Intelligent precise nerve blood pressure reduction rehabilitation robot

Abstract

The invention belongs to the technical field of rehabilitation medical treatment, and discloses an intelligent precise nerve blood pressure reduction rehabilitation robot which comprises a physiological monitoring module used for collecting electrocardiosignals, respiratory rhythm signals and skeletal electromyographic signals of a patient in real time; the pressure sensing array is used for judging the bedridden state of the patient according to the pressure distribution symmetry and the duration time; the infrared detection unit is used for distinguishing a human body from a static object through infrared reflection waveform characteristics; the REM judgment module is electrically connected with the physiological monitoring module, the pressure sensing array and the infrared detection unit and is used for fusing multi-source signals to analyze an REM stage; the data marking module is electrically connected with the REM judgment module and is used for marking the REM time period in the physiological data and executing right falling processing on the heart rate data in the time period; the nerve regulation and control module is used for dynamically adjusting a stimulation scheme according to the physiological parameters in the non-REM period; the data verification module is used for carrying out afterward backtracking on the marked REM time period; the invention solves the problem of marking and excluding the heart rate data of the REM period.

Description

Technical Field

[0001] This solution belongs to the field of rehabilitation medical technology, specifically involving an intelligent and precise neurohypertensive rehabilitation robot. Background Technology

[0002] Neurohypertensive rehabilitation is a comprehensive rehabilitation method that lowers blood pressure by regulating the function of the autonomic nervous system. Its core lies in balancing the activity of the sympathetic and vagus nerves. Excessive excitation of the sympathetic nervous system can lead to vasoconstriction, increased heart rate, and elevated blood pressure. Neurohypertensive rehabilitation inhibits sympathetic nerve activity through aerobic training and psychological intervention, while simultaneously enhancing vagal tone, thereby relieving small artery spasm and improving vascular compliance.

[0003] During rehabilitation, multiple data points need to be measured to assess the therapeutic effect. One of these is the measurement of heart rate and heart rate variability (HRV), which is used to quantify changes in sympathetic / vagal tone. A resting heart rate ≥80 beats / min indicates excessive sympathetic nerve activation and requires intervention. When measuring heart rate, dynamic electrocardiogram recording or wearable devices (such as smartwatches or chest strap monitors) are generally used to collect data for 24 hours to capture the dynamic balance of sympathetic / vagal nerves under diurnal rhythms. Heart rate monitoring generally requires coverage of three key time periods: (1) the resting baseline period (3:00-7:00 AM), when the vagus nerve is dominant and the heart rate is lowest; (2) the daytime activity period (6:00-8:00 PM), where the sympathetic nerve stress response is assessed through heart rate fluctuations under exercise load; and (3) the nighttime recovery period (8:00-3:00 AM), where the magnitude of heart rate decrease and heart rate variability (HRV) are monitored to assess the recovery efficiency of neural regulation.

[0004] During the resting baseline and nighttime recovery phases of heart rate monitoring, patients are asleep. Sleep consists of cyclical non-rapid eye movement (NREM) and rapid eye movement (REM) phases, with the duration of REM phases gradually increasing as the night progresses, especially in the early morning.

[0005] The REM phase is characterized by the release of acetylcholine from cholinergic neurons in the pontine reticular formation, triggering low-amplitude fast-wave brain activity in the cerebral cortex (approximately a state of wakefulness). Simultaneously, transient activation of the sympathetic nervous system mediated by the locus coeruleus induces increased heart rate fluctuations (typically 60 beats / min during NREM deep sleep, which can surge to 80-100 beats / min during REM), accompanied by respiratory rhythm disturbances and a transient increase in blood pressure. This physiological fluctuation stems from the dissociation of neural regulation of somatic muscle tone inhibition (spinal motor neuron hyperpolarization) and extraocular muscle activation (rapid eye movements) in the central nervous system; essentially, it is a normal physiological process of memory integration and brain metabolic regulation. However, if ECG recording equipment detects data during this period, it may be misinterpreted as pathological sympathetic hyperactivity, resulting in abnormal heart rate assessment. Therefore, it is necessary to label and exclude heart rate data from the REM phase. Summary of the Invention

[0006] The purpose of this solution is to provide an intelligent and precise neuro-hypertensive rehabilitation robot to label and exclude heart rate data during REM periods.

[0007] To achieve the above objectives, this solution provides an intelligent and precise neuro-hypertensive rehabilitation robot, comprising:

[0008] The physiological monitoring module is used to collect patients' electrocardiogram signals, respiratory rhythm signals, and skeletal muscle electrocardiogram signals in real time;

[0009] A pressure sensor array is installed in the head, torso, and foot areas of the hospital bed to determine the patient's bed rest status by measuring the symmetry of pressure distribution and duration.

[0010] An infrared detection unit is used to distinguish between human bodies and static objects by the characteristics of infrared reflection waveforms.

