviation analysis on the instantaneous phase-locked value.

[0023] The stability index is compared with a preset first stability threshold.

[0024] When the stability index continuously exceeds the first stability threshold, the patient's effective attention state is determined to be effective;

[0025] When the stability index does not continuously exceed the first stability threshold, the patient's effective attention state is determined to be invalid.

[0026] In a preferred embodiment, the instantaneous phase-locked value is calculated using the following formula:

[0027] ;

[0028] In the formula, Indicates at time The instantaneous phase-locked value, Represents the imaginary unit. This represents the number of adjacent time points used for local estimation. Indicates at time No. The instantaneous phase of the steady-state visual evoked potential signal in the pillow region at each adjacent time point. Indicates at time The instantaneous phase of the sine-cosine reference signal, Represents a complex exponential function. This represents the modulo operation for complex numbers.

[0029] In a preferred embodiment, when the intent feature parsing module performs cooperative time-frequency analysis on the motion-related cortical potential signal to obtain the energy change sequence of the motion-related cortical potential signal when the attention effective state is effective, it is specifically used for:

[0030] Based on the event-related characteristics of the motion-related cortical potential signal, a time analysis window for the motion-related cortical potential signal is established with the moment when the effective attention state is received as the effective marker as the center.

[0031] Within the time analysis window, continuous wavelet transform is performed on the motion-related cortical potential signal to obtain the time-frequency energy distribution of the motion-related cortical potential signal;

[0032] The energy data of a specific frequency in the time-frequency energy distribution are sorted by spectral integration to obtain the energy change sequence of the motion-related cortical potential signal.

[0033] In a preferred embodiment, when the lateralization feature extraction module performs motor cortical biochemical analysis on the patient based on the spatial distribution of the energy change sequence to obtain the patient's contralateral innervation pattern, it is specifically used for:

[0034] Based on the electrode spatial location information of the energy change sequence, a first set of electrode channels covering the left motor cortex region of the patient and a second set of electrode channels covering the right motor cortex region of the patient are defined.

[0035] Extract the energy change sequences corresponding to the first group of electrode channels and the energy change sequences corresponding to the second group of electrode channels, respectively.

[0036] The difference between the average energy of the first group of electrode channels and the second group of electrode channels is quantified to obtain the lateralization energy difference of the patient.

[0037] The sign and amplitude of the lateralization energy difference are coupled and analyzed to obtain the contralateral control pattern of the patient.

[0038] In a preferred embodiment, when the graded intention decision module performs graded intention decision-making on the patient based on the effective attention state, the energy change sequence, and the contralateral dominance pattern to obtain the patient's preliminary motor intention, it is specifically used for:

[0039] The stability of the effective attention state is assessed to obtain the patient's attention weighting factor.

[0040] The normalized amplitude characteristics of the energy change sequence are analyzed statistically to obtain the patient's energy confidence factor.

[0041] A significant quantitative analysis of the contralateral control pattern was performed to obtain the spatial lateralization factor of the patient.

[0042] The patient's overall intent credibility is calculated based on the attention weighting factor, the energy confidence factor, and the spatial lateralization factor.

[0043] Based on the comprehensive intent credibility, parametric motion mapping is performed on the contralateral control pattern to obtain the patient's preliminary motion intent.

[0044] In a preferred embodiment, the formula for calculating the credibility of the comprehensive intent is as follows:

[0045] ;

[0046] In the formula, This indicates the credibility of the overall intent. This represents the attention weight factor. This represents the energy confidence factor. This represents the spatial lateralization factor. This represents the preset regularization coefficient. This represents a pre-defined, extremely small positive constant. Represents the hyperbolic tangent function. This represents the arctangent function.

[0047] In a preferred embodiment, when the control command generation module generates the final control command for the target device based on the initial movement intention and the patient's real-time physiological feedback signal, it is specifically used for:

[0048] Receive the initial movement intention and the patient's real-time physiological feedback signals collected by the target device's sensors;

[0049] Physiological state features are extracted from the real-time physiological feedback signal to obtain the patient's real-time body pressure distribution information and heart rate variability index;

[0050] Based on the real-time body pressure distribution information, a biomechanical analysis is performed on the patient's current posture to obtain the patient's stability margin estimation data;

[0051] Based on the stability margin estimation data and the heart rate variability index, the safety of the preliminary exercise intention is verified.

[0052] Based on the verified preliminary motion intent, the drive parameters of the target device are encoded to obtain the final control command of the target device.

[0053] To address the aforementioned problems, this invention also provides an adaptive control method for an intelligent nursing bed based on multimodal EEG intention recognition, the method comprising:

[0054] S1. Perform multi-dimensional feature parallel separation on the raw EEG signals of the patient's occipital brain region and motor cortex brain region to obtain the patient's occipital steady-state visual evoked potential signal and motor-related cortical potential signal.

[0055] S2. Perform phase-locked stability analysis on the steady-state visual evoked potential signal in the occipital region to determine the patient's effective attention state;

[0056] S3. When the attention effective state is effective, perform cooperative time-frequency analysis on the motion-related cortical potential signal to obtain the energy change sequence of the motion-related cortical potential signal;

[0057] S4. Based on the spatial distribution of the energy change sequence, perform motor cortex biochemical analysis on the patient to obtain the patient's contralateral innervation pattern;

[0058] S5. Based on the effective attention state, the energy change sequence, and the contralateral dominance pattern, perform graded intention decision-making on the patient to obtain the patient's preliminary motor intention;

[0059] S6. Based on the initial movement intention and the patient's real-time physiological feedback signal, generate the final control command for the target device.

[0060] Compared with the prior art, the present invention has the following beneficial effects:

[0061] 1. This invention separates the original EEG signal in parallel using multi-dimensional features, effectively eliminating power frequency interference and physiological artifacts, accurately screening dual-modal signal components and completing signal synthesis, combining phase-locked stability analysis to determine attention state, and extracting precise energy change sequences through collaborative time-frequency analysis, significantly improving the purity of EEG signal processing and the accuracy of intention feature extraction.

[0062] 2. This invention calculates the credibility of comprehensive intent by integrating attention, energy, and spatial multi-dimensional factors, achieving hierarchical intent decision-making and precise motion mapping. It then combines real-time physiological feedback signals to conduct biomechanical analysis and safety verification, generating control commands adapted to the patient's dynamic state. This significantly improves the adaptability, reliability, and safety of nursing bed control, and enhances the convenience for patients. Attached Figure Description

[0063] Figure 1 This is a system architecture diagram of an intelligent nursing bed adaptive control system based on multimodal EEG intention recognition, provided in an embodiment of the present invention.

