Intelligent nursing bed based on brain-computer interface and control method thereof

Abstract

The application relates to the technical field of brain-computer interfaces, and discloses an intelligent nursing bed based on a brain-computer interface and a control method thereof, the method comprising the following steps: S1: collecting original electroencephalogram signals and extracting mu rhythm features and beta rhythm features; S2: decoding the mu rhythm features and the beta rhythm features into corresponding nursing instructions, and dividing the nursing instructions into high-confidence instructions, medium-confidence instructions and low-confidence instructions; S3: when the nursing instruction is a high-confidence instruction, directly outputting the nursing instruction and corresponding action parameters, and directly executing a nursing action without physiological signal verification; when the nursing instruction is a medium-confidence instruction, calculating a cognitive load level and generating a comprehensive decision result; and when the nursing instruction is a low-confidence instruction, rejecting the nursing instruction and not triggering any nursing action, thereby solving the problem of false triggering caused by electroencephalogram signal fluctuation, and realizing intelligent processing of nursing operations.

Description

Technical Field

[0001] This invention relates to the field of brain-computer interface technology, and in particular to a brain-computer interface-based intelligent nursing bed and its control method. Background Technology

[0002] People with disabilities and postoperative rehabilitation patients are bedridden for a long time and have a strong need for self-triggering of nursing functions such as back raising, turning over, leg raising, and massage. However, existing smart nursing beds rely on handheld devices, bedside buttons, or assistance from others. Those with limited upper limb function cannot achieve self-control, and the timeliness of nursing care is heavily dependent on the response speed of others.

[0003] Existing BCI decoding models are designed only for single motor intentions, failing to cover the needs of multiple nursing functions. Furthermore, relying solely on EEG signal decoding is prone to false triggering rates exceeding 5% due to signal fluctuations, lacking a decision-making mechanism that integrates physiological state data with EEG signals to reduce the risk of false triggering. Additionally, existing BCI help-seeking functions require users to actively initiate motor imagery, with decoding delays exceeding 500 milliseconds. In sudden situations where users are confused or their motor imagery abilities are impaired, they cannot quickly trigger a help-seeking response, delaying rescue opportunities. Summary of the Invention

[0004] This invention provides an intelligent nursing bed based on a brain-computer interface and its control method. This invention further solves the problem of false triggering caused by fluctuations in brain electrical signals and realizes intelligent processing of nursing operations.

[0005] The first aspect of this invention provides a brain-computer interface-based intelligent nursing bed control method, the method comprising: S1: Acquire raw EEG signals and extract μ-rhythm and β-rhythm features; S2: Decode the μ rhythm features and the β rhythm features into corresponding nursing instructions, and classify the nursing instructions into high-confidence instructions, medium-confidence instructions and low-confidence instructions; S3: When the nursing instruction is a high-confidence instruction, the nursing instruction and corresponding action parameters are directly output, and the nursing action is executed directly without physiological signal verification; when the nursing instruction is a medium-confidence instruction, the cognitive load level is calculated and a comprehensive decision result is generated; when the nursing instruction is a low-confidence instruction, the nursing instruction is rejected and no nursing action is triggered.

[0006] In conjunction with the first aspect, in a first implementation of the first aspect of the present invention, step S1 includes: S11: Acquire raw EEG signals and remove electrooculography artifacts from the raw EEG signals to obtain preprocessed EEG signals; S12: Perform a short-time Fourier transform on the preprocessed EEG signal to calculate the μ rhythm features of the μ band and the β rhythm features of the β band.

[0007] In conjunction with the first aspect, in a second implementation of the first aspect of the present invention, step S2 includes: S21: Calculate the probability of getting up, turning over, raising legs, massaging, and seeking emergency help based on the μ rhythm features and the β rhythm features; S22: Find the maximum probability value among the back-raising probability, the turning-over probability, the leg-raising probability, the massage probability, and the emergency help probability, take the instruction type corresponding to the maximum probability value as the nursing instruction, and take the maximum probability value as the confidence level; S23: Compare the confidence level with preset high confidence thresholds and low confidence thresholds. When the confidence level is greater than or equal to the high confidence threshold, determine the nursing instruction as a high confidence instruction. When the confidence level is between the low confidence threshold and the high confidence threshold, determine the nursing instruction as a medium confidence instruction. When the confidence level is less than the low confidence threshold, determine the nursing instruction as a low confidence instruction.

[0008] In conjunction with the first aspect, in a third implementation of the first aspect of the present invention, step S3 includes: S31: When the nursing instruction is a high-confidence instruction, the nursing instruction and corresponding action parameters are directly output, and the physiological signal verification is skipped and the nursing action is executed directly. S32: When the nursing instruction is a moderate confidence instruction, calculate the cognitive load level, and perform a comprehensive analysis of the EEG confidence score, stress verification score, heart rate verification score, and cognitive load verification score based on the nursing instruction to obtain a comprehensive decision result; S33: When the nursing instruction is a low-confidence instruction, rejecting the nursing instruction does not trigger any nursing action.

[0009] In conjunction with the first aspect, in a fourth implementation of the first aspect of the present invention, step S32 includes: S321: The sensor array collects mattress pressure, the heart rate is collected by the heart rate sensor, and the ratio of theta wave power to alpha wave power in the prefrontal cortex EEG signal is calculated to obtain the cognitive load level. S322: Calculate the stress verification score, heart rate verification score, and cognitive load verification score based on the nursing instructions and the cognitive load level, respectively. S323: The EEG confidence score, the stress verification score, the heart rate verification score, and the cognitive load verification score are weighted and summed to obtain a comprehensive decision result.

[0010] In conjunction with the first aspect, in a fifth implementation of the first aspect of the present invention, step S321 includes: S3211: Collects pressure data of the user's torso, hips, and legs at a preset sampling frequency using a sensor array to obtain mattress pressure; S3212: Measures heart rate using photoplethysmography (PPG) technology via a heart rate sensor; S3213: Extract the first power spectral density of the theta wave band and the second power spectral density of the alpha wave band from the EEG signal of the prefrontal region, calculate the ratio of the first power spectral density to the second power spectral density, and obtain the cognitive load level.

[0011] In conjunction with the first aspect, in the sixth implementation of the first aspect of the present invention, step S322 includes: S3221: When the nursing instruction is a turning instruction, the pressure centroid coordinates are calculated based on the mattress pressure, and the pressure verification score is determined based on the matching relationship between the pressure centroid coordinates and the turning direction. S3222: When the nursing instruction is a back-raising instruction, extract the pressure data of the buttock area in the mattress pressure, calculate the pressure standard deviation, and determine the pressure verification score based on the pressure standard deviation and the preset stability threshold. S3223: When the nursing instruction is a leg-lifting instruction, extract the pressure data of the leg area in the mattress pressure to verify the leg posture and determine the pressure verification score; S3224: When the nursing instruction is a massage instruction, the heart rate is compared with the preset heart rate range to determine the heart rate verification score, and the cognitive load level is compared with the preset load threshold to determine the cognitive load verification score; S3225: When the nursing instruction is an emergency help instruction, assess the degree of urgency based on the heart rate and the cognitive load level, and determine the heart rate verification score and the cognitive load verification score.