[0011] The REM determination module is electrically connected to the physiological monitoring module, pressure sensor array, and infrared detection unit. It is used to fuse multi-source signals to analyze the REM stage and is determined by a sudden increase in the standard deviation of heart rate variability exceeding 50 milliseconds, a coefficient of variation of respiratory rhythm disorder index not less than 15%, and a duration of skeletal muscle electrophysiological inhibition exceeding 60 seconds.

[0012] The data labeling module, electrically connected to the REM determination module, is used to label REM periods in physiological data and perform weight reduction processing on the heart rate data of those periods;

[0013] The neuromodulation module is used to dynamically adjust the stimulation protocol based on physiological parameters during non-REM phases.

[0014] The data verification module is used to retrospectively analyze the marked REM periods and compare them with the characteristics of the synchronously acquired multichannel physiological data. If the mislabeling rate exceeds 5%, parameter correction is triggered.

[0015] The principle and effectiveness of this solution are as follows: A physiological monitoring module collects the patient's electrocardiogram (ECG), respiratory rhythm, and skeletal muscle electromyography (EMG) signals. Combined with a pressure sensor array, it monitors the symmetry and duration of pressure distribution in the head, trunk, and foot areas of the bed, determining the patient's bedridden status and eliminating interference from static heavy objects or changes in body position. Simultaneously, an infrared detection unit distinguishes between the human body and static objects based on infrared reflection waveform characteristics, avoiding the influence of external objects on sleep state analysis. Building upon this, the REM (Rapid Eye Movement) determination module integrates multi-source signals, identifying the REM sleep stage based on characteristics such as a sudden increase in the standard deviation of heart rate variability exceeding 50 milliseconds (reflecting increased heart rate fluctuations due to transient sympathetic nerve activation), a respiratory rhythm disorder index coefficient of variation not less than 15% (reflecting the dissociation characteristics of respiratory regulation during REM sleep), and a skeletal muscle EMG inhibition duration exceeding 60 seconds (consistent with the physiological phenomenon of somatic muscle tone inhibition during REM sleep). The data labeling module, based on the REM assessment results, marks the REM period in the physiological data and performs downweighting on the heart rate data during that period to reduce the risk of physiological heart rate fluctuations during REM being misjudged as pathological sympathetic hyperactivity. The neuromodulation module adjusts the stimulation protocol based on physiological parameters (such as heart rate, heart rate variability, and diurnal variation of systolic blood pressure) during non-REM periods (when heart rate data more accurately reflects the state of autonomic nervous regulation) to intervene in the sympathetic-vagal balance. The data verification module performs retrospective analysis of the marked REM periods by comparing the characteristics of synchronously acquired multichannel physiological data. If the mislabeling rate exceeds 5%, parameter correction is triggered, forming a feedback loop of "detection-judgment-labeling-intervention-verification" to optimize the system's accuracy in recognizing the REM phase.

[0016] Furthermore, the pressure sensing array determines that the pressure source is a human body by the pressure distribution symmetry being greater than 0.8 and the duration being no less than 10 seconds, and the time synchronization error between the pressure signal and the electrocardiogram signal is no more than 2 seconds to trigger confirmation of the presence of a human body.

[0017] The principle and effect of this solution are as follows: By setting a head pressure sensor with a range of 0 to 15 kg in the occipital region, the system captures the head micro-movement frequency and pressure waveform from 0.5 to 3 Hz, utilizing the frequency characteristics of natural head micro-movements during sleep to distinguish between the human body and static heavy objects. A matrix of pressure sensors in the torso region calculates the pressure center of gravity shift; when the shift is less than 10 cm / min, it is determined to be static heavy object pressure and eliminated, avoiding interference from non-human factors. Pressure sensors with a range of 0 to 10 kg are set in the foot region; when a sudden drop in foot pressure exceeding 50% is detected while torso pressure remains constant, it is determined to be a postural change rather than a REM stage characteristic. Thus, through synchronous monitoring of sensors in multiple regions (head, torso, and feet), normal body movement, postural changes, and foreign object interference can be identified, providing the REM assessment module with clean pressure data after interference removal, ensuring the accuracy of sleep state analysis based on pressure signals and subsequent heart rate data labeling.

[0018] The principle of the pressure sensing array in this scheme is as follows: by setting judgment conditions that the symmetry of the pressure distribution is greater than 0.8 and the duration is not less than 10 seconds, the scheme utilizes the natural symmetry and stable and continuous characteristics of the pressure distribution when a person is lying down to distinguish between human pressure and interference from static heavy objects or debris. Simultaneously, the time synchronization error between the pressure signal and the electrocardiogram signal is required to be no more than 2 seconds. Based on the natural correlation between human physiological signals (such as slight changes in body pressure accompanying heartbeat), the scheme triggers confirmation of human presence, eliminating pressure signal interference caused by non-physiological factors. This provides reliable basic data on the bedridden state for REM stage analysis, avoiding REM judgment deviations caused by misjudging human presence or interference from foreign objects, and ensuring the accuracy of subsequent heart rate data labeling and neural modulation.