[0064] Figure 2 This is a flowchart illustrating an adaptive control method for an intelligent nursing bed based on multimodal EEG intention recognition, provided as an embodiment of the present invention.

[0065] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0066] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments belong to some, but not all, embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0067] The terminology used in the embodiments of this invention is for the purpose of describing particular embodiments only and is not intended to limit the invention. The singular forms “said” and “the” as used in the embodiments of this invention and the appended claims are also intended to include the plural forms, and “multiple” generally includes at least two unless the context clearly indicates otherwise.

[0068] Depending on the context, the word "if" or "if" as used here can be interpreted as "when," "when," "in response to determination," or "in response to detection." Similarly, depending on the context, the phrase "if determination" or "if detection (of the stated condition or event)" can be interpreted as "when determination," "in response to determination," "when detection (of the stated condition or event)," or "in response to detection (of the stated condition or event)."

[0069] Furthermore, the timing of the steps in the following method embodiments is merely an example and not a strict limitation.

[0070] In practice, the server-side equipment deployed in an intelligent nursing bed adaptive control system based on multimodal EEG intention recognition may consist of one or more devices. This intelligent nursing bed adaptive control system based on multimodal EEG intention recognition can be implemented as a business instance, a virtual machine, or hardware devices. For example, this intelligent nursing bed adaptive control system based on multimodal EEG intention recognition can be implemented as a business instance deployed on one or more devices in a cloud node. Simply put, this intelligent nursing bed adaptive control system based on multimodal EEG intention recognition can be understood as software deployed on a cloud node, used to provide each user terminal with an intelligent nursing bed adaptive control system based on multimodal EEG intention recognition. Alternatively, this intelligent nursing bed adaptive control system based on multimodal EEG intention recognition can also be implemented as a virtual machine deployed on one or more devices in a cloud node. This virtual machine contains application software for managing each user terminal. Alternatively, this intelligent nursing bed adaptive control system based on multimodal EEG intention recognition can also be implemented as a server consisting of numerous identical or different types of hardware devices, with one or more hardware devices set up to provide each user terminal with an intelligent nursing bed adaptive control system based on multimodal EEG intention recognition.

[0071] In terms of implementation, an adaptive control system for an intelligent nursing bed based on multimodal EEG intention recognition and the user terminal are mutually adapted. Specifically, if the adaptive control system is implemented as an application installed on a cloud service platform, the user terminal acts as a client establishing a communication connection with that application; or if the system is implemented as a website, the user terminal acts as a webpage; or if the system is implemented as a cloud service platform, the user terminal acts as a mini-program within an instant messaging application.

[0072] like Figure 1 The diagram shown is a system architecture diagram of an intelligent nursing bed adaptive control system based on multimodal EEG intention recognition, provided by an embodiment of the present invention.

[0073] The intelligent nursing bed adaptive control system 100 based on multimodal EEG intention recognition described in this invention can be set up in a cloud server. In terms of implementation, it can be used as one or more service devices, or as an application installed in the cloud (e.g., a mobile service operator's server, server cluster, etc.), or it can be developed into a website. Depending on the functions implemented, the intelligent nursing bed adaptive control system 100 based on multimodal EEG intention recognition may include a signal separation module 101, a state determination module 102, an intention feature parsing module 103, a side-channel feature extraction module 104, a hierarchical intention decision module 105, and a control command generation module 106. The module described in this invention can also be called a unit, which refers to a series of computer program segments that can be executed by the processor of an electronic device and can perform a fixed function, stored in the memory of the electronic device.

[0074] In this embodiment of the invention, an adaptive control system for an intelligent nursing bed based on multimodal EEG intention recognition allows each of the aforementioned modules to be implemented independently and to call upon other modules. This "calling" can be understood as a module connecting to multiple modules of another type and providing corresponding services to those connected modules. This embodiment of the invention provides an adaptive control system for an intelligent nursing bed based on multimodal EEG intention recognition. Without modifying the program code, the applicable scope of the adaptive control system architecture can be adjusted by adding modules and directly calling them, achieving cluster-based horizontal expansion. This allows for quick and flexible expansion of the adaptive control system. In practical applications, the aforementioned modules can be set in the same device or different devices, or they can be set in a virtual device, such as a service instance in a cloud server.

[0075] The following describes, with reference to specific embodiments, the various components and specific workflows of an adaptive control system for an intelligent nursing bed based on multimodal EEG intention recognition:

[0076] The signal separation module 101 is used to perform multi-dimensional feature parallel separation on the original EEG signals of the patient's occipital brain region and motor cortex brain region to obtain the patient's occipital steady-state visual evoked potential signal and motor-related cortical potential signal.

[0077] In this embodiment of the invention, when the signal separation module performs multi-dimensional feature parallel separation of the raw EEG signals from the patient's occipital region and motor cortex to obtain the patient's occipital region steady-state visual evoked potential signal and motor-related cortical potential signal, it is specifically used for:

[0078] Simultaneously collect raw electroencephalogram (EEG) signals from the patient's occipital region and motor cortex.

[0079] By removing power frequency interference and physiological artifacts from the original EEG signal, a clean EEG signal is obtained from the original EEG signal.

[0080] Blind source separation was performed on the clean EEG signal to obtain the patient's independent signal source components;

[0081] Based on the scalp topology distribution map of the independent signal source components, the first type of component set of the occipital brain region and the second type of component set of the motor cortex brain region are screened out.

[0082] Based on the matching relationship between the peak power spectrum of the first type of component set and the preset visual stimulus frequency, the corresponding matching frequency signal components in the first type of component set are superimposed in the time domain to obtain the steady-state visual evoked potential signal of the patient's occipital region.

[0083] Event-related potential analysis is performed on the second type of component set, and waveform synthesis is performed on the analyzed motor preparation feature components to obtain the motor-related cortical potential signal of the patient.

[0084] Silver chloride electrodes conforming to EEG acquisition standards were fixed in the scalp areas corresponding to the occipital region and the motor cortex, respectively. All electrodes were simultaneously connected to the same multi-channel EEG acquisition device through shielded wires. The device was pre-set with a unified sampling start trigger command. When the trigger command was issued, the electrodes of all channels started to acquire EEG signals synchronously. The acquired unprocessed electrical signals were the raw EEG signals of the occipital and motor cortex regions of the patient.

