A non-contact blood pressure detection method and system based on MIMOUWB

Abstract

The application provides a kind of non-contact blood pressure detection method and system based on UWB, through the time division duplex transmission ultra-wideband impulse signal of double transmitting antenna, through four-way receiving antenna synchronous acquisition echo signal reflected by human body to be measured part, output four-way synchronous baseband I / Q signal;Four-way synchronous baseband I / Q signal is carried out channel consistency calibration, direct current, static clutter suppression, differential motion compensation processing, and the multi-channel echo signal after pre-processing is obtained;The multi-channel echo signal after pre-processing is positioned to be measured blood vessel area by spatial spectrum estimation, and the continuous phase change sequence is extracted by multi-channel joint phase unwrapping, and it is converted into the sub-millimeter level micro-displacement pulse wave timing signal corresponding to human blood vessel pulsation, and multi-channel pulse wave spatial feature parameter is extracted;Based on the pre-calibrated multi-channel feature-blood pressure mapping model, the systolic pressure and diastolic pressure values of human body are calculated according to the multi-channel pulse wave spatial feature parameter;Output and store blood pressure detection result.

Description

Technical Field

[0001] This invention relates to the field of ultra-wideband technology, and in particular to a non-contact blood pressure detection method and system based on MIMO UWB. Background Technology

[0002] Blood pressure is one of the core physiological parameters for assessing human cardiovascular health. Real-time and accurate blood pressure monitoring is of great significance for early screening of hypertension, chronic disease management, and early warning of cardiovascular diseases.

[0003] In existing technologies, mainstream blood pressure measurement methods are divided into two categories: invasive and non-invasive. Invasive methods offer the highest accuracy but are invasive procedures, suitable only for hospital intensive care settings and unsuitable for routine monitoring. Among non-invasive methods, the most widely used is the cuff-type electronic blood pressure monitor, which measures blood pressure based on oscillometric or Korotkoff sound methods. However, the measurement process requires inflating the cuff, causing discomfort to the user and preventing continuous, dynamic blood pressure monitoring. Furthermore, the tightness of the cuff and the user's posture can affect measurement accuracy.

[0004] In recent years, non-contact physiological parameter detection technology has developed rapidly, among which ultra-wideband (UWB) radar technology has been widely used in non-contact heart rate and respiration detection. UWB radar emits nanosecond-level ultra-wideband impulse pulse signals, which have the advantages of strong penetration, low power consumption, high ranging accuracy, and strong anti-interference ability, making it very suitable for non-contact detection of micro-displacements on the human body surface.

[0005] Therefore, it is necessary to provide a non-contact blood pressure detection method and system based on MIMO UWB to effectively solve the above problems. Summary of the Invention

[0006] This invention addresses the problems and shortcomings of existing technologies by providing a non-contact blood pressure detection method and system based on MIMO UWB, which achieves high-precision micro-displacement monitoring through the processing of multi-channel combined signals.

[0007] This invention provides a non-contact blood pressure detection method based on MIMO UWB, the method comprising: Ultra-wideband impulse pulse signals are transmitted in time-division duplex mode using dual transmitting antennas, and echo signals reflected from the human body part under test are collected synchronously through four receiving antennas, outputting four synchronous baseband I / Q signals. The four synchronous baseband I / Q signals are subjected to channel consistency calibration, DC removal, static clutter suppression, and differential motion compensation to obtain preprocessed multi-channel echo signals. Spatial spectrum estimation is performed on the preprocessed multi-channel echo signal to locate the blood vessel region to be tested. Continuous phase change sequence is extracted by multi-channel joint phase dewinding and converted into sub-millimeter-level micro-displacement pulse wave time sequence signal corresponding to human blood vessel pulsation. Multi-channel pulse wave spatial feature parameters are then extracted. Based on the pre-calibrated multi-channel feature-blood pressure mapping model, the systolic and diastolic blood pressure values ​​of the human body are calculated according to the spatial feature parameters of the multi-channel pulse wave. Output and store blood pressure test results to enable communication and interaction with the host terminal.

[0008] Preferably, the process of performing channel consistency calibration, DC removal, static clutter suppression, and differential motion compensation on the four synchronous baseband I / Q signals includes: The amplitude and phase of the four synchronous baseband I / Q signals are calibrated to eliminate hardware deviations between channels; By differentially processing the echo signals from the main and auxiliary transmitters, millimeter-level macro-displacement interference caused by human breathing and slight body swaying is removed, while sub-millimeter-level micro-displacement signals of vascular pulsation are preserved.