[0012] In conjunction with the first aspect, in the seventh implementation of the first aspect of the present invention, after step S323, the method further includes: S324: Based on the comprehensive decision-making results and the nursing instructions, respectively drive the electric push rod to adjust the backboard to the first angle, or drive the airbag turning system to perform the turning action according to the inflation sequence, or drive the thigh board electric push rod to adjust the thigh board to the second angle, or drive the vibration massage component to adjust the massage intensity according to the cognitive load level to perform the massage action, or when the nursing instruction is an emergency help instruction, activate the sound and light alarm and push the help data packet to the bound mobile terminal.

[0013] In conjunction with the first aspect, in the eighth implementation of the first aspect of the present invention, after step S3, the following is further included: S4: Calculate the theta wave power of the original EEG signal in real time, and calculate the ratio of the difference between the theta wave power and the baseline theta wave power to the baseline theta wave power to obtain the surge amplitude; S5: Determine whether the surge is greater than or equal to a preset threshold and whether the duration is greater than or equal to a preset duration. When both conditions are met, an emergency help command is automatically triggered. S6: After triggering the emergency help command, activate the buzzer and LED light array to provide an audible and visual alarm, push a help data packet containing EEG data summary and location information to the bound mobile terminal, and switch all nursing actions to safe mode.

[0014] A second aspect of the present invention provides an intelligent nursing bed, the intelligent nursing bed comprising: The acquisition module is used to acquire raw EEG signals and extract μ-rhythm and β-rhythm features; The decoding module is used to decode the μ rhythm features and the β rhythm features into corresponding nursing instructions, and to classify the nursing instructions into high-confidence instructions, medium-confidence instructions and low-confidence instructions; The graded output module is used to directly output the nursing instruction and its corresponding action parameters when the nursing instruction is a high-confidence instruction, and to directly execute the nursing action without verifying the physiological signal; when the nursing instruction is a medium-confidence instruction, it calculates the cognitive load level and generates a comprehensive decision result; when the nursing instruction is a low-confidence instruction, it rejects the nursing instruction and does not trigger any nursing action.

[0015] Compared with existing technologies, this invention solves the problem of false triggering caused by EEG signal fluctuations. When the decoding confidence level is at a moderate level, physiological data such as mattress pressure distribution, heart rate, and prefrontal cognitive load level are collected simultaneously. Validation rules are designed for different nursing instruction types: the turning instruction verifies the matching of body posture and turning direction using pressure centroid coordinates; the back-raising instruction verifies postural stability using the standard deviation of pressure in the buttock area; and massage and emergency help instructions assess physiological stress state using heart rate and cognitive load levels. The EEG confidence level and multidimensional validation scores are weighted and fused to generate a comprehensive decision result, reducing the false triggering rate. Based on a passive triggering mechanism of theta wave power relative to baseline surge amplitude and duration, this invention can automatically trigger emergency help instructions and link audio-visual alarms and remote push notifications even when the user's motor imagination is impaired due to a disease attack, solving the response lag problem of traditional active triggering methods. The massage intensity parameters are adaptively adjusted according to the cognitive load level, automatically reducing stimulation intensity under high cognitive load conditions to avoid increasing psychological burden, thus achieving intelligent processing of nursing operations. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] The structures, proportions, sizes, etc., shown in the accompanying drawings of this specification are only for the purpose of assisting those skilled in the art in understanding and reading the content disclosed in the specification, and are not intended to limit the conditions under which the present invention can be implemented. Therefore, they have no substantial technical significance. Any modifications to the structure, changes in the proportions, or adjustments to the size, without affecting the effects and objectives that the present invention can produce, should still fall within the scope of the technical content disclosed in the present invention.

[0018] Figure 1 This is a flowchart illustrating the intelligent nursing bed control method based on a brain-computer interface provided in an embodiment of the present invention. Figure 2 This is a schematic diagram of the distribution of EEG electrodes provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the theta wave power relative to the baseline surge amplitude monitoring and emergency call triggering mechanism provided in an embodiment of the present invention; Figure 4 This is a schematic block diagram of the structure of the intelligent nursing bed provided in an embodiment of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the 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.

[0020] The flowchart shown in the attached diagram is for illustrative purposes only and does not necessarily include all content and operations / steps, nor does it necessarily have to be performed in the order described. For example, some operations / steps can be broken down, combined, or partially merged, so the actual execution order may change depending on the actual situation.

[0021] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0022] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination of one or more of the associated listed items, and all possible combinations, and includes such combinations. See also Figure 1 One embodiment of the intelligent nursing bed control method based on brain-computer interface in this invention includes: S1: Acquire raw EEG signals and extract μ-rhythm and β-rhythm features; S2: Decode the μ rhythm features and β rhythm features into corresponding nursing instructions, and classify the nursing instructions into high-confidence instructions, medium-confidence instructions and low-confidence instructions; S3: When the nursing instruction is a high-confidence instruction, the nursing instruction and corresponding action parameters are directly output, and the nursing action is executed directly without physiological signal verification; when the nursing instruction is a medium-confidence instruction, the cognitive load level is calculated and a comprehensive decision result is generated; when the nursing instruction is a low-confidence instruction, the nursing instruction is rejected and no nursing action is triggered.

[0023] Figure 2 This is a top view of the electrode layout of the 16-channel EEG acquisition system used in this invention. The figure shows the spatial location and functional zones of each electrode on the scalp. The Fp1 and Fp2 electrodes in the prefrontal cortex are located in front of the forehead, forming the prefrontal cognitive load monitoring area. They are used to acquire theta and alpha wave signals and calculate the theta / alpha power ratio to assess the user's cognitive load level. The C3 and C4 electrodes in the central region are located in the motor cortex of the left and right hemispheres, respectively. The C3 electrode, located in the left motor cortex, is responsible for acquiring right-side limb motor imagery signals and decoding commands such as "raise your back" and "roll over." The C4 electrode, located in the right motor cortex, is responsible for acquiring left-side limb motor imagery signals and decoding massage commands. Both electrodes are used to extract μ-rhythm and β-rhythm features. The Cz electrode, located in the C# midmotor area at the center of the top of the head, is responsible for decoding the bi-leg motor imagery task corresponding to the leg-raising command. CMS and DRL are the reference electrodes. Other electrodes, including F3, F4, F7, F8, FC5, FC6, T7, T8, P3, P4, P7, P8, O1, and O2, are distributed in the frontal, temporal, parietal, and occipital regions, forming a complete 16-channel acquisition array with a sampling frequency of 500Hz, a signal resolution of 16 bits, and an input impedance greater than or equal to 10MΩ.

[0024] In one specific embodiment, step S1 includes: S11: Acquire raw EEG signals and remove electrooculography artifacts from the raw EEG signals to obtain preprocessed EEG signals; S12: Perform short-time Fourier transform on the preprocessed EEG signal to calculate the μ rhythm features of the μ band and the β rhythm features of the β band.