[0085] For the acquired raw EEG signals, a passive RC filter circuit was used to filter out 50Hz power frequency interference. This circuit consists of a specific frequency response network composed of resistors and capacitors, allowing only effective EEG frequency signals outside 50Hz to pass through, directly blocking power frequency interference signals. For EEG artifacts, waveform segments with amplitudes exceeding 50μV and exhibiting biphasic waveforms were identified, and their amplitudes were replaced with the average amplitude of adjacent normal signal segments. For EMG artifacts, components with frequencies above 30Hz and amplitude fluctuations exceeding 20μV were identified, and these components were processed using a moving average method with a sliding window set to 10ms. The fluctuations were smoothed by calculating the average signal within the window. After the above targeted processing, a clean EEG signal with power frequency interference and physiological artifacts removed was obtained.

[0086] The clean EEG signal is input into a preset signal decomposition process. This process is based on the inherent differences in waveform duration and amplitude variation patterns of EEG signals generated by different signal sources. The clean EEG signal is segmented into segments. First, signal segments with similar waveform characteristics are identified. Then, these segments are classified into independent signal units. Each independent signal unit corresponds to only a single physiological signal source. These independent signal units are the patient's independent signal source components.

[0087] Using EEG signal topological imaging technology, a scalp topological distribution map is drawn for each independent signal source component. This map is based on the two-dimensional coordinates of the patient's scalp, and different colors are used to indicate the energy distribution intensity of the signal components. At the same time, standard scalp range coordinates of the occipital brain region and the motor cortex brain region are pre-stored. The scalp topological distribution map of each independent signal source component is compared with the above two preset range coordinates. Independent signal source components whose energy distribution area falls within the scalp range coordinates of the occipital brain region are integrated into the first type of component set, and independent signal source components whose energy distribution area falls within the scalp range coordinates of the motor cortex brain region are integrated into the second type of component set.

[0088] Frequency spectrum analysis is performed on each signal component in the first type of component set. By scanning the frequency distribution of the signal, the power value corresponding to each frequency point is recorded to form a power spectrum curve for each signal component. The frequency point with the largest power value is extracted from the power spectrum curve as the power spectrum peak of the signal component. The preset visual stimulation frequency is set to 10Hz. The power spectrum peak of each signal component is compared with 10Hz. When the difference between the two is less than or equal to 0.5Hz, the signal component is determined to be a matched frequency signal component. All matched frequency signal components are precisely aligned in chronological order according to the time axis so that the start time and duration of each signal component are completely consistent. Then, the amplitudes of each aligned signal component at the same time point are added to obtain the composite signal after time domain superposition. This composite signal is the patient's occipital region steady-state visual evoked potential signal.

[0089] Each signal component in the second type of component set is decomposed into a time series. The preset time window related to movement preparation is from 1 second before the movement intention to 0.5 seconds after the movement intention. The waveform changes of each signal component within this time window are analyzed. The preset characteristic waveform of movement preparation is that the amplitude gradually increases from the beginning of the time window, reaches a peak at the middle of the time window, and then gradually decreases. Signal components that meet this waveform characteristic are selected as movement preparation characteristic components. All movement preparation characteristic components are precisely aligned according to the corresponding time on the time axis to ensure that the time marks of each component are completely consistent. Then, the arithmetic mean of the amplitude of all movement preparation characteristic components at each time point is calculated. The average value of each time point is used as the signal amplitude at that time to construct a continuous signal waveform, which is the patient's movement-related cortical potential signal.

[0090] The beneficial effects include ensuring the temporal synchronization of the original EEG signals from the occipital region and the motor cortex through synchronous acquisition; subsequent targeted interference and artifact removal processes accurately removed irrelevant interference components, significantly improving signal quality; blind source separation and scalp topology-based screening methods enabled accurate classification of signal components from different brain regions; quantitative matching and temporal superposition of power spectrum peak values ​​and preset visual stimulation frequencies ensured accurate extraction of steady-state visual evoked potential signals from the occipital region; and event-related potential analysis and waveform synthesis of motor preparation feature components reliably obtained motor-related cortical potential signals. The entire implementation process is specific and reproducible, effectively ensuring the purity and integrity of the two target signals, providing a high-quality and highly reliable signal foundation for further processing of various modules in the system, and ensuring the stable operation of the entire intelligent nursing bed adaptive control system.

[0091] The state determination module 102 is used to perform phase-locked stability analysis on the steady-state visual evoked potential signal in the occipital region to determine the patient's effective attention state.

[0092] In this embodiment of the invention, when the state determination module performs phase-locked stability analysis on the steady-state visual evoked potential signal in the occipital region to determine the patient's effective attention state, it is specifically used for:

[0093] A sine-cosine reference signal with the same frequency as the preset visual stimulus is generated by a digital oscillator.

[0094] Calculate the instantaneous phase-locked value between the steady-state visual evoked potential signal in the pillow region and the sine-cosine reference signal;

[0095] The stability index of the instantaneous phase-locked value is obtained by performing in-window standard deviation analysis on the instantaneous phase-locked value.

[0096] The stability index is compared with a preset first stability threshold.

[0097] When the stability index continuously exceeds the first stability threshold, the patient's effective attention state is determined to be effective;

[0098] When the stability index does not continuously exceed the first stability threshold, the patient's effective attention state is determined to be invalid.

[0099] The formula for calculating the instantaneous phase-locked value is as follows:

[0100] ;

[0101] In the formula, Indicates at time The instantaneous phase-locked value, Represents the imaginary unit. This represents the number of adjacent time points used for local estimation. Indicates at time No. The instantaneous phase of the steady-state visual evoked potential signal in the pillow region at each adjacent time point. Indicates at time The instantaneous phase of the sine-cosine reference signal, Represents a complex exponential function. This represents the modulo operation for complex numbers.

[0102] The digital oscillator stores the value of a preset visual stimulus frequency, which is set to 10 Hz. The digital oscillator generates periodically changing electrical signals through internal circuitry, forming sine wave signals and cosine wave signals. The frequencies of both signals are exactly the same as 10 Hz, and the phase difference between the sine wave signal and the cosine wave signal is 90 degrees. The combination of these two signals is the sine-cosine reference signal with the same frequency as the preset visual stimulus frequency.

[0103] Ten adjacent time points are pre-set for local estimation. At each specific time point, the instantaneous phase of the steady-state visual evoked potential signal in the pillow region corresponding to each of the ten adjacent time points is extracted one by one. At the same time, the instantaneous phase of the sine-cosine reference signal at the same time point is also extracted. For each adjacent time point, the instantaneous phase of the steady-state visual evoked potential signal in the pillow region at that time point is subtracted from the instantaneous phase of the sine-cosine reference signal to obtain the phase difference value corresponding to each adjacent time point. Each phase difference value is substituted into a complex exponential function for transformation to obtain the complex result corresponding to each adjacent time point. The complex results corresponding to the ten adjacent time points are accumulated and summed. The total value obtained by summing is divided by 10 to obtain the complex average value. Then, the modulus of the complex average value is obtained through mathematical operations. The non-negative value obtained is the instantaneous phase-locked value at that time.