[0009] Preferably, the step of spatial spectrum estimation to locate the target blood vessel region from the preprocessed multi-channel echo signal, extracting a continuous phase change sequence through multi-channel joint phase dewinding, converting it into a sub-millimeter-level micro-displacement pulse wave time-series signal corresponding to human blood vessel pulsation, and extracting multi-channel pulse wave spatial feature parameters includes: Using the MIMO spatial spectrum estimation algorithm, the DOA direction of arrival of the preprocessed multi-channel echo signal is estimated to locate the target vascular region of the radial artery or precordial region and extract the corresponding effective echo signal. An improved branching method based on the phase consistency constraint of the preprocessed multi-channel echo signal is used to perform joint phase unwrapping on the multi-channel echo signal, eliminating phase jumps under low signal-to-noise ratio and obtaining a continuous phase change sequence without jumps. Based on the linear relationship between phase change and displacement, the phase change sequence is converted into a sub-millimeter-level micro-displacement time sequence signal corresponding to human blood vessel pulsation, i.e., a multi-channel pulse wave signal. Extract the time-domain features of a single-channel pulse wave, as well as spatial feature parameters including the pulse wave phase difference, amplitude ratio, pulse wave propagation time difference (PATD), and peak arrival time difference between the four receiving channels.

[0010] Preferably, the pre-calibrated multi-channel feature-blood pressure mapping model is a multi-feature fusion regression model based on the LightGBM algorithm; The input features of the model include single-channel temporal features and multi-channel spatial features. The model is trained by a population sample covering different ages, genders, body mass indexes, and exercise states. The label of the training sample is the true blood pressure value measured by a standard mercury sphygmomanometer. The model supports personalized calibration by users. By inputting at least three sets of standard blood pressure measurements and corresponding multi-channel pulse wave feature parameters, the model parameters are fine-tuned and optimized based on transfer learning to improve individual detection accuracy.

[0011] Preferably, the output and storage of blood pressure detection results includes: Wireless data transmission is achieved with mobile APP and host computer via Bluetooth communication unit; An alarm is triggered by an audible and visual alarm unit when a blood pressure value exceeds a preset normal range.

[0012] Preferably, the method further includes: performing orthogonal demodulation on the effective echo signal to obtain a multi-channel phase sequence wrapped in the range of [-π, π].

[0013] Preferably, the phase change sequence is converted into a sub-millimeter-level micro-displacement time-series signal corresponding to human blood vessel pulsation, specifically calculated using the following formula:

[0014] Where Δd is the sub-millimeter-level micro-displacement time-series signal. The wavelength corresponding to the center frequency of the UWB signal. It is a continuous, non-jumping phase change sequence.

[0015] Preferably, the sample data is preprocessed by min-max normalization to eliminate differences in feature dimensions, and the training set and test set are divided in an 8:2 ratio.

[0016] Preferably, the model after pre-calibration and personalized fine-tuning is iteratively optimized, and the feature weights and iteration parameters of the model are dynamically adjusted based on multiple batches of measured data.

[0017] This invention also provides a non-contact blood pressure detection system based on MIMO UWB, the system comprising: The MIMO UWB radar front-end module is used to transmit ultra-wideband impulse pulse signals in time-division duplex mode through dual transmitting antennas, and synchronously collect echo signals reflected from the human body part to be tested through four receiving antennas, and output four synchronous baseband I / Q signals. The multi-channel synchronous preprocessing module is used to perform channel consistency calibration, DC removal, static clutter suppression, and differential motion compensation on the four synchronous baseband I / Q signals to obtain the preprocessed multi-channel echo signals. The MIMO spatial domain pulse wave extraction module is used to perform spatial spectrum estimation on the preprocessed multi-channel echo signal to locate the target blood vessel region. It extracts a continuous phase change sequence through multi-channel joint phase dewinding, converts it into a sub-millimeter-level micro-displacement pulse wave time sequence signal corresponding to human blood vessel pulsation, and extracts multi-channel pulse wave spatial feature parameters. The multi-channel feature fusion blood pressure inversion module is used to calculate the systolic and diastolic blood pressure values ​​of the human body based on the pre-calibrated multi-channel feature-blood pressure mapping model and the spatial feature parameters of the multi-channel pulse wave. The output interaction module is used to output and store blood pressure test results and realize communication and interaction with the host terminal.