[0025] Specifically, a non-invasive EEG acquisition helmet that meets the requirements for bioelectrical signal detection is used as the hardware foundation. The helmet is equipped with a dry electrode array to improve wearing comfort, and has a high input impedance (e.g., ≥10MΩ) to reduce interference from skin resistance. It continuously acquires raw EEG signals from the user's scalp surface at a sampling frequency of 500Hz. During signal acquisition, time-domain signals from multiple electrode channels are recorded, including the target brain-derived signal components, as well as artifacts generated by eye movements (such as blinking and gaze shifts), some facial electromyography interference, and power supply noise. Artifact removal is performed on the raw EEG signals. Spatial demixing of the raw signals is performed based on an independent component analysis (ICA) algorithm. Multiple channel signals are input into an ICA model, and several independent components are extracted. Template matching and feature discrimination are performed on each component. Utilizing the high correlation of eye movements in the frontal pole channels (e.g., Fp1, Fp2), the Pearson correlation coefficient between each component and a standard eye template is calculated. When the correlation coefficient exceeds a set threshold (e.g., 0.8), the component is determined to be an eye artifact. Components identified as artifacts are zeroed out or suppressed during signal reconstruction. The preprocessed EEG signal, free of EEG artifacts, is then regenerated through inverse transform of the reconstructed signal. The retained signal preserves the authentic neural oscillation characteristics while reducing the influence of non-brain-derived signals. A short-time Fourier transform (SFT) is performed on the preprocessed EEG signal for each channel, and the μ-rhythm (8–13 Hz) and β-rhythm (13–30 Hz) in the time-frequency domain are decomposed. The SFT uses a fixed-length sliding window (e.g., 1 second window length, 50% overlap) to segment the signal, and a fast Fourier transform (FFT) is performed on each segment to obtain the amplitude spectrum or power spectral density of the corresponding frequency band. The spectral data within each time window is integrated, and the average power values ​​within the μ-band and β-band are calculated to obtain the μ-rhythm and β-rhythm features for that time window. To enhance the recognition of motor intention, rhythmic power is extracted from motor cortical channels such as C3 and C4, and a moving average is performed based on multiple time windows to suppress short-term fluctuations.

[0026] In this embodiment, between steps S11 and S12, the following steps are also included: S111: Calculate signal quality indices for each of the 16 channels of the preprocessed EEG signal. By calculating the signal-to-noise ratio and impedance stability of each channel signal, identify inferior channels with signal quality below a preset quality threshold and superior channels with signal quality above a preset quality threshold; S112: Apply a spatial filtering algorithm to the identified inferior channels and reconstruct the signal based on the weighted average of the signals of adjacent superior channels. The spatial filtering weight coefficient is inversely proportional to the spatial distance between channels and directly proportional to the signal quality index of adjacent channels to obtain the reconstructed channel signal; S113: Compress the superior channel and reconstructed channel signals into channel descriptors using global average pooling. Calculate the attention weight coefficients of each channel using one-dimensional convolution. Multiply the attention weight coefficients element-wise with each channel signal to achieve adaptive enhancement of motor cortex-related channels and adaptive suppression of prefrontal cortex-irrelevant channels, obtaining a quality-optimized EEG signal for short-time Fourier transform.

[0027] In one specific embodiment, step S2 includes: S21: Calculate the probability of getting up, turning over, raising legs, massaging, and seeking emergency help based on μ-rhythm features and β-rhythm features; S22: Find the maximum probability value among the probability of lifting the back, turning over, raising the leg, massaging, and emergency assistance, take the instruction type corresponding to the maximum probability value as the nursing instruction, and take the maximum probability value as the confidence level; S23: Compare the confidence level with preset high confidence thresholds and low confidence thresholds. When the confidence level is greater than or equal to the high confidence threshold, the nursing instruction is determined to be a high confidence instruction. When the confidence level is between the low confidence threshold and the high confidence threshold, the nursing instruction is determined to be a medium confidence instruction. When the confidence level is less than the low confidence threshold, the nursing instruction is determined to be a low confidence instruction.

[0028] Specifically, a multi-class probabilistic decoding model based on μ-rhythm and β-rhythm features is constructed. The μ-rhythm and β-rhythm features are used as input feature vectors and input into the trained deep neural network model. The model structure adopts a convolutional neural network architecture with residual connections, containing multiple convolutional layers, batch normalization layers, and ReLU nonlinear activation functions. This effectively extracts the spatiotemporal distribution relationship between μ / β rhythms and the activity patterns of specific brain regions, thereby enhancing the model's ability to discriminate different nursing intentions. A Softmax function is set in the output layer to perform multi-class probability predictions for five types of nursing action intentions: lifting the back, turning over, raising the leg, massaging, and emergency assistance. The Softmax function maps the input linear combination vector to a probability distribution between 0 and 1, with the sum of all five probabilities being 1, resulting in five probability values ​​representing the five types of action intentions, denoted as P1 (lifting the back), P2 (turning over), P3 (raising the leg), P4 (massaging), and P5 (emergency assistance). Each probability value represents the degree of matching between the current EEG features and the corresponding action imagery pattern. The five predicted probabilities are compared, and the argmax operation is performed to determine the maximum value P. max And determine its position index i in the five types of actions. max , i.e. i max =argmax(P1,P2,P3,P4,P5). The nursing action corresponding to the location index is the type of nursing instruction currently recognized by the EEG decoding system, and the maximum probability value P... max This is considered the decoding confidence level of the instruction, used to measure the reliability of the judgment result. The nursing instruction and the confidence level are encapsulated into a structured recognition result. The confidence level P... max With the set high confidence threshold T high and low confidence threshold T low Comparative analysis was conducted, T high Set to 0.85, T low The confidence level was set to 0.60, and a three-level confidence scale was established accordingly. When the confidence level P... max Greater than or equal to the high confidence threshold T high When the nursing instruction is deemed highly reliable, it is directly labeled as a high-confidence instruction, allowing execution to skip the fusion verification step; when the confidence level is at a low threshold T... low With high threshold T high Between, i.e., 0.60 ≤ P max When the confidence level is less than 0.85, the current judgment result is considered to have some reasonableness but still carries a risk of misidentification. In this case, the instruction is marked as a medium-confidence instruction, and fusion verification using physiological signals or cognitive load data is required to improve control safety. When the confidence level is less than the low-confidence threshold T... low That is, P maxWhen the value is less than 0.60, the current EEG feature matching is determined to be unstable or ambiguous, and the decoding result lacks sufficient credibility. At this time, the current instruction is determined to be a low-confidence instruction and rejected to prevent misidentification and triggering of incorrect nursing operations.