[0104] The duration of the time window is set to 500 milliseconds. The instantaneous phase-locked values ​​calculated sequentially are divided into 500-millisecond time windows according to the time sequence. Each window contains several instantaneous phase-locked values. First, the arithmetic mean of all instantaneous phase-locked values ​​in each window is calculated. Then, the difference between each instantaneous phase-locked value and the average value of the window is calculated. Each difference is squared. Then, all the squared differences are summed. The sum is divided by the number of instantaneous phase-locked values ​​in the window. Finally, the square root of the result is taken. The resulting value is the stability index of the instantaneous phase-locked values ​​corresponding to the window.

[0105] The preset first stability threshold is set to 0.1. The stability index calculated for each time window is compared with 0.1. The relationship between each stability index and 0.1 is recorded one by one to determine whether each stability index is greater than 0.1 or less than or equal to 0.1.

[0106] The judgment time for the stability index to continuously exceed the first stability threshold is set at 2 seconds. The total duration corresponding to the continuous time window where the stability index is greater than 0.1 is counted. When the total duration reaches 2 seconds or more, the patient's attention is judged to be effective.

[0107] When the total duration of consecutive time windows with a stability index greater than 0.1 does not reach 2 seconds, or when there is at least one time window with a stability index less than or equal to 0.1 within any consecutive 2-second time period, the patient's effective attentional state is determined to be invalid.

[0108] The beneficial effects are as follows: the digital oscillator generates a sine-cosine reference signal based on a clearly preset visual stimulus frequency, ensuring the accuracy and consistency of the phase comparison benchmark; the calculation process of the instantaneous phase-locked value fully utilizes the phase information of multiple adjacent time points, and through a series of clear steps such as phase difference conversion, summation, averaging, and modulus taking, it accurately quantifies the phase synchronization degree between the steady-state visual evoked potential signal in the occipital region and the reference signal; the in-window standard deviation analysis, through a fixed window duration and unified calculation logic, objectively reflects the fluctuation of the instantaneous phase-locked value, and the obtained stability index is reproducible; the clear setting of the first stability threshold and the duration of continuous judgment makes the judgment criteria for the effective state of attention unique and objective, avoiding ambiguous judgments; the entire implementation process is clear in its steps and the processing methods are fully disclosed, with clear basis and specific methods for each step of the operation, ensuring the rigor of the phase-locked stability analysis, accurately and reliably judging the patient's effective state of attention, providing a precise judgment basis for the subsequent intention feature analysis module to work only when attention is effective, ensuring the pertinence and effectiveness of the subsequent processing flow of the entire system, and improving the overall accuracy of the adaptive control of the intelligent nursing bed.

[0109] The intent feature parsing module 103 is used to perform cooperative time-frequency analysis on the motion-related cortical potential signal when the attention effective state is effective, so as to obtain the energy change sequence of the motion-related cortical potential signal;

[0110] In this embodiment of the invention, when the intent feature parsing module performs cooperative time-frequency analysis on the motion-related cortical potential signal to obtain the energy change sequence of the motion-related cortical potential signal when the attention effective state is effective, it is specifically used for:

[0111] Based on the event-related characteristics of the motion-related cortical potential signal, a time analysis window for the motion-related cortical potential signal is established with the moment when the effective attention state is received as the effective marker as the center.

[0112] Within the time analysis window, continuous wavelet transform is performed on the motion-related cortical potential signal to obtain the time-frequency energy distribution of the motion-related cortical potential signal;

[0113] The energy data of a specific frequency in the time-frequency energy distribution are sorted by spectral integration to obtain the energy change sequence of the motion-related cortical potential signal.

[0114] Based on the event-related characteristics of motor-related cortical potential signals, the moment when the effective state of attention is received is taken as the midpoint of time. The total duration of the time analysis window is set to 2 seconds, of which the duration before the marked moment is 1 second and the duration after the marked moment is 1 second. By defining this time range, the analysis interval of the motor-related cortical potential signal is clarified, and the time analysis window of the motor-related cortical potential signal is formed.

[0115] Morlet wavelet is selected as the basis function for continuous wavelet transform. This basis function is composed of a sine wave and a Gaussian envelope. The basis function is convolved with the motion-related cortical potential signal within the time analysis window point by point along the time axis. During the operation, the scale parameter of the basis function is kept to change continuously. Through the interaction between the basis function and the signal at different scales, the energy values ​​of the signal at different time points and different frequencies are obtained. These energy values ​​are arranged according to the correspondence between time and frequency to form the time-frequency energy distribution of the motion-related cortical potential signal.

[0116] A specific frequency range of 0.5Hz to 30Hz is preset, which covers the main effective frequency components of the motion-related cortical potential signal. Based on the time-frequency energy distribution, all energy data in the frequency range of 0.5Hz to 30Hz are extracted one by one at each time point. The energy data in the frequency range corresponding to each time point are accumulated and calculated to obtain the spectral integral value of each time point. According to the time sequence within the time analysis window, the spectral integral values ​​of all time points are arranged in sequence to form the energy change sequence of the motion-related cortical potential signal.

[0117] The beneficial effects include: defining the time analysis window centered on the effective attention marker moment, accurately locking the signal interval related to motor intention, avoiding interference from irrelevant time segments on the effective signal, and ensuring the relevance of the analysis object; using Morlet wavelet basis functions for continuous wavelet transform, and comprehensively and meticulously capturing the energy change characteristics of motor-related cortical potential signals in the time-frequency domain through pointwise convolution of the basis functions with the signal and continuous adjustment of scale parameters, ensuring the integrity and accuracy of the time-frequency energy distribution; clearly defining specific frequency ranges and performing spectral integration sorting, focusing on the effective energy components related to motion, eliminating the influence of high-frequency noise and low-frequency interference, and the energy change sequence formed in chronological order can realistically and coherently reflect the dynamic energy change law of motor-related cortical potential signals; the entire implementation process is clear in steps, specific and reproducible in operation, and the processing method of each step has clear basis and specific operating specifications, providing accurate and reliable energy feature data for the spatial distribution analysis of the subsequent side-effect feature extraction module and the intention judgment of the hierarchical intention decision module, effectively improving the accuracy and stability of the entire system in interpreting the patient's motor intention.