[0019] Furthermore, the pressure sensing array includes:

[0020] A head pressure sensor with a range of 0 to 15 kg is placed in the occipital region to detect pressure waveforms and the micro-motion frequency of the human head, wherein the micro-motion frequency range is 0.5 to 3 Hz.

[0021] A trunk matrix pressure sensor array, located in the trunk area, eliminates static heavy object pressure by offsetting the center of gravity of pressure by less than 10 cm per minute.

[0022] The foot pressure sensor, with a range of 0 to 10 kg, is located in the foot area. If the foot pressure suddenly drops by more than 50% while the torso pressure remains unchanged, it is determined to be a change in body position.

[0023] The principle and effect of this scheme are as follows: A head pressure sensor with a range of 0 to 15 kg is placed in the occipital region to detect pressure waveforms and the micro-movement frequency of the human head from 0.5 to 3 Hz. By utilizing the physiological characteristic frequency range of natural micro-movement of the head during sleep, the source of head pressure can be distinguished as the human body rather than a static heavy object. A matrix pressure sensor group in the torso region calculates the pressure center of gravity offset. When the offset is less than 10 cm per minute, it is determined to be static heavy object pressure and excluded, avoiding interference from non-human factors in determining the bed rest status. A pressure sensor with a range of 0 to 10 kg in the foot region detects a sudden drop in foot pressure exceeding 50% while the torso pressure remains unchanged, which is determined to be a change in body position rather than the electromyographic inhibition characteristics of the REM stage. Thus, through the coordinated monitoring of pressure distribution, micro-movement frequency, and center of gravity offset by multiple sensors in the head, torso, and feet, interference factors such as static heavy object pressure and changes in body position can be eliminated.

[0024] Furthermore, the infrared detection unit includes an adjustable-height infrared transmitter and receiver, and uses infrared reflection waveform characteristics with a wavelength of 850 nanometers and a pulse width of no more than 10 milliseconds to distinguish between human bodies and static objects; when the patient is lying down, the receiver signal strength is not lower than 80% of the initial value; when the infrared signal strength is lower than 80% of the initial value and the pressure distribution symmetry is less than 0.6 or the duration is less than 5 seconds, it is determined to be foreign interference and the REM judgment function is disabled, and when the interference is removed, the REM judgment function is automatically restored and the data during the interference period is marked.

[0025] The principle and effect of this solution are as follows: An infrared signal with a wavelength of 850 nanometers and a pulse width of no more than 10 milliseconds is emitted through an adjustable-height infrared transmitter and receiver. Utilizing the difference in infrared light reflection waveforms between the human body and static objects (human body reflection signals exhibit dynamic biological tissue characteristics), the receiver signal strength is maintained at at least 80% of the initial value when the patient is lying down to confirm the presence of a human body. When the infrared signal strength is lower than 80% of the initial value and the pressure distribution symmetry is less than 0.6 or the duration is less than 5 seconds, it is determined to be interference from foreign objects (because static foreign objects cannot maintain a stable and symmetrical pressure distribution and sufficiently strong infrared reflection). In this case, the REM assessment function is disabled to avoid misjudgment. When the interference is resolved (e.g., signal strength and pressure symmetry are restored after removing foreign objects), the REM assessment function is automatically restored, and the data during the interference period is marked. Thus, through verification of the infrared signal and pressure data, the REM stage analysis is based solely on real human body data, avoiding misjudgment of REM and mismarking of heart rate data caused by environmental foreign objects.

[0026] Furthermore, the REM determination module includes:

[0027] A multimodal data fusion engine aligns the time series of pressure sensor signals, electrocardiogram signals, electromyogram signals, and infrared detection signals;

[0028] The sleep state activation unit initiates the REM analysis process when the head, torso, and foot pressure sensors all detect that the torso pressure ratio is not less than 50% and the infrared signal is not blocked.

[0029] The REM feature analyzer determines the REM stage based on a sudden increase in the standard deviation of heart rate variability exceeding 50 milliseconds, a coefficient of variation of the respiratory rhythm disorder index not less than 15%, and an electromyographic inhibition duration exceeding 60 seconds.

[0030] The principle and effect of this solution are as follows: A multimodal data fusion engine aligns the time series of pressure sensor signals, ECG signals, EMG signals, and infrared detection signals; the sleep state activation unit sets trigger conditions, initiating the REM analysis process only when the head, trunk, and foot pressure sensors all detect a trunk pressure ratio of no less than 50% (indicating the body is in a stable lying position) and the infrared signal is not obstructed (excluding interference from debris), thus avoiding invalid analysis in non-sleep states or under interfering environments; the REM feature analyzer identifies the REM sleep stage based on multi-dimensional criteria, including a sudden increase in the standard deviation of heart rate variability exceeding 50 milliseconds (reflecting heart rate fluctuations caused by transient sympathetic nerve activation during REM), a respiratory rhythm disorder index coefficient of variation of no less than 15% (reflecting the dissociation of respiratory regulation mechanisms during REM), and an EMG inhibition duration exceeding 60 seconds (consistent with the physiological characteristics of somatic muscle tone inhibition during REM). This distinguishes between physiological heart rate fluctuations during REM and pathological sympathetic nerve overactivity, avoiding abnormal heart rate assessments due to misjudgment caused by a single data dimension.