[0029] In this embodiment, after step S12 and before step S21, the following steps are also included: S13: Establishing a mapping relationship between 5 types of motor imagery tasks and nursing instructions, wherein the back-raising instruction corresponds to the task of imagining raising the right arm, the turning-over instruction corresponds to the task of imagining turning the head to the left, the leg-raising instruction corresponds to the task of imagining raising both legs, the massage instruction corresponds to the task of imagining clenching a fist, and the emergency help instruction corresponds to the task of imagining forcefully waving both arms; S14: During the training phase, guiding the user to complete the repeated execution of each type of motor imagery task, collecting the corresponding μ rhythm features and β rhythm features in real time as training samples, and labeling each training sample with the corresponding nursing instruction category label; S15: Analyzing the specific brain regions activated by each motor imagery task, extracting the μ rhythm inhibition features of the right motor cortex channel corresponding to the back-raising task, the β rhythm enhancement features of the bilateral motor cortex channels corresponding to the turning-over task, and the μ / β rhythm joint features of the central motor cortex channel corresponding to the leg-raising task, to obtain the differentiated EEG feature patterns of each nursing instruction.

[0030] After step S12 and before step S21, the method further includes: S16: Inputting μ-rhythm features and β-rhythm features into three parallel temporal convolutional branches. Each branch uses a first convolutional kernel size covering a first time window to capture μ-rhythm slow wave features, a second convolutional kernel size covering a second time window to capture β-rhythm fast wave features, and a third convolutional kernel size covering a third time window to capture γ-rhythm transient features. The temporal features output from the three branches are adaptively weighted and fused through an attention mechanism to obtain multi-timescale fused features; S17: Inputting the multi-timescale fused features into three parallel spatial convolutional branches. Each branch uses a full-channel convolutional kernel to model the global brain connectivity pattern and a half-channel convolutional kernel to focus on the interaction pattern between the left and right hemispheres. S18: Local channel convolution kernels extract local electrode cluster features of the motor cortex. The spatial features output from the three branches are adaptively weighted and fused through an attention mechanism to obtain multi-spatial scale fusion features. The multi-spatial scale fusion features are input into the spatial and channel reconstruction convolution module. The learnable scaling factor of each channel is generated by grouping and normalization and compared with a preset threshold. The channel is divided into information-rich channels and information-redundant channels. Group convolution is performed on the information-rich channels and then pointwise convolution is performed. Pointwise convolution is performed on the information-redundant channels and then group convolution is performed. The two convolution results are cross-multiplied and concatenated to obtain redundancy-removing and enhanced features for calculating the probability of getting up, turning over, raising legs, massage, and emergency help.

[0031] In one specific embodiment, step S3 includes: S31: When the nursing instruction is a high-confidence instruction, the nursing instruction and corresponding action parameters are output directly, and the physiological signal verification is skipped and the nursing action is executed directly. S32: When the nursing instruction is a moderate confidence instruction, calculate the cognitive load level, and perform a comprehensive analysis of the EEG confidence score, stress verification score, heart rate verification score, and cognitive load verification score based on the nursing instruction to obtain a comprehensive decision result; S33: When a nursing instruction is a low-confidence instruction, rejecting the nursing instruction does not trigger any nursing action.

[0032] Specifically, when a nursing instruction is a high-confidence instruction, meaning the confidence value is greater than or equal to a preset high-confidence threshold (e.g., 0.85), it is considered that the EEG characteristics and the target motor imagery pattern are highly matched, and the risk of misjudgment is negligible. Therefore, the identified nursing instruction type and its preset action parameters are directly encapsulated as execution instructions in a structured manner and sent to each sub-module of the bed execution unit via the CAN bus or equivalent control interface, triggering physical execution modules such as electric push rods, airbag systems, or massage components to run according to the specified parameters. If the nursing instruction is a medium-confidence instruction, with the confidence value between the preset low and high thresholds (e.g., 0.60~0.84), it enters the fusion decision path for multimodal validation analysis. The system calculates the current user's cognitive load level by determining the ratio of the theta wave (4–8Hz) to alpha wave (8–13Hz) power in the prefrontal channels (e.g., Fp1 and Fp2). After extracting the power spectral density using short-time Fourier transform, the θ / α ratio is calculated and used as the cognitive load index CL. A CL greater than 1.5 indicates a high cognitive load, 0.8 to 1.5 indicates a moderate load, and below 0.8 indicates a low load. Simultaneously, the system reads the mattress pressure sensor array to obtain the real-time pressure distribution matrix and extracts characteristic indicators matching the command type. For example, a rolling over command verifies whether the body's pressure center of mass deviates from the original posture area, or a back-raising command verifies whether the pressure distribution in the buttocks area is uniform and stable. Based on these indicators, a pressure verification score S is calculated. 压力 Simultaneously, the current heart rate value (HR) is obtained from the heart rate sensor, and a heart rate verification score (S) is derived based on its deviation from the standard range. 心率 If the heart rate is under stress (e.g., >100 bpm or <60 bpm), the score will decrease accordingly. The EEG confidence score C... eeg With three verification scores (S 压力 S 心率 S 负荷 The integrated analysis is performed using a weighted formula, i.e., the comprehensive score w1×C is calculated. eeg + w2×S 压力 + w3×S 心率 + w4×S 负荷The weights are set to 0.5, 0.2, 0.15, and 0.15 respectively. When the comprehensive score is greater than the decision threshold (e.g., 0.7), the medium confidence instruction is determined to be credible and verified, and the system proceeds to the execution path and outputs the action parameters. Otherwise, it is determined to be an untrusted instruction, not executed, and a system feedback prompt is generated to inform the user that the current instruction has not been executed, thus avoiding accidental triggering. If the nursing instruction is identified as a low confidence instruction, i.e., the confidence level is less than the low threshold (e.g., 0.60), the current instruction is directly determined to be an untrusted or invalid instruction, and it will not enter the fusion verification process. No action control instructions will be issued to the execution unit, and an "instruction rejected" status code will be returned immediately. At the same time, the current control queue will be cleared to ensure that the system will not cause the device to operate malfunction due to signal misunderstanding or fluctuations in the user's mental state.

[0033] In one specific embodiment, step S32 includes: S321: The sensor array collects mattress pressure, the heart rate is collected by the heart rate sensor, and the ratio of theta wave power to alpha wave power in the prefrontal cortex EEG signal is calculated to obtain the cognitive load level. S322: Calculate the stress verification score, heart rate verification score, and cognitive load verification score based on nursing instructions and cognitive load levels, respectively. S323: Weighted summation of EEG confidence score, stress verification score, heart rate verification score, and cognitive load verification score to obtain a comprehensive decision result.