[0118] The lateralization feature extraction module 104 is used to perform motor cortex chemometry analysis on the patient based on the spatial distribution of the energy change sequence to obtain the patient's contralateral innervation pattern.

[0119] In this embodiment of the invention, when the lateralization feature extraction module performs motor cortical biochemical analysis on the patient based on the spatial distribution of the energy change sequence to obtain the patient's contralateral innervation pattern, it is specifically used for:

[0120] Based on the electrode spatial location information of the energy change sequence, a first set of electrode channels covering the left motor cortex region of the patient and a second set of electrode channels covering the right motor cortex region of the patient are defined.

[0121] Extract the energy change sequences corresponding to the first group of electrode channels and the energy change sequences corresponding to the second group of electrode channels, respectively.

[0122] The difference between the average energy of the first group of electrode channels and the second group of electrode channels is quantified to obtain the lateralization energy difference of the patient.

[0123] The sign and amplitude of the lateralization energy difference are coupled and analyzed to obtain the contralateral control pattern of the patient.

[0124] The coordinates of the scalp region corresponding to the left motor cortex of the patient were pre-defined as follows: the anteroposterior diameter is 10 cm to 16 cm posterior to the glabella, and the lateral diameter is 2 cm to 6 cm left of the midline of the head. The coordinates of the scalp region corresponding to the right motor cortex were also pre-defined as follows: the anteroposterior diameter is 10 cm to 16 cm posterior to the glabella, and the lateral diameter is 2 cm to 6 cm right of the midline of the head. Based on the spatial position information of the electrodes corresponding to the energy change sequence, the spatial coordinates of each acquisition electrode were checked one by one. Electrodes whose coordinates fall within the scalp region corresponding to the left motor cortex were classified as the first group of electrode channels, and electrodes whose coordinates fall within the scalp region corresponding to the right motor cortex were classified as the second group of electrode channels. The first group of electrode channels covering the left motor cortex region and the second group of electrode channels covering the right motor cortex region were clearly defined.

[0125] Based on the defined classification of the first and second groups of electrode channels, energy change sequences corresponding one-to-one with each electrode channel in the first group are selected from all energy change sequences. These sequences are then integrated to form the energy change sequences corresponding to the first group of electrode channels. At the same time, energy change sequences corresponding one-to-one with each electrode channel in the second group are selected and integrated to form the energy change sequences corresponding to the second group of electrode channels, ensuring that the two groups of sequences are completely matched with their respective electrode channels.

[0126] For each energy change sequence corresponding to the first group of electrode channels, the energy values ​​of all time points included in it are counted. The sum of all energy values ​​is divided by the total number of time points to obtain the average energy of each sequence. Then, the average energy of all sequences in the first group is summed and divided by the number of electrode channels in the first group to obtain the average energy of the first group of electrode channels. Using the same calculation method, the average energy of each energy change sequence in the second group is obtained first, and then the average energy of the second group of electrode channels is calculated. The specific value obtained by subtracting the average energy of the second group of electrode channels from the average energy of the first group of electrode channels is the lateralization energy difference of the patient.

[0127] The preset amplitude determination threshold is 0.5 microvolts. First, the sign of the lateralization energy difference is determined, and then the relationship between its amplitude and 0.5 microvolts is determined, performing a coupling analysis of sign and amplitude. When the sign of the lateralization energy difference is positive and the amplitude is greater than 0.5 microvolts, the patient's contralateral control pattern is determined to be right-sided motor control. When the sign of the lateralization energy difference is negative and the absolute value of the amplitude is greater than 0.5 microvolts, the patient's contralateral control pattern is determined to be left-sided motor control. When the amplitude of the lateralization energy difference is less than or equal to 0.5 microvolts, regardless of whether the sign is positive or negative, the patient's contralateral control pattern is determined to be without a clear contralateral control pattern.

[0128] The beneficial effects are as follows: defining electrode channel groups by clearly defining scalp region coordinates ensures the precise correspondence between the first and second groups of electrode channels and the left and right motor cortexes, providing an accurate spatial division basis for subsequent analysis; extracting energy change sequences by group ensures the specificity and uniqueness of the sequence source and avoids confusion of signals from different regions; the calculation process of average energy is clear and logically rigorous, and the lateralization energy difference can objectively reflect the energy difference between the two motor cortexes; the coupling analysis of sign and amplitude, combined with clear judgment thresholds, makes the judgment results of contralateral control patterns consistent and reproducible; the entire implementation process is open and detailed, effectively extracting the lateralization features of the patient's motor cortex, providing reliable spatial feature data for the graded intention decision module to accurately judge the patient's motor intention, and improving the accuracy of the system in recognizing the patient's motor intention.

[0129] The graded intention decision module 105 is used to make graded intention decisions for the patient based on the effective attention state, the energy change sequence and the contralateral dominance pattern, so as to obtain the patient's preliminary motor intention.

[0130] In this embodiment of the invention, when the graded intention decision module performs graded intention decision-making on the patient based on the effective attention state, the energy change sequence, and the contralateral dominance pattern to obtain the patient's preliminary motor intention, it is specifically used for:

[0131] The stability of the effective attention state is assessed to obtain the patient's attention weighting factor.

[0132] The normalized amplitude characteristics of the energy change sequence are analyzed statistically to obtain the patient's energy confidence factor.

[0133] A significant quantitative analysis of the contralateral control pattern was performed to obtain the spatial lateralization factor of the patient.

[0134] The patient's overall intent credibility is calculated based on the attention weighting factor, the energy confidence factor, and the spatial lateralization factor.

[0135] Based on the comprehensive intent credibility, parametric motion mapping is performed on the contralateral control pattern to obtain the patient's preliminary motion intent.

[0136] The formula for calculating the credibility of the comprehensive intent is as follows:

[0137] ;

[0138] In the formula, This indicates the credibility of the overall intent. This represents the attention weight factor. This represents the energy confidence factor. This represents the spatial lateralization factor. This represents the preset regularization coefficient. This represents a preset, minimal positive constant. Represents the hyperbolic tangent function. This represents the arctangent function.

[0139] A stable evaluation standard for effective attention was established. The duration of effective attention and the number of fluctuations in ineffective attention were statistically analyzed. The threshold for stable effective attention duration was set at 2 seconds, and the threshold for the number of fluctuations was set at 3. If the continuous effective attention duration reached or exceeded 2 seconds and the number of fluctuations was 0, the attention weight factor was set to 1.0; if the continuous effective attention duration reached 2 seconds and the number of fluctuations was 1, the attention weight factor was set to 0.8; if the continuous effective attention duration was 1.5 seconds and the number of fluctuations was 0, the attention weight factor was set to 0.7; if the continuous effective attention duration was less than 1 second, the attention weight factor was set to 0.1 regardless of the number of fluctuations. The patient's attention weight factor was obtained through this quantitative rule.