[0031] Furthermore, the data tagging module includes:

[0032] The timestamp traceability unit uses the IEEE 1588 protocol to align the timeline of multi-source data, with a time error of no more than ±5 milliseconds;

[0033] The dynamic weighting processor assigns a confidence weight of 0.3 to 0.5 to the heart rate data during the REM marking period. If a decrease in pressure distribution symmetry of more than 30% is detected within 10 minutes after marking, the marking period is automatically corrected.

[0034] The principle and effect of this scheme are as follows: By using the IEEE 1588 protocol to align the time axis of multi-source data through the timestamp tracing unit, the time error of ECG, respiration, EMG, pressure, and infrared signals is controlled within ±5 milliseconds, thus unifying the time reference for REM period marking. The dynamic weighting processor assigns a confidence weight of 0.3 to 0.5 to the heart rate data of the period marked by the REM judgment module (the weight of non-REM period data is 1.0). The weight difference is used to reduce the interference of physiological heart rate fluctuations during REM period on the overall assessment. At the same time, it is set to automatically correct the marked period when the pressure distribution symmetry decreases by more than 30% within 10 minutes after marking. By identifying changes in body position or interference events through dynamic changes in pressure signals, it corrects mismarking that may be caused by motion artifacts. This filters and marks the heart rate data of REM period, avoids misjudging normal heart rate fluctuations during REM period as pathological sympathetic nerve overactivation, and improves the accuracy of assessment indicators such as heart rate variability in reflecting the state of autonomic nervous system function.

[0035] Furthermore, the neural modulation module includes:

[0036] The transcranial magnetic stimulation unit is used to dynamically adjust the stimulation intensity based on the standard deviation of heart rate variability when the minimum heart rate during the non-REM phase is less than 60 beats per minute: when the standard deviation of heart rate variability is <50ms, the intensity is increased to 90% of the exercise threshold; otherwise, stimulation at 80% of the exercise threshold and a frequency of 10 Hz is used.

[0037] The ultrasonic stimulation unit is used to adjust the power density according to the diurnal difference in systolic blood pressure exceeding 20 mmHg: 250 mW / cm² is activated when the difference is >30 mmHg. 2 Power, otherwise use 200mW / cm 2 Co-stimulation with power and frequency of 1 MHz;

[0038] The emergency braking unit is used to trigger a 4 Hz theta wave resonant magnetic field and a 187 Hz bone conduction sound wave in cases of orthostatic hypotension.

[0039] The principle and effect of this scheme are as follows: When the minimum heart rate during the non-REM phase is below 60 beats per minute, the transcranial magnetic stimulation unit dynamically adjusts the stimulation intensity based on the standard deviation of heart rate variability (increasing to 90% of the exercise threshold when the standard deviation of heart rate variability is <50ms, otherwise using 80% of the exercise threshold and a frequency of 10Hz). This utilizes the regulatory effect of magnetic stimulation on cortical nerves to enhance vagal nerve tone and thus reduce heart rate. When the diurnal difference in systolic blood pressure exceeds 20 mmHg, the ultrasound stimulation unit adjusts the power density according to the difference (activating at 250mW / cm² when the difference is >30mmHg). 2 Power, otherwise use 200mW / cm 2 The device utilizes sound wave energy (power, 1 MHz frequency) to improve vascular endothelial function and compliance, thereby balancing sympathetic nerve activity. Upon detecting orthostatic hypotension, the emergency braking unit triggers a 4 Hz theta wave resonant magnetic field and a 187 Hz bone conduction sound wave. Through a neural feedback mechanism, it rapidly regulates autonomic nerve function to raise blood pressure. This adjustment, based on real physiological parameters during the non-REM phase (with REM phase interference removed), intervenes in the sympathetic-vagal nerve imbalance, improving the therapeutic effect of neurohypertensive rehabilitation while reducing the risk of over- or under-intervention due to misinterpretation of REM phase data.

[0040] Furthermore, it also includes a human body pressure distribution feature library. The data verification module reduces the misjudgment rate due to foreign object interference by comparing the cosine similarity between real-time pressure data and historical normal samples.

[0041] The principle and effect of this solution are as follows: By establishing a human pressure distribution feature library, the pressure distribution vector features of historical normal samples are stored. In real-time monitoring, the data verification module uses a cosine similarity algorithm to compare the data collected by the current pressure sensor array with the samples in the feature library dimension by dimension, and calculates the cosine value of the angle between the vector spaces (the threshold is set to >0.85). When the similarity is higher than the threshold, it is determined to be a normal human pressure distribution. If it is lower than the threshold, it indicates that there may be foreign interference.