[0034] Specifically, a 16×16 piezoresistive sensor array deployed under the mattress is used to collect real-time two-dimensional pressure distribution data of the user's contact area with the bed surface at a sampling frequency of 10Hz. Parameters such as the pressure centroid coordinates, mean pressure value, and standard deviation of the area are extracted to characterize the current posture. Simultaneously, a heart rate sensor module is activated, using photoplethysmography (PPG) to collect fingertip reflection signals and calculate the heart rate (HR) value, ensuring signal stability and the continuity of physiological rhythm. Preprocessed EEG signals are obtained from prefrontal cortex channels (such as Fp1 and Fp2), and short-time Fourier transform is performed to extract the power spectral density of theta waves (4–8Hz) and alpha waves (8–13Hz), calculating the ratio CL = Pθ / Pα. This ratio serves as a cognitive load indicator, with a higher CL indicating a higher cognitive load or state of mental stress. The physiological data is interpreted according to the identified nursing instruction type, and three verification scores are calculated based on the current cognitive load level. For the pressure verification score S... 压力 The nursing intent is matched with pressure distribution patterns. For example, for the turning instruction, if the pressure centroid shifts significantly laterally and the distribution matches the expected turning, the score is 1; otherwise, it is 0. For the back-raising action, the uniformity of pressure in the buttock area is assessed; if the standard deviation is below a set threshold, a full score is awarded. Heart rate verification score S 心率The calculation is based on the degree to which the HR value deviates from the normal range (e.g., 60-100 bpm). A linear interpolation method is used to map the heart rate to a continuous value between 0 and 1; the closer the heart rate is to the ideal physiological state, the higher the score. Meanwhile, the cognitive load verification score S... 负荷 The assessment depends on the range of the CL value. A CL less than 0.8 indicates stable cognitive state and a score of 1; a CL greater than 1.5 indicates mental stress or inattention and a score of 0; scores in other ranges vary continuously between 0 and 1. The scores of the three physiological dimensions are compared with the confidence level C of the previous EEG classification model output. eeg The results are combined using a linear weighting strategy, and the calculation expression is: Overall Score = w1 × C eeg + w2×S 压力 +w3×S 心率 + w4×S 负荷 The weighting parameters were set as w1=0.5, w2=0.2, w3=0.15, and w4=0.15 to highlight the dominant role of EEG recognition results in decision-making. Simultaneously, stress matching, heart rate stability, and cognitive state were used as redundancy safeguards to improve the robustness of instruction reliability judgment. When the comprehensive score is higher than a set threshold (e.g., 0.70), the moderate-confidence nursing instruction is considered to have sufficient execution value under multidimensional verification and is implemented; if the score does not meet the requirements, the verification is deemed a failure, and the instruction is rejected.

[0035] In this embodiment, between steps S322 and S323, the following steps are also included: S3226: When the nursing instruction is a turning instruction, calculate the normalized coordinate value of the pressure centroid coordinate in the lateral position of the mattress, and determine whether the normalized coordinate value matches the decoded turning direction. If the turning direction is to the left but the normalized coordinate value is less than the first threshold, it indicates that the user is already on the left side of the mattress, and the pressure verification score is assigned to zero; if the normalized coordinate value is greater than or equal to the first threshold, the pressure verification score is assigned to one; S3227: When the nursing instruction is a back-raising instruction, extract the pressure data within the row index range and column index range of the buttock area, calculate the pressure variance, and then... The force variance is compared with the preset stability threshold. If the pressure variance is greater than the preset stability threshold, it indicates that the posture is unstable. The pressure verification score is set to zero and a delayed execution flag is set. If the pressure variance is less than or equal to the preset stability threshold, the pressure verification score is set to one. S3228: The preset weight coefficient and preset decision threshold in step S323 are dynamically adjusted according to the historical verification pass rate. When multiple consecutive verification failures cause the historical verification pass rate to be lower than the preset pass rate threshold, the weight coefficient of EEG confidence is reduced and the weight coefficient of physiological verification score is increased. When the historical verification pass rate is higher than the preset pass rate threshold, the initial weight coefficient configuration is restored.

[0036] In one specific embodiment, step S321 includes: S3211: Collects pressure data of the user's torso, hips, and legs at a preset sampling frequency using a sensor array to obtain mattress pressure; S3212: Measures heart rate using photoplethysmography (PPG) technology via a heart rate sensor; S3213: Extract the first power spectral density of the theta wave band and the second power spectral density of the alpha wave band from the EEG signal of the prefrontal lobe region, calculate the ratio of the first power spectral density to the second power spectral density, and obtain the cognitive load level.

[0037] Specifically, a 16x16 piezoresistive sensor array deployed inside the mattress synchronously collects pressure data from the user's torso, hips, and legs at a preset sampling frequency of 10Hz. Each unit node in the sensor array outputs the pressure value of its local contact surface. Based on the distribution structure of different areas within the array, pressure information covering the user's upper back is extracted from rows 6 to 10, rows 11 to 13 are the hip contact area, and rows 14 to 16 correspond to the leg support section. A weighted average of the pressure values ​​from each area is used to calculate the overall mattress pressure distribution map. Combining the pressure centroid coordinates and local standard deviation indicators, real-time posture data is constructed. The heart rate acquisition process is then initiated, using a photoplethysmography (PPG) sensor module configured on the side of the bed or the edge of the mattress to perform non-invasive transmissive blood flow detection on the fingertips or earlobes using the 940nm infrared band. The sensor acquires light intensity reflection signals at a frequency of 100Hz, analyzes the arterial pulsation cycle through amplitude changes, extracts continuous RR interval sequences, and calculates the heart rate (HR) per minute in real time based on this. Simultaneously, it records the heart rate variability (HRV) index for stress state analysis. A parallel spectral analysis module for the prefrontal cortex EEG signals is initiated, selecting signal windows from channels Fp1 and Fp2. A Butterworth bandpass filter is applied to retain the θ and α frequency bands in the range of 4Hz to 13Hz. Short-time Fourier transform is used to decompose the filtered signal into time windows and estimate the spectrum. The first power spectral density Pθ is calculated for the 4–8Hz band, and the second power spectral density Pα is calculated for the 8–13Hz band. The cognitive load index CL = Pθ / Pα is constructed using the ratio of Pθ to Pα. This ratio is used in cognitive psychology to quantify the degree of individual attentional resource consumption and mental load state. A CL greater than 1.5 indicates a high cognitive load or stress state, a CL between 0.8 and 1.5 represents a moderate load, and a CL less than 0.8 indicates a relaxed and stable state.

[0038] In one specific embodiment, step S322 includes: S3221: When the nursing instruction is a turning instruction, the pressure centroid coordinates are calculated based on the mattress pressure, and the pressure verification score is determined based on the matching relationship between the pressure centroid coordinates and the turning direction. S3222: When the nursing instruction is a back-raising instruction, extract the pressure data of the buttock area from the mattress pressure, calculate the pressure standard deviation, and determine the pressure verification score based on the pressure standard deviation and the preset stability threshold. S3223: When the nursing instruction is a leg-lifting instruction, extract the pressure data of the leg area from the mattress pressure to verify the leg posture and determine the pressure verification score; S3224: When the nursing instruction is a massage instruction, compare the heart rate with the preset heart rate range to determine the heart rate verification score, and compare the cognitive load level with the preset load threshold to determine the cognitive load verification score. S3225: When the nursing instruction is an emergency help instruction, assess the degree of urgency based on heart rate and cognitive load levels, and determine the heart rate verification score and cognitive load verification score.