[0140] Divide the energy values ​​at all time points in the energy change sequence by the maximum energy value in the sequence to obtain the normalized amplitude feature corresponding to each time point. Divide the 2-second time window of the energy change sequence into four 0.5-second time intervals. Calculate the average value of the normalized amplitude feature in each time interval. Compare the difference between the average value of the next time interval and the average value of the previous time interval. If the difference reaches 20% of the average value of the previous time interval, it is determined to be an upward trend. If the difference is between -20% and 20%, it is determined to be a stable trend. If the difference is less than -20%, it is determined to be a downward trend. Count the number of time intervals with an upward trend. When 3 or more of the 4 intervals show an upward trend, the energy confidence factor is set to 0.9; when 2 intervals show an upward trend, it is set to 0.7; when 1 interval shows an upward trend, it is set to 0.4; and when 0 intervals show an upward trend, it is set to 0.2. This yields the patient's energy confidence factor.

[0141] A predefined quantification standard for lateralization energy difference was established. If the contralateral control pattern was right-sided limb motor control and the corresponding lateralization energy difference amplitude was greater than 0.5 μV, the spatial lateralization factor was set to 0.8; when the lateralization energy difference amplitude was between 0.3 μV and 0.5 μV, the spatial lateralization factor was set to 0.6. If the contralateral control pattern was left-sided limb motor control and the lateralization energy difference amplitude was greater than 0.5 μV, the spatial lateralization factor was set to 0.8; when the lateralization energy difference amplitude was between 0.3 μV and 0.5 μV, the spatial lateralization factor was set to 0.6. If the contralateral control pattern was not clearly defined, the spatial lateralization factor was set to 0.2 regardless of the lateralization energy difference amplitude. The patient's spatial lateralization factor was obtained through this quantification rule.

[0142] With a preset regularization coefficient of 0.5 and a minimum positive constant of 0.0001, the first part of the calculation is performed by multiplying the attention weight factor and the energy confidence factor. The spatial lateralization factor is subtracted from the result by 1, and the minimum positive constant is added to obtain the denominator. The product is then divided by the denominator and multiplied by the regularization coefficient to obtain the first part of the calculation result. The second part of the calculation is performed by squaring the attention weight factor and the energy confidence factor, adding them together, adding the minimum positive constant, and calculating the square root of the sum to obtain the denominator. The spatial lateralization factor is then divided by the denominator, and the quotient is transformed using the arctangent function to obtain the second part of the calculation result. The first part of the calculation result is added to the second part of the calculation result, and the sum is transformed using the hyperbolic tangent function. The resulting value is the patient's overall intention credibility.

[0143] The preset threshold for determining the reliability of the overall intention is 0.6. If the reliability of the overall intention is greater than 0.6 and the contralateral control mode is right-sided limb motor control, the parameterized movement is mapped to a lifting action of the right-side actuator of the nursing bed, with a lifting angle set to 30 degrees. If the contralateral control mode is left-sided limb motor control, it is mapped to a lifting action of the left-side actuator of the nursing bed, with a lifting angle of 30 degrees. If the reliability of the overall intention is between 0.4 and 0.6, regardless of whether the contralateral control mode is left-sided or right-sided limb motor control, it is mapped to a lifting action of the corresponding side actuator, with a lifting angle of 15 degrees. If the reliability of the overall intention is less than 0.4, regardless of the contralateral control mode, it is mapped to a preliminary motor intention without a clear motor movement. The patient's preliminary motor intention is obtained through this parameterized mapping rule.

[0144] The beneficial effects are as follows: the evaluation of the attention weight factor is based on clear standards of duration and number of fluctuations; the calculation of the energy confidence factor is achieved through normalized amplitude trend statistics; the quantification of the spatial lateralization factor combines the contralateral dominance pattern and the amplitude of the lateralization energy difference. The acquisition process of the three factors is clear, standard, and reproducible. The calculation of the comprehensive intention credibility integrates the information of the three factors, and the objectivity of the calculation results is ensured through clear step-by-step operation logic and preset parameters. The parameterized motion mapping is based on the threshold division of the comprehensive intention credibility and the correspondence of the contralateral dominance pattern, so that the output of the initial motion intention has clear action direction and specific parameters. The entire implementation process is detailed and logically rigorous, with clear operational basis and quantitative standards for each step. It fully integrates the multi-dimensional features of attention, energy, and space, effectively improving the accuracy and reliability of the initial motion intention recognition, and laying a solid foundation for the control command generation module to generate the final control command by combining physiological feedback signals.

[0145] The control command generation module 106 is used to generate the final control command for the target device based on the initial movement intention and the patient's real-time physiological feedback signal.

[0146] In this embodiment of the invention, when the control command generation module generates the final control command for the target device based on the initial movement intention and the patient's real-time physiological feedback signal, it is specifically used for:

[0147] Receive the initial movement intention and the patient's real-time physiological feedback signals collected by the target device's sensors;

[0148] Physiological state features are extracted from the real-time physiological feedback signal to obtain the patient's real-time body pressure distribution information and heart rate variability index;

[0149] Based on the real-time body pressure distribution information, a biomechanical analysis is performed on the patient's current posture to obtain the patient's stability margin estimation data;

[0150] Based on the stability margin estimation data and the heart rate variability index, the safety of the preliminary exercise intention is verified.

[0151] Based on the verified preliminary motion intent, the drive parameters of the target device are encoded to obtain the final control command of the target device.

[0152] An array of body pressure sensors and a photoelectric heart rate sensor are pre-installed on the target device. The array of body pressure sensors are evenly distributed on the contact surface of the nursing bed, and the photoelectric heart rate sensor is fixed on a monitoring strap worn by the patient's wrist. The two sensors synchronously collect signals at a sampling frequency of 100 Hz. The collected body pressure signal and heart rate signal together constitute the patient's real-time physiological feedback signal. The control command generation module simultaneously receives the preliminary movement intention output by the hierarchical intention decision module and the above-mentioned real-time physiological feedback signal through the data transmission interface, ensuring the time synchronization of the two information.