[0042] Furthermore, the time series alignment error of the multimodal data fusion engine does not exceed ±100 milliseconds when fusing pressure sensor signals, electrocardiogram signals, electromyogram signals, and infrared detection signals.

[0043] The principle and effect of this solution are as follows: real-time alignment of the time series of pressure sensor signals, electrocardiogram (ECG) signals, electromyogram (EMG) signals, and infrared detection signals, controlling the time deviation of different signals within ±100 milliseconds, ensuring that the data of each modality correspond in the time dimension (e.g., the peak value of the ECG R wave is strictly synchronized with the synchronous onset point of EMG inhibition and the pressure fluctuation time). This enables the REM determination module to perform joint analysis based on spatiotemporally consistent multi-source data (e.g., synchronously capturing the instantaneous correlation between sudden increases in heart rate variability, respiratory rhythm disorders, and EMG inhibition), avoiding feature mismatch caused by time misalignment (e.g., incorrectly associating non-REM heart rate fluctuations with REM EMG inhibition), thereby improving the accuracy of REM phase determination and reducing the risk of misjudgment caused by data asynchrony.

[0044] Furthermore, the determination of the duration of inhibition of skeletal muscle electrical signals is based on the standard that the amplitude of electrical signals of large muscle groups such as the tibialis anterior and biceps brachii is less than 5 microvolts.

[0045] The principle and effect of this scheme are as follows: Based on the physiological characteristics of the central nervous system's inhibition of somatic muscle tone during the REM phase (significantly weakened skeletal muscle activity except for extraocular muscles), the scheme uses an electromyographic amplitude of less than 5 microvolts for large muscle groups such as the tibialis anterior and biceps brachii as the criterion for electromyographic inhibition. By monitoring the intensity of electromyographic signals of large muscle groups, when the amplitude is detected to be consistently below this threshold, it is determined that the electromyographic inhibition state has entered the REM phase. This not only conforms to the physiological mechanism of hyperpolarization of somatic motor neurons during the REM phase, but also verifies the effectiveness of the 5 microvolt threshold in distinguishing between physiological electromyographic inhibition and normal electromyographic activity in the non-REM phase through clinical data. This provides a reliable electromyographic characteristic basis for the REM determination module, avoids misjudging low electromyographic activity (such as resting state) or interference signals in the non-REM phase as REM inhibition, improves the accuracy of REM phase identification, and ensures that the data labeling module can accurately remove heart rate data during the REM period. Attached Figure Description

[0046] Figure 1This is a schematic diagram of an intelligent and precise neuro-hypertensive rehabilitation robot according to the present invention. Detailed Implementation

[0047] The following will describe the concept and technical effects of the present invention clearly and completely with reference to embodiments, so as to fully understand the purpose, features and effects of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are all within the scope of protection of the present invention.

[0048] Example:

[0049] like Figure 1 As shown, this embodiment provides an intelligent and precise neuro-hypertensive rehabilitation robot, including:

[0050] Physiological monitoring module: ECG signal acquisition uses medical-grade Ag / AgCl electrodes (impedance ≤10kΩ), sampling frequency 500Hz, configured with an IIR bandpass filter (0.5-40Hz, stopband attenuation ≥40dB) and db4 wavelet noise reduction, achieving R-wave detection accuracy ≥99.5%. Respiratory rhythm monitoring uses a thoracic and abdominal impedance breathing band, sampling frequency 100Hz. The respiratory waveform envelope is extracted using Hilbert transform, and the coefficient of variation (CV) of the difference between adjacent respiratory cycles is calculated, with a normal range of 5%-15%. Skeletal muscle electromyography (EMG) signal acquisition uses surface EMG electrodes (10mm diameter, 20mm center-to-center distance), covering large muscle groups such as the tibialis anterior and biceps brachii, sampling frequency 1000Hz. After 50Hz power frequency notch filtering (attenuation ≥30dB), the EMG amplitude is calculated using the root mean square (RMS) algorithm, with <5μV for ≥60 seconds as the criterion for REM phase EMG inhibition. All signals are converted by a 24-bit ADC (signal-to-noise ratio ≥110dB) and transmitted to the central processing unit via the SPI interface, with a time synchronization error ≤±1μs (using the IEEE 1588v2 precision clock protocol).