[0039] Specifically, when the identified nursing instruction is a turning instruction, the two-dimensional pressure distribution matrix provided by the mattress pressure sensor array is analyzed. By weighted summation of the spatial coordinates of all effective pressure nodes, the coordinates of the pressure centroid (x) at the current moment are calculated. c , y c ), set the x-coordinate of the centroid to c Spatial matching is performed with the turning direction (e.g., "left" or "right") indicated in the nursing action. If it is identified as "turning to the left" and the center of pressure is significantly biased towards the right side of the mattress, the posture is considered to have the necessity of turning and dynamic consistency. In this case, the pressure verification score S is assigned. 压力 Assign a value of 1; if the centroid itself is already in the target side region, then it is determined that the attitude is already in the target position or there is a recognition error, S 压力 Assign a value of 0. When the nursing instruction is a back-raising instruction, extract a submatrix of the buttock region from the pressure matrix (e.g., rows 10 to 13, columns 6 to 10), and calculate the standard deviation σ of the local region pressure data. 臀部 To reflect the stability of the current pressure distribution, if σ 臀部 If the force is less than the set stability threshold (e.g., 2 kPa), it indicates that the force is uniform and the user's posture is stable, which can support the safety of lifting the back action. 压力 Assign a value of 1; if the standard deviation is large, it indicates that the posture is unstable or there is a risk of tilting, slipping, etc., then S 压力 Assign a value of 0. When the nursing instruction is a leg-lifting instruction, analyze the pressure response of the leg area in the mattress pressure array. For example, extract the pressure data from rows 14 to 16 at the bottom of the matrix, and assess whether there are stable support points and symmetrical distribution trends in this area. If the pressure distribution on both sides of the leg is uniform and the peak areas are symmetrically arranged, then it is determined that the current leg posture is suitable for leg-lifting adjustment, and thus an S value is given. 压力=1; If the leg area shows signs of pressure relief, unilateral load, or complete lack of pressure, it indicates that the lower limb condition is not suitable for adjustment. S 压力 The value is then assigned to 0. When the nursing instruction is a massage instruction, the score is verified from two dimensions. The current heart rate (HR) is compared to whether it is within the preset safe massage heart rate range (e.g., 70–100 bpm). If the HR is within the safe range, then S... 心率 Assign a value of 1 otherwise; compare the cognitive load level CL with the set load threshold (e.g., CL=1.5). If CL<1.5, it indicates that the user is in a relaxed or moderately stressed state and is suitable for receiving stimulation, then S... 负荷 Assign a value of 1; if CL > 1.5, it indicates mental stress. 负荷 The value is assigned to 0. When the nursing instruction is an emergency help instruction, a preliminary assessment of the current health risk is made based on two physiological dimensions: heart rate and cognitive load. If the heart rate is significantly outside the safe range (e.g., greater than 120 bpm or less than 50 bpm), then S... 心率 Assign a value of 1 if CL is high, otherwise assign 0; a significant increase in CL indicates accompanying cognitive abnormalities, confusion, or sudden neurological events. 负荷 The value is assigned as 1, and vice versa, it is 0.

[0040] In one specific embodiment, after step S323, the method further includes: S324: Based on the comprehensive decision-making results and nursing instructions, the electric push rod is driven to adjust the backboard to the first angle, or the airbag turning system is driven to perform turning actions according to the inflation sequence, or the electric push rod of the thigh board is driven to adjust the thigh board to the second angle, or the vibration massage component is driven to adjust the massage intensity according to the cognitive load level to perform massage actions, or when the nursing instruction is an emergency help instruction, an audible and visual alarm is activated and a help data packet is pushed to the bound mobile terminal.

[0041] Specifically, when the nursing instruction is a backrest lifting instruction and the overall score exceeds the execution threshold, the main controller encapsulates the instruction into a data structure and broadcasts it to the electric push rod control module via the CAN bus at a baud rate of 1Mbps. This drives the backrest push rod to adjust to the target first angle at a constant angular velocity, with the angle range between 0° and 70°. The controller's built-in absolute angle encoder provides real-time feedback on the current backrest angle and uses a proportional-integral-derivative (PID) closed-loop control algorithm to dynamically correct the angle error, ensuring that the steady-state error between the actual angle and the first angle is controlled within 1°, guaranteeing smooth movement and meeting human comfort requirements. When the nursing instruction is a turning instruction, the airbag turning system is activated according to the inflation strategy sequence. By controlling the inflation pressure and sequence of 12 independent airbags, the sides, shoulders, and legs of the body are gradually lifted, forming a directional turning thrust. Real-time data from the mattress pressure array is collected to monitor changes in the pressure centroid trajectory. If the turning is not completed or the turning angle is insufficient, the airbag inflation time is automatically extended for closed-loop compensation to ensure that the user completes an effective turning. When the nursing instruction is a leg-lift instruction, the system controls the electric push rod of the thigh plate to adjust the thigh plate to a specified second angle by adjusting the extension and retraction of the push rod. The control accuracy is maintained within ±1°, and the pressure in the leg area is continuously monitored to avoid the risk of excessive lifting causing pressure collapse or postural instability. If the nursing instruction is a massage instruction, the cognitive load index (CL) is used as an input parameter to match the massage intensity level. When the CL is below 0.8, it is set to a strong stimulation mode (high frequency and high amplitude); when the CL is between 0.8 and 1.5, it is set to a medium stimulation mode; when the CL is above 1.5, it automatically switches to a soothing stimulation mode (low frequency and low amplitude) to prevent excessive stimulation under high cognitive load conditions from inducing discomfort. Subsequently, the controller starts the vibration motor and runs it at the set intensity for a fixed duration (e.g., 5 minutes). During the process, pressure changes and abnormal stress fluctuations are collected in real time to determine whether the operation needs to be interrupted. If the nursing instruction is an emergency help instruction, the audible and visual alarm module is activated. The audible alarm emits a warning sound of more than 80dB through a buzzer, and the visual alarm alerts nearby nursing staff by flashing red high frequency through an LED array. At the same time, the data interaction unit encapsulates the help data packet and sends it to the bound mobile terminal through the WiFi module. The data packet contains the user's unique identifier, trigger type, EEG and heart rate summary data, location information, and trigger timestamp.

[0042] In this embodiment, after step S324, the method further includes: S325: Based on the comparison results between the cognitive load level and the preset high load threshold and low load threshold, when the cognitive load level is greater than the preset high load threshold, it is determined to be a high cognitive load state and a massage intensity reduction coefficient is calculated; when the cognitive load level is between the preset low load threshold and the preset high load threshold, it is determined to be a medium cognitive load state and a standard massage intensity is maintained; when the cognitive load level is less than the preset low load threshold, it is determined to be a low cognitive load state and a massage intensity increase coefficient is calculated; and the amplitude and frequency parameters of the vibration massage component are adjusted according to the calculated coefficients; S326: When the nursing instruction is a bowel and bladder care instruction, a timing control sequence is generated including activating the buttock support push rod at the first moment, activating the urine suction device at the second moment, and activating the mattress heating module at the third moment. The sequence includes a second moment that is delayed by a preset delay time relative to the first moment to ensure that the buttocks are supported in place, and a third moment that is synchronized with the second moment. The negative pressure of the urine suction device is stabilized within the preset working negative pressure range through a pressure regulating valve, and the heating temperature of the mattress heating module is controlled within the preset temperature range through temperature sensor feedback. S327: The sequence control sequence and synchronization clock signal are broadcast to the electric push rod controller, airbag turning system controller, vibration massage component controller, buttocks support push rod controller, urine suction device controller and mattress heating module controller via CAN bus. The sequence control sequence and synchronization clock signal are received in real time from the execution status data and fault flag bits reported by each controller. When the completion flag bit of the preceding action is detected to be true and the fault flag bit is false, the execution command of the subsequent action is triggered to ensure that the multi-unit collaborative execution error is less than the preset collaborative error threshold.