[0153] The received real-time physiological feedback signals are classified and processed. For body pressure signals, the voltage signals collected by each body pressure sensor are converted into corresponding pressure values. According to the two-dimensional coordinate positions of the sensors on the contact surface of the nursing bed, all pressure values ​​are arranged in coordinate order to form the patient's real-time body pressure distribution information. For heart rate signals, the time interval between consecutive heartbeats is extracted, and the standard deviation of the time interval is calculated. The obtained value is the patient's heart rate variability index. The entire extraction process is strictly executed according to fixed signal conversion rules and calculation logic.

[0154] Based on real-time body pressure distribution information, the coordinates of the pressure center point of the contact surface between the patient's body and the nursing bed are determined. Using this center point as a reference, a biomechanical model of the patient's current posture is constructed. The straight-line distance from this center point to each point on the edge of the nursing bed contact surface is calculated. The smallest straight-line distance is selected as the stability margin estimate of the patient's current posture. This data directly reflects the stability of the patient's current posture.

[0155] The preset stability margin safety threshold is 5 cm, and the heart rate variability safety range is 10 ms to 100 ms. The obtained stability margin estimate data is compared with 5 cm, and the heart rate variability index is compared with the 10 ms to 100 ms range. When the stability margin estimate data is greater than 5 cm and the heart rate variability index is within the 10 ms to 100 ms range, the initial exercise intention is determined to meet the safety requirements, and the safety verification is passed. When the stability margin estimate data is less than or equal to 5 cm or the heart rate variability index exceeds the 10 ms to 100 ms range, the initial exercise intention is determined to not meet the safety requirements, and the safety verification is failed.

[0156] For preliminary motion intentions that have passed safety verification, the corresponding drive parameters are determined according to the type of motion. If the preliminary motion intention is to raise the left actuator of the nursing bed by 30 degrees, the corresponding drive parameters are set to a motor speed of 60 revolutions per minute and a running time of 5 seconds. If it is to raise the right actuator by 30 degrees, the drive parameters are also set to a motor speed of 60 revolutions per minute and a running time of 5 seconds. If it is to raise by 15 degrees, the drive parameters are set to a motor speed of 30 revolutions per minute and a running time of 2.5 seconds. According to the binary encoding rules that the target device drive system can recognize, the above drive parameters are converted into corresponding coded signals, which are the final control commands of the target device.

[0157] The beneficial effects include: real-time physiological feedback signals are acquired through clearly defined sensor types, installation locations, and sampling frequencies, ensuring signal accuracy and synchronization; physiological state feature extraction employs fixed transformation rules and computational logic, making the acquisition of real-time body pressure distribution information and heart rate variability indicators reproducible; biomechanical analysis, based on clear benchmarks and calculation methods, provides stability margin estimates that objectively reflect patient posture stability; safety verification, through specific threshold and range definitions, ensures the safety and feasibility of initial movement intentions; drive parameter encoding follows unified rules, and the generated final control commands can be accurately recognized and executed by the target device. The entire implementation process is specific, logically rigorous, and fully integrates the patient's real-time physiological state to optimize initial movement intentions, effectively improving the safety, accuracy, and adaptability of target device control, and ensuring patient comfort and safety during use.

[0158] Reference Figure 2 The diagram shown is a flowchart illustrating an adaptive control method for an intelligent nursing bed based on multimodal EEG intention recognition, according to an embodiment of the present invention. In this embodiment, the adaptive control method for an intelligent nursing bed based on multimodal EEG intention recognition includes:

[0159] S1. Perform multi-dimensional feature parallel separation on the raw EEG signals of the patient's occipital brain region and motor cortex brain region to obtain the patient's occipital steady-state visual evoked potential signal and motor-related cortical potential signal.

[0160] S2. Perform phase-locked stability analysis on the steady-state visual evoked potential signal in the occipital region to determine the patient's effective attention state;

[0161] S3. When the attention effective state is effective, perform cooperative time-frequency analysis on the motion-related cortical potential signal to obtain the energy change sequence of the motion-related cortical potential signal;

[0162] S4. Based on the spatial distribution of the energy change sequence, perform motor cortex biochemical analysis on the patient to obtain the patient's contralateral innervation pattern;

[0163] S5. Based on the effective attention state, the energy change sequence, and the contralateral dominance pattern, perform graded intention decision-making on the patient to obtain the patient's preliminary motor intention;

[0164] S6. Based on the initial movement intention and the patient's real-time physiological feedback signal, generate the final control command for the target device.

[0165] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0166] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0167] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. An adaptive control system for an intelligent nursing bed based on multimodal EEG intention recognition, characterized in that, The system includes a signal separation module, a state determination module, an intent feature parsing module, a side-channel feature extraction module, a hierarchical intent decision-making module, and a control command generation module, wherein: The signal separation module is used to perform multi-dimensional feature parallel separation of the raw EEG signals of the patient's occipital brain region and motor cortex brain region to obtain the patient's occipital steady-state visual evoked potential signal and motor-related cortical potential signal. The state determination module is used to perform phase-locked stability analysis on the steady-state visual evoked potential signal in the occipital region to determine the patient's effective attention state. The intent feature parsing module is used to perform cooperative time-frequency analysis on the motion-related cortical potential signal when the attention effective state is effective, so as to obtain the energy change sequence of the motion-related cortical potential signal; The lateralization feature extraction module is used to perform motor cortex chemometry analysis on the patient based on the spatial distribution of the energy change sequence to obtain the patient's contralateral innervation pattern. The graded intention decision module is used to make graded intention decisions for the patient based on the effective attention state, the energy change sequence, and the contralateral dominance pattern, so as to obtain the patient's preliminary motor intention. The control command generation module is used to generate the final control command for the target device based on the initial movement intention and the patient's real-time physiological feedback signal.

2. The intelligent nursing bed adaptive control system based on multimodal EEG intention recognition as described in claim 1, characterized in that, The signal separation module, when performing multi-dimensional feature parallel separation of the raw EEG signals from the patient's occipital region and motor cortex to obtain the patient's occipital region steady-state visual evoked potential signal and motor-related cortical potential signal, is specifically used for: Simultaneously collect raw electroencephalogram (EEG) signals from the patient's occipital region and motor cortex. By removing power frequency interference and physiological artifacts from the original EEG signal, a clean EEG signal is obtained from the original EEG signal. Blind source separation was performed on the clean EEG signal to obtain the patient's independent signal source components; Based on the scalp topology distribution map of the independent signal source components, the first type of component set of the occipital brain region and the second type of component set of the motor cortex brain region are screened out. Based on the matching relationship between the peak power spectrum of the first type of component set and the preset visual stimulus frequency, the corresponding matching frequency signal components in the first type of component set are superimposed in the time domain to obtain the steady-state visual evoked potential signal of the patient's occipital region. Event-related potential analysis is performed on the second type of component set, and waveform synthesis is performed on the analyzed motor preparation feature components to obtain the motor-related cortical potential signal of the patient.