[0051] Pressure sensor array: The pressure sensor array consists of three sub-arrays: head (3×3 matrix), torso (10×10 matrix), and feet (5×5 matrix). It employs piezoresistive sensors (range 0-15kg, linearity ±0.1%FS, hysteresis <0.05%FS) with a spacing of 10cm. Pressure distribution symmetry is calculated (formula: S = 1 - |P|). l -P r | / (P l +P r ), P l P rThe pressure is calculated as the average pressure on both sides, and an S ≥ 0.8 value is considered a valid human body distribution. Simultaneously, the pressure center of gravity shift is monitored, and the center of gravity trajectory is predicted using Kalman filtering (state transition matrix Q = 0.01I, measurement noise matrix R = 0.1I). A body position change marker is triggered when the shift velocity > 10 cm / min. The foot pressure sensor uses a threshold drop detection algorithm (drop rate > 2 kg / s and duration > 2 seconds) to identify a leg-raising motion. The array is calibrated hourly, and cross-checking between adjacent sensors (triggered three repeated checks when the difference > 30% of the baseline value; if still abnormal, the fault point is marked and KNN interpolation (k = 3) is used) ensures accurate pressure data.

[0052] Infrared Detection Unit: The infrared detection unit consists of an 850nm wavelength transmitting and receiving pair (transmitting power 50mW, receiving sensitivity 0.5A / W), mounted on an electric push rod bracket (height adjustment range 30-80cm, step accuracy 0.5°, positioning error <±0.1°). The system compensates for light interference through an ambient light sensor (measurement range 0-100klux, accuracy ±3%). When the ambient light intensity is >200lux, the transmitting power is automatically increased by 10% and a narrowband filter (bandwidth ±10nm, center wavelength 850nm) is activated. Human presence detection uses a dynamic threshold algorithm: the baseline signal strength is the average value of samples taken continuously for 10 minutes under interference-free conditions. When the real-time signal is ≥80% of the baseline value and the duration is >5 seconds, it is considered a valid human signal. If the signal strength is <70%, intelligent obstacle avoidance adjustment is triggered, and the electric push rod scans the optimal detection position using an Archimedean spiral trajectory (step size 2cm, angle increment 5°) until the signal recovers to ≥85%. The temperature compensation algorithm dynamically adjusts the threshold based on the ambient temperature (within the range of -20℃ to 50℃ and a measurement accuracy of ±0.5℃) (the threshold increases by 2% for every 5℃ increase) to ensure system stability.

[0053] REM determination module: 1. Initial screening: When the pressure sensors of the head, torso, and feet detect that the torso pressure accounts for ≥50% and the infrared signal is normal (signal strength ≥80% of the baseline value), REM analysis is started. 2. Feature extraction: (1) HRV analysis: Calculate the standard deviation of adjacent NN intervals (SDNN), and use a moving window (window size 30 seconds, sliding step size 5 seconds) to calculate the burst rate. A burst > 50ms and an absolute value > 70ms is judged as a REM feature; (2) Respiratory disturbance index: Calculate the CV of the difference between adjacent respiratory cycles. The CV is required to be ≥ 15% and the respiratory rate fluctuation is > 5 breaths / min; (3) Electromyographic inhibition: The RMS amplitude of large muscle groups < 5μV lasts for ≥ 60 seconds, and the β / α power ratio is < 0.8. 3. Data validation: The 30-minute window data is validated again using an LSTM network. The input features include HRV time-domain indicators (SDNN, RMSSD, pNN50), respiratory rate variability (standard deviation, kurtosis) and electromyographic spectrum features (β / α power ratio, median frequency). SDNN: Standard deviation of normal NN intervals, reflecting the overall variability of HRV; RMSSD: Root mean square of the difference between adjacent NN intervals, mainly reflecting the high-frequency components of HRV. pNN50: The percentage of adjacent NN intervals with a difference of >50ms, reflecting the high-frequency component of HRV.

[0054] Data Tagging Module: Data tagging uses dual timestamps: the time of physiological event occurrence (accuracy ±1ms) and system processing time (synchronized to UTC time using NTP protocol, error <±100μs). Data from the REM-marked period is weighted hierarchically: HRV data confidence weight is set to 0.3-0.5 (1.0 for non-REM periods), and respiratory rate data weight is set to 0.7-0.9. When a change in body position (pressure symmetry decrease >30%) or a decrease in signal quality (SNR <10dB) is detected, the data repair process is initiated: 1. Missing data is completed using cubic spline interpolation (when the missing rate is <20%), with an interpolation error <±5%. 2. If the missing rate is >20%, the data for that period is marked as unusable, and manual review is triggered. 3. All correction operations generate blockchain evidence (SHA-256 hash value), which is packaged and uploaded to a private blockchain every hour to ensure data traceability. 4. The system calculates the signal-to-noise ratio (SNR) of each channel's data. When the SNR is less than 10dB, the weight of the data in that channel is automatically reduced (e.g., when the ECG SNR is less than 8dB, the weight of the HRV parameter is reduced to 0.2).

[0055] The neuromodulation module implements personalized interventions based on non-REM physiological parameters.

[0056] 1. Transcranial magnetic stimulation (TMS):

[0057] (1) When the HRV standard deviation is <50ms, an incremental dose scheme is adopted: the initial intensity is 80% of the motion threshold (MT), and it is increased by 2% every 5 minutes until the induced α wave power enhancement is >30% (by EEG monitoring, α wave frequency band 8-13Hz), and the maximum intensity is ≤95%MT; (2) The frequency is 10Hz, the pulse width is 200μs, the double coil figure-eight design is adopted (outer diameter 70mm), the focusing depth is 4cm, and the electric field intensity distribution uniformity is >90%; (3) The stimulation sequence adopts the repetitive pulse mode, with 10 pulses per series and a series interval of 500ms.