[0043] In one specific embodiment, after step S3, the method further includes: S4: Calculate the theta wave power of the raw EEG signal in real time, calculate the ratio of the difference between the theta wave power and the baseline theta wave power to the baseline theta wave power, and obtain the surge amplitude; S5: Determine whether the surge is greater than or equal to a preset threshold and whether the duration is greater than or equal to a preset duration. When both conditions are met, an emergency help command will be automatically triggered. S6: After triggering the emergency help command, activate the buzzer and LED light array for sound and light alarm, push the help data package containing EEG data summary and location information to the bound mobile terminal, and switch all nursing actions to safe mode.

[0044] Specifically, by selecting channels closely related to consciousness fluctuations in helmet electrodes (such as F3, F4, Fp1, Fp2, etc.) and extracting EEG signals in the 4–8 Hz frequency band, short-time Fourier transform or Welch power spectrum estimation methods are used to calculate the average power of the theta wave band in real time within a time series with a window length of 500 ms and a sliding step size of 100 ms. This average power is denoted as the theta wave power Pθ(t) at the current moment. An individualized baseline theta wave power Pθ0 is pre-constructed based on the user's 24-hour historical EEG data in a resting state. The baseline value is taken as the median of past resting samples to improve adaptability to inter-individual differences. The difference between the current theta wave power and the baseline value, ΔPθ = Pθ(t), is calculated in real time. Pθ0, and the ratio of the difference to the baseline Rθ = ΔPθ / Pθ0, the latter being the theta wave power surge amplitude, used to quantify the deviation between the user's current EEG signal and their resting state. The surge amplitude Rθ is then compared with a set surge threshold R... t (e.g., 50%) is compared, and the cumulative time length Δt of the surge state is continuously recorded in the time domain, only if Rθ is greater than or equal to R t And the duration of Δt exceeds the set duration threshold T. t (For example, 1 second) only after this time is it determined that an abnormal change has occurred in the current EEG state, corresponding to a sudden high-risk physiological state such as confusion, epileptic seizure, or stroke precursors, thus triggering an emergency help command. Once the emergency help command is automatically triggered, a multi-path linkage execution process is initiated. The local audible and visual alarm module is activated via the IO control interface. The buzzer emits an intermittent sound at a frequency of 2.5kHz, with a volume not lower than 80dB, and the LED array flashes a red warning signal at a frequency of 2Hz to immediately attract the attention of caregivers or family members in the room. Simultaneously, the data interaction unit calls the communication module to push a standard-format help data packet to the bound mobile terminal via WiFi or Bluetooth. The help data packet is encapsulated in JSON format and includes the current trigger time, user unique identifier, trigger type (theta wave abnormality), theta wave and beta wave power summary within 10 seconds, prefrontal cortex power curve data, device number, GPS location information, etc., ensuring that remote caregivers or monitoring platforms receive the alarm and locate the user within 3 seconds. At the same time, all ongoing nursing actions are forcibly switched to safety mode. For example, the backrest is slowly lowered to a 15° supine position to prevent aspiration, the thigh board is returned to a horizontal position to reduce the burden on the lower back, all power operations such as turning over, massage, and hip support are stopped, and the electric push rod is locked in the current position to prevent mechanical accidental movement. The entire system enters emergency waiting state until the user's condition recovers or manual intervention takes over control.

[0045] Figure 3This invention illustrates the real-time monitoring curve and automatic triggering mechanism of theta wave power for emergency health risk detection. The horizontal axis represents the monitoring time (in seconds), and the vertical axis represents the percentage of theta wave power relative to the baseline theta wave power. The solid line represents the theta wave power curve, the dashed line marks the preset threshold line, and the gray grid area represents the area where the triggering condition is met. Under normal conditions (0-12 seconds), theta wave power fluctuates slightly around the baseline level of 100%, indicating that the user's EEG signal is stable. When a sudden health risk event occurs, the theta wave power rises sharply from the 13th second, with the surge exceeding the preset threshold (marked as 50% in the figure). The curve enters the gray triggering condition area, at which point the theta wave power reaches approximately 150% and continues for more than the preset duration (1 second). This automatically determines an emergency health risk and triggers an emergency help command. No active motor imagery is required from the user, achieving a passive and rapid response with a response delay of less than or equal to 100 milliseconds. Immediately, an audible and visual alarm and remote push function are activated, providing emergency help to disabled users when they are confused or have impaired motor imagery abilities.

[0046] In this embodiment, between steps S4 and S5, the following steps are also included: S41: Simultaneously activating the EEG abnormality detection path and the heart rate abnormality detection path; in the EEG abnormality detection path, using a sliding window method at preset update intervals to calculate the surge amplitude at the current moment and recording the cumulative duration of the surge amplitude continuously exceeding a preset threshold; in the heart rate abnormality detection path, extracting continuous heartbeat interval sequences and calculating the standard deviation and root mean square of heart rate variability; S42: Determining whether the EEG abnormality detection path meets the first triggering condition that the surge amplitude is greater than or equal to the preset threshold and the cumulative duration is greater than or equal to the preset duration; determining whether the heart rate abnormality detection path meets the condition that the standard deviation of heart rate variability is small. S43: An emergency help command is triggered when either the first or second trigger condition is met, based on a second trigger condition that is a preset standard deviation threshold and lasts for a duration longer than a preset heart rate monitoring duration; S44: A help data packet with different priorities is generated according to the detection path type that triggers the emergency help command. When the EEG abnormality detection path is triggered, the help data packet is marked as an EEG abnormality type and contains the theta wave power curve data of the most recent time period; when the heart rate abnormality detection path is triggered, the help data packet is marked as a cardiac risk type and contains heart rate variability indicators and real-time heart rate values; when both paths are triggered simultaneously, the help data packet is marked as a high-risk type and the highest push priority is set.

[0047] The above describes the control method for the intelligent nursing bed based on the brain-computer interface in the embodiments of the present invention. The intelligent nursing bed in the embodiments of the present invention will be described below. Please refer to [link / reference]. Figure 4 One embodiment of the intelligent nursing bed in this invention includes: Acquisition module 401 is used to acquire raw EEG signals and extract μ rhythm features and β rhythm features; The decoding module 402 is used to decode the μ rhythm features and β rhythm features into corresponding nursing instructions, and to classify the nursing instructions into high-confidence instructions, medium-confidence instructions and low-confidence instructions; The graded output module 403 is used to directly output the nursing instruction and corresponding action parameters when the nursing instruction is a high-confidence instruction, and to directly execute the nursing action without verifying the physiological signal; when the nursing instruction is a medium-confidence instruction, it calculates the cognitive load level and generates a comprehensive decision result; when the nursing instruction is a low-confidence instruction, it rejects the nursing instruction and does not trigger any nursing action.