3. The intelligent nursing bed adaptive control system based on multimodal EEG intention recognition as described in claim 2, characterized in that, When the state determination module performs phase-locked stability analysis on the steady-state visual evoked potential signal in the occipital region to determine the patient's effective attention state, it is specifically used for: A sine-cosine reference signal with the same frequency as the preset visual stimulus is generated by a digital oscillator. Calculate the instantaneous phase-locked value between the steady-state visual evoked potential signal in the pillow region and the sine-cosine reference signal; The stability index of the instantaneous phase-locked value is obtained by performing in-window standard deviation analysis on the instantaneous phase-locked value. The stability index is compared with a preset first stability threshold. When the stability index continuously exceeds the first stability threshold, the patient's effective attention state is determined to be effective; When the stability index does not continuously exceed the first stability threshold, the patient's effective attention state is determined to be invalid.

4. The intelligent nursing bed adaptive control system based on multimodal EEG intention recognition as described in claim 3, characterized in that, The formula for calculating the instantaneous phase-locked value is as follows: ； In the formula, Indicates at time The instantaneous phase-locked value, Represents the imaginary unit. This represents the number of adjacent time points used for local estimation. Indicates at time No. The instantaneous phase of the steady-state visual evoked potential signal in the pillow region at each adjacent time point. Indicates at time The instantaneous phase of the sine-cosine reference signal, Represents a complex exponential function. This represents the modulo operation for complex numbers.

5. The intelligent nursing bed adaptive control system based on multimodal EEG intention recognition as described in claim 1, characterized in that, When the intent feature parsing module performs cooperative time-frequency analysis on the motion-related cortical potential signal to obtain the energy change sequence of the motion-related cortical potential signal when the attention effective state is valid, it is specifically used for: Based on the event-related characteristics of the motion-related cortical potential signal, a time analysis window for the motion-related cortical potential signal is established with the moment when the effective attention state is received as the effective marker as the center. Within the time analysis window, continuous wavelet transform is performed on the motion-related cortical potential signal to obtain the time-frequency energy distribution of the motion-related cortical potential signal; The energy data of a specific frequency in the time-frequency energy distribution are sorted by spectral integration to obtain the energy change sequence of the motion-related cortical potential signal.

6. The intelligent nursing bed adaptive control system based on multimodal EEG intention recognition as described in claim 1, characterized in that, When the lateralization feature extraction module performs motor cortical biochemical analysis on the patient based on the spatial distribution of the energy change sequence to obtain the patient's contralateral innervation pattern, it is specifically used for: Based on the electrode spatial location information of the energy change sequence, a first set of electrode channels covering the left motor cortex region of the patient and a second set of electrode channels covering the right motor cortex region of the patient are defined. Extract the energy change sequences corresponding to the first group of electrode channels and the energy change sequences corresponding to the second group of electrode channels, respectively. The difference between the average energy of the first group of electrode channels and the second group of electrode channels is quantified to obtain the lateralization energy difference of the patient. The sign and amplitude of the lateralization energy difference are coupled and analyzed to obtain the contralateral control pattern of the patient.

7. The intelligent nursing bed adaptive control system based on multimodal EEG intention recognition as described in claim 1, characterized in that, When the graded intention decision-making module performs graded intention decision-making on the patient based on the effective attention state, the energy change sequence, and the contralateral dominance pattern to obtain the patient's preliminary motor intention, it is specifically used for: The stability of the effective attention state is assessed to obtain the patient's attention weighting factor. The normalized amplitude characteristics of the energy change sequence are analyzed statistically to obtain the patient's energy confidence factor. A significant quantitative analysis of the contralateral control pattern was performed to obtain the spatial lateralization factor of the patient. The patient's overall intent credibility is calculated based on the attention weighting factor, the energy confidence factor, and the spatial lateralization factor. Based on the comprehensive intent credibility, parametric motion mapping is performed on the contralateral control pattern to obtain the patient's preliminary motion intent.

8. The intelligent nursing bed adaptive control system based on multimodal EEG intention recognition as described in claim 7, characterized in that, The formula for calculating the credibility of the comprehensive intent is as follows: ； In the formula, This indicates the credibility of the overall intent. This represents the attention weight factor. This represents the energy confidence factor. This represents the spatial lateralization factor. This represents the preset regularization coefficient. This represents a preset, minimal positive constant. Represents the hyperbolic tangent function. This represents the arctangent function.

9. The intelligent nursing bed adaptive control system based on multimodal EEG intention recognition as described in claim 1, characterized in that, When the control command generation module generates the final control command for the target device based on the initial movement intention and the patient's real-time physiological feedback signals, it is specifically used for: Receive the initial movement intention and the patient's real-time physiological feedback signals collected by the target device's sensors; Physiological state features are extracted from the real-time physiological feedback signal to obtain the patient's real-time body pressure distribution information and heart rate variability index; Based on the real-time body pressure distribution information, a biomechanical analysis is performed on the patient's current posture to obtain the patient's stability margin estimation data; Based on the stability margin estimation data and the heart rate variability index, the safety of the preliminary exercise intention is verified. Based on the verified preliminary motion intent, the drive parameters of the target device are encoded to obtain the final control command of the target device.

10. An adaptive control method for an intelligent nursing bed based on multimodal EEG intention recognition, characterized in that, The method for using the intelligent nursing bed adaptive control system based on multimodal EEG intention recognition as described in claim 1, wherein the method is as follows: S1. Perform multi-dimensional feature parallel separation on the raw EEG signals of the patient's occipital brain region and motor cortex brain region to obtain the patient's occipital steady-state visual evoked potential signal and motor-related cortical potential signal. S2. Perform phase-locked stability analysis on the steady-state visual evoked potential signal in the occipital region to determine the patient's effective attention state; S3. When the attention effective state is effective, perform cooperative time-frequency analysis on the motion-related cortical potential signal to obtain the energy change sequence of the motion-related cortical potential signal; S4. Based on the spatial distribution of the energy change sequence, perform motor cortex biochemical analysis on the patient to obtain the patient's contralateral innervation pattern; S5. Based on the effective attention state, the energy change sequence, and the contralateral dominance pattern, perform graded intention decision-making on the patient to obtain the patient's preliminary motor intention; S6. Based on the initial movement intention and the patient's real-time physiological feedback signal, generate the final control command for the target device.

Images