[0058] 2. Ultrasound stimulation:

[0059] (1) When the diurnal variation in systolic blood pressure is >20 mmHg, activate 200-250 mW / cm 2 Power (use 250mW / cm when the difference is >30mmHg) 2 (2) Frequency 1MHz, duty cycle 50%, use a circular probe (diameter 2cm), sound beam diffusion angle <15°, coupling agent acoustic impedance matched human tissue (1.5MRayl); (3) Each treatment lasts 5 minutes, repeated once every hour.

[0060] 3. Emergency braking mechanism:

[0061] (1) When orthostatic hypotension is detected (systolic blood pressure drops by more than 20 mmHg and lasts for more than 15 seconds), a 4 Hz theta wave TMS (80% MT) and a 187 Hz bone conduction sound wave (intensity 60 dB) are triggered simultaneously for 30 seconds; (2) If blood pressure does not recover by more than 10% within 30 seconds, the alarm system is automatically activated (audio-visual alarm + SMS notification to medical staff); (3) All intervention parameters are dynamically updated every 5 minutes to form a closed-loop feedback regulation with regulation gain Kp = 0.5, Ki = 0.1, and Kd = 0.2.

[0062] Data validation module:

[0063] 1. Intra-module self-verification:

[0064] (1) The pressure sensor is zero-point calibrated every hour (the deviation between the output value and the reference value when no load is < ±0.1%FS), and the zero-point drift is < ±0.01%FS / ℃; (2) The ECG signal R wave detection adopts dual algorithm cross-validation (Pan-Tompkins and Christov algorithm), and a re-detection is triggered when the inconsistency rate is >1%; (3) The electromyography signal is verified by spectrum analysis method, and the power ratio of the β wave band (13-30Hz) is determined to be a valid signal when it is >80%.

[0065] 2. Cross-module cross-validation: (1) Infrared detection and pressure distribution jointly confirm the presence of the human body: when the infrared signal is normal but the pressure symmetry is <0.6, it is judged as foreign interference; (2) Validation of the synchronicity between electromyographic inhibition and HRV changes: the time difference between the onset of electromyographic inhibition in the REM period and the sudden increase in HRV should be ≤10 seconds, otherwise a suspicious period is marked; (3) Validation of the correlation between respiratory rate and chest wall fluctuation pressure: when the correlation coefficient is >0.8, it is judged as a valid respiratory signal.

[0066] 3. Historical data pattern matching: (1) Establish a pressure distribution feature library containing 1000 samples, and calculate the cosine similarity between real-time data and samples in the library dimension by dimension (the feature vector contains 16 parameters such as pressure mean, standard deviation, skewness, and kurtosis); (2) When the similarity is <0.85, update the feature library weekly (add 10 valid samples); (3) Use One-Class SVM to detect abnormal patterns (kernel function is RBF, γ=0.1, ν=0.1), and issue an early warning when Mahalanobis distance is >3σ.

[0067] The above descriptions are merely embodiments of the present invention, and common knowledge regarding specific structures and characteristics is not elaborated upon here. It should be noted that those skilled in the art can make various modifications and improvements without departing from the structure of the present invention, and these should also be considered within the scope of protection of the present invention. These modifications and improvements will not affect the effectiveness of the present invention or the practicality of the patent. The scope of protection claimed in this application should be determined by the content of its claims, and the specific embodiments described in the specification can be used to interpret the content of the claims.

Claims

1. An intelligent and precise neuro-hypertensive rehabilitation robot, characterized in that, include: The physiological monitoring module is used to collect patients' electrocardiogram signals, respiratory rhythm signals, and skeletal muscle electrocardiogram signals in real time; A pressure sensor array is installed in the head, torso, and foot areas of the hospital bed to determine the patient's bed rest status by measuring the symmetry of pressure distribution and duration. An infrared detection unit is used to distinguish between human bodies and static objects by the characteristics of infrared reflection waveforms. The REM determination module is electrically connected to the physiological monitoring module, pressure sensor array, and infrared detection unit. It is used to fuse multi-source signals to analyze the REM stage and is determined by a sudden increase in the standard deviation of heart rate variability exceeding 50 milliseconds, a coefficient of variation of respiratory rhythm disorder index not less than 15%, and a duration of skeletal muscle electrophysiological inhibition exceeding 60 seconds. The data labeling module, electrically connected to the REM determination module, is used to label REM periods in physiological data and perform weight reduction processing on the heart rate data of those periods; The neuromodulation module is used to dynamically adjust the stimulation protocol based on physiological parameters during non-REM periods; The data verification module is used to retrospectively analyze the marked REM periods and compare them with the characteristics of the synchronously acquired multichannel physiological data. If the mislabeling rate exceeds 5%, parameter correction is triggered.