[0048] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0049] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause an electronic device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0050] The above-described 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 the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for controlling an intelligent nursing bed based on a brain-computer interface, characterized in that, include: S1: Acquire raw EEG signals and extract μ-rhythm and β-rhythm features; S2: Decode the μ rhythm features and the β rhythm features into corresponding nursing instructions, and classify the nursing instructions into high-confidence instructions, medium-confidence instructions and low-confidence instructions; S3: When the nursing instruction is a high-confidence instruction, the nursing instruction and corresponding action parameters are directly output, and the nursing action is executed directly without physiological signal verification; when the nursing instruction is a medium-confidence instruction, the cognitive load level is calculated and a comprehensive decision result is generated; when the nursing instruction is a low-confidence instruction, the nursing instruction is rejected and no nursing action is triggered.

2. The intelligent nursing bed control method based on brain-computer interface according to claim 1, characterized in that, Step S1 includes: S11: Acquire raw EEG signals and remove electrooculography artifacts from the raw EEG signals to obtain preprocessed EEG signals; S12: Perform a short-time Fourier transform on the preprocessed EEG signal to calculate the μ rhythm features of the μ band and the β rhythm features of the β band.

3. The intelligent nursing bed control method based on brain-computer interface according to claim 1, characterized in that, Step S2 includes: S21: Calculate the probability of getting up, turning over, raising legs, massaging, and seeking emergency help based on the μ rhythm features and the β rhythm features; S22: Find the maximum probability value among the back-raising probability, the turning-over probability, the leg-raising probability, the massage probability, and the emergency help probability, take the instruction type corresponding to the maximum probability value as the nursing instruction, and take the maximum probability value as the confidence level; S23: Compare the confidence level with preset high confidence thresholds and low confidence thresholds. When the confidence level is greater than or equal to the high confidence threshold, determine the nursing instruction as a high confidence instruction. When the confidence level is between the low confidence threshold and the high confidence threshold, determine the nursing instruction as a medium confidence instruction. When the confidence level is less than the low confidence threshold, determine the nursing instruction as a low confidence instruction.

4. The intelligent nursing bed control method based on brain-computer interface according to claim 1, characterized in that, Step S3 includes: S31: When the nursing instruction is a high-confidence instruction, the nursing instruction and corresponding action parameters are directly output, and the physiological signal verification is skipped and the nursing action is executed directly. S32: When the nursing instruction is a moderate confidence instruction, calculate the cognitive load level, and perform a comprehensive analysis of the EEG confidence score, stress verification score, heart rate verification score, and cognitive load verification score based on the nursing instruction to obtain a comprehensive decision result; S33: When the nursing instruction is a low-confidence instruction, rejecting the nursing instruction does not trigger any nursing action.

5. The intelligent nursing bed control method based on brain-computer interface according to claim 4, characterized in that, Step S32 includes: S321: The sensor array collects mattress pressure, the heart rate is collected by the heart rate sensor, and the ratio of the theta wave power to the alpha wave power in the prefrontal cortex EEG signal is calculated to obtain the cognitive load level. S322: Calculate the stress verification score, heart rate verification score, and cognitive load verification score based on the nursing instructions and the cognitive load level, respectively. S323: The EEG confidence score, the stress verification score, the heart rate verification score, and the cognitive load verification score are weighted and summed to obtain a comprehensive decision result.

6. The intelligent nursing bed control method based on brain-computer interface according to claim 5, characterized in that, Step S321 includes: S3211: Collects pressure data of the user's torso, hips, and legs at a preset sampling frequency using a sensor array to obtain mattress pressure; S3212: Measures heart rate using photoplethysmography (PPG) technology via a heart rate sensor; S3213: Extract the first power spectral density of the theta wave band and the second power spectral density of the alpha wave band from the EEG signal of the prefrontal region, calculate the ratio of the first power spectral density to the second power spectral density, and obtain the cognitive load level.

7. The intelligent nursing bed control method based on brain-computer interface according to claim 5, characterized in that, Step S322 includes: S3221: When the nursing instruction is a turning instruction, the pressure centroid coordinates are calculated based on the mattress pressure, and the pressure verification score is determined based on the matching relationship between the pressure centroid coordinates and the turning direction. S3222: When the nursing instruction is a back-raising instruction, extract the pressure data of the buttock area in the mattress pressure, calculate the pressure standard deviation, and determine the pressure verification score based on the pressure standard deviation and the preset stability threshold. S3223: When the nursing instruction is a leg-lifting instruction, extract the pressure data of the leg area in the mattress pressure to verify the leg posture and determine the pressure verification score; S3224: When the nursing instruction is a massage instruction, the heart rate is compared with the preset heart rate range to determine the heart rate verification score, and the cognitive load level is compared with the preset load threshold to determine the cognitive load verification score; S3225: When the nursing instruction is an emergency help instruction, assess the degree of urgency based on the heart rate and the cognitive load level, and determine the heart rate verification score and the cognitive load verification score.

8. The intelligent nursing bed control method based on brain-computer interface according to claim 1, characterized in that, Following step S323, the following is also included: S324: Based on the comprehensive decision-making results and the nursing instructions, respectively drive the electric push rod to adjust the backboard to the first angle, or drive the airbag turning system to perform the turning action according to the inflation sequence, or drive the thigh board electric push rod to adjust the thigh board to the second angle, or drive the vibration massage component to adjust the massage intensity according to the cognitive load level to perform the massage action, or when the nursing instruction is an emergency help instruction, activate the sound and light alarm and push the help data packet to the bound mobile terminal.

9. The intelligent nursing bed control method based on brain-computer interface according to claim 1, characterized in that, Following step S3, the following is also included: S4: Calculate the theta wave power of the original EEG signal in real time, and calculate the ratio of the difference between the theta wave power and the baseline theta wave power to the baseline theta wave power to obtain the surge amplitude; S5: Determine whether the surge is greater than or equal to a preset threshold and whether the duration is greater than or equal to a preset duration. When both conditions are met, an emergency help command is automatically triggered. S6: After triggering the emergency help command, activate the buzzer and LED light array to provide an audible and visual alarm, push a help data packet containing EEG data summary and location information to the bound mobile terminal, and switch all nursing actions to safe mode.

10. An intelligent nursing bed, characterized in that, A method for controlling a brain-computer interface-based intelligent nursing bed as described in any one of claims 1-9, comprising: The acquisition module is used to acquire raw EEG signals and extract μ-rhythm and β-rhythm features; The decoding module is used to decode the μ rhythm features and the β rhythm features into corresponding nursing instructions, and to classify the nursing instructions into high-confidence instructions, medium-confidence instructions and low-confidence instructions; The graded output module is used to directly output the nursing instruction and its corresponding action parameters when the nursing instruction is a high-confidence instruction, and to directly execute the nursing action without verifying the physiological signal; when the nursing instruction is a medium-confidence instruction, it calculates the cognitive load level and generates a comprehensive decision result; when the nursing instruction is a low-confidence instruction, it rejects the nursing instruction and does not trigger any nursing action.

Images