2. The intelligent and precise neuro-hypertensive rehabilitation robot according to claim 1, characterized in that: The pressure sensing array determines that the pressure source is a human body by the pressure distribution symmetry being greater than 0.8 and the duration being no less than 10 seconds, and the time synchronization error between the pressure signal and the electrocardiogram signal is no more than 2 seconds to trigger confirmation of the presence of a human body.

3. The intelligent and precise neuro-hypertensive rehabilitation robot according to claim 2, characterized in that: The pressure sensing array includes: A head pressure sensor with a range of 0 to 15 kg is placed in the occipital region to detect pressure waveforms and the micro-motion frequency of the human head, wherein the micro-motion frequency range is 0.5 to 3 Hz. A trunk matrix pressure sensor array, located in the trunk area, eliminates static heavy object pressure by offsetting the center of gravity of pressure by less than 10 cm per minute. The foot pressure sensor, with a range of 0 to 10 kg, is located in the foot area. If the foot pressure suddenly drops by more than 50% while the torso pressure remains unchanged, it is determined to be a change in body position.

4. The intelligent and precise neuro-hypertensive rehabilitation robot according to claim 1, characterized in that: The infrared detection unit includes an adjustable-height infrared transmitter and receiver. It uses infrared reflection waveform characteristics with a wavelength of 850 nanometers and a pulse width of no more than 10 milliseconds to distinguish between human bodies and static objects. When the patient is lying down, the receiver signal strength is not lower than 80% of the initial value. When the infrared signal strength is lower than 80% of the initial value and the pressure distribution symmetry is less than 0.6 or the duration is less than 5 seconds, it is determined to be foreign interference and the REM judgment function is disabled. When the interference is removed, the REM judgment function is automatically restored and the data during the interference period is marked.

5. The intelligent and precise neuro-hypertensive rehabilitation robot according to claim 1, characterized in that: The REM determination module includes: A multimodal data fusion engine aligns the time series of pressure sensor signals, electrocardiogram signals, electromyogram signals, and infrared detection signals; The sleep state activation unit initiates the REM analysis process when the head, torso, and foot pressure sensors all detect that the torso pressure ratio is not less than 50% and the infrared signal is not blocked. The REM feature analyzer determines the REM stage based on a sudden increase in the standard deviation of heart rate variability exceeding 50 milliseconds, a coefficient of variation of the respiratory rhythm disorder index not less than 15%, and an electromyographic inhibition duration exceeding 60 seconds.

6. The intelligent and precise neuro-hypertensive rehabilitation robot according to claim 1, characterized in that: The data tagging module includes: The timestamp traceability unit uses the IEEE 1588 protocol to align the timeline of multi-source data, with a time error of no more than ±5 milliseconds; The dynamic weighting processor assigns a confidence weight of 0.3 to 0.5 to the heart rate data during the REM marking period. If a decrease in pressure distribution symmetry of more than 30% is detected within 10 minutes after marking, the marking period is automatically corrected.

7. The intelligent and precise neuro-hypertensive rehabilitation robot according to claim 1, characterized in that: The neural modulation module includes: The transcranial magnetic stimulation unit is used to dynamically adjust the stimulation intensity based on the standard deviation of heart rate variability when the minimum heart rate during the non-REM phase is less than 60 beats per minute: when the standard deviation of heart rate variability is <50ms, the intensity is increased to 90% of the exercise threshold; otherwise, stimulation at 80% of the exercise threshold and a frequency of 10 Hz is used. The ultrasonic stimulation unit is used to adjust the power density according to the diurnal difference in systolic blood pressure exceeding 20 mmHg: 250 mW / cm² is activated when the difference is >30 mmHg. 2 Power, otherwise use 200mW / cm 2 Co-stimulation with power and frequency of 1 MHz; The emergency braking unit is used to trigger a 4 Hz theta wave resonant magnetic field and a 187 Hz bone conduction sound wave in cases of orthostatic hypotension.

8. The intelligent and precise neuro-hypertensive rehabilitation robot according to claim 1, characterized in that: It also includes a human body pressure distribution feature library. The data verification module reduces the misjudgment rate due to foreign object interference by comparing the cosine similarity between real-time pressure data and historical normal samples.

9. The intelligent and precise neuro-hypertensive rehabilitation assessment module robot according to claim 5, characterized in that: The multimodal data fusion engine achieves a time series alignment error of no more than ±100 milliseconds when fusing pressure sensor signals, electrocardiogram signals, electromyogram signals, and infrared detection signals.

10. The intelligent and precise neuro-hypertensive rehabilitation robot according to claim 1, characterized in that: The determination of the duration of inhibition of skeletal muscle electrical signals is based on the standard that the amplitude of electrical signals of large muscle groups such as the tibialis anterior and biceps brachii is less than 5 microvolts.

Images