[0018] Compared with the prior art, the technical solution of the embodiments of the present invention has the following beneficial effects: This invention provides a non-contact blood pressure detection method and system based on MIMO UWB. The method includes: transmitting ultra-wideband impulse pulse signals via time-division duplex transmission using dual transmitting antennas; synchronously acquiring echo signals reflected from the target body part via four receiving antennas; and outputting four synchronous baseband I / Q signals. The four synchronous baseband I / Q signals are then processed with channel consistency calibration, DC removal, static clutter suppression, and differential motion compensation to obtain pre-processed multi-channel echo signals. Spatial spectrum estimation is performed on the pre-processed multi-channel echo signals to locate the target blood vessel region. A continuous phase change sequence is extracted through multi-channel joint phase unwrapping and converted into a sub-millimeter-level micro-displacement pulse wave time-series signal corresponding to human vascular pulsation. Multi-channel pulse wave spatial feature parameters are extracted. Based on a pre-calibrated multi-channel feature-blood pressure mapping model, the systolic and diastolic blood pressure values ​​are calculated according to the multi-channel pulse wave spatial feature parameters. The blood pressure detection results are output and stored, enabling communication and interaction with a host terminal. High-precision micro-displacement monitoring is achieved through multi-channel joint signal processing. Furthermore, a multi-channel synchronous preprocessing and MIMO spatial domain signal processing link was designed. Macro displacement interference was removed by dual-transmission differential motion compensation, the target blood vessel region was located by MIMO spatial spectrum estimation, and sub-millimeter-level micro-displacement detection under low signal-to-noise ratio was achieved by multi-channel joint phase unwrapping. The pulse wave signal can be stably extracted under slight human body shaking and non-resting state, and the anti-interference capability has achieved a qualitative breakthrough. Furthermore, the dual transmitting antennas enable time-division beam scanning and differential motion calibration, and the four receiving channels enable spatial coherent accumulation, resulting in a 60% increase in beam coverage and a signal-to-noise ratio improvement of over 15dB. Furthermore, spatial features such as multi-channel phase difference and conduction time difference are extracted and combined with single-channel temporal features to construct a multi-channel feature set. A multi-feature fusion model is built based on the LightGBM algorithm, which significantly improves the model's generalization ability. At the same time, a personalized calibration process based on transfer learning and iterative optimization and deployment adaptation steps are designed. The detection accuracy meets the ISO 81060-2:2018 medical grade standard. The average absolute error of systolic blood pressure on the test set is ≤2.9mmHg, and the average absolute error of diastolic blood pressure is ≤2.5mmHg, which is significantly better than the existing technical solutions and is suitable for multi-scenario deployment. Furthermore, the detection distance covers 0.1m-1.5m, supports detection through up to 3 layers of thin clothing, and can be adapted to various scenarios such as wrist-worn, desktop, wall-mounted, and bedside monitoring. The hardware is highly customizable, does not require binding to a specific chip, and has extremely strong practical application value. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention, but not all embodiments. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a flowchart illustrating a preferred embodiment of a non-contact blood pressure detection method based on MIMO UWB. Figure 2 This is a schematic diagram of a non-contact blood pressure detection system based on MIMO UWB according to a preferred embodiment of the present invention; Figure 3 This is a schematic diagram comparing the measured results of MIMO spatial domain pulse wave signals in a non-contact blood pressure detection method based on MIMO UWB according to a preferred embodiment of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. 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.

[0022] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0023] This invention addresses the problems and shortcomings of existing technologies by providing a non-contact blood pressure detection method and system based on MIMO UWB, which achieves high-precision micro-displacement monitoring through the processing of multi-channel combined signals.

[0024] Figure 1 This is a schematic flowchart of a non-contact blood pressure detection method based on MIMO UWB, according to a preferred embodiment of the present invention. Now refer to... Figure 1 This invention provides a non-contact blood pressure detection method based on MIMO UWB, the method comprising: Step S101: Transmit ultra-wideband impulse pulse signals through dual transmitting antennas in time-division duplex mode, and synchronously collect echo signals reflected from the human body part to be tested through four receiving antennas, and output four synchronous baseband I / Q signals. Step S102: Perform channel consistency calibration, DC removal, static clutter suppression, and differential motion compensation on the four synchronous baseband I / Q signals to obtain the preprocessed multi-channel echo signal; Step S103: Spatial spectrum estimation is performed on the preprocessed multi-channel echo signal to locate the blood vessel region to be tested. The continuous phase change sequence is extracted by multi-channel joint phase dewinding and converted into the sub-millimeter-level micro-displacement pulse wave time sequence signal corresponding to human blood vessel pulsation. The spatial feature parameters of the multi-channel pulse wave are extracted. Step S104: Based on the pre-calibrated multi-channel feature-blood pressure mapping model, calculate the systolic and diastolic blood pressure values ​​of the human body according to the multi-channel pulse wave spatial feature parameters; Step S105: Output and store the blood pressure test results to achieve communication and interaction with the host terminal.

[0025] Specifically, the system is first initialized, configuring the operating parameters of the MIMO UWB radar, setting the time-division duplex timing of the dual transmit antennas, and completing the synchronization calibration and channel consistency calibration of the four receive channels. Then, the two transmit antennas are controlled to alternately transmit ultra-wideband impulse pulse signals conforming to domestic spectrum standards. The echo signals reflected from the tested parts of the human body are synchronously acquired through the four receive antennas, outputting four synchronous baseband I / Q signals. Multi-channel synchronous preprocessing is then performed on the four synchronous baseband I / Q signals, including channel consistency calibration, DC removal, static clutter suppression, differential motion compensation, and bandpass filtering, to obtain the preprocessed multi-channel I / Q signals. Multi-channel echo signals are processed; MIMO spatial spectrum estimation is performed on the preprocessed multi-channel echo signals to locate the target blood vessel region; continuous phase change sequences are extracted through multi-channel joint phase dewinding and converted into sub-millimeter-level micro-displacement pulse wave signals corresponding to blood vessel pulsation; multi-channel pulse wave spatial feature parameters are extracted; the extracted multi-channel pulse wave feature parameters are input into a pre-calibrated multi-channel feature-blood pressure mapping model to calculate the systolic and diastolic blood pressure values; the blood pressure detection results are output and stored, alarms are triggered for abnormal blood pressure values, and the data is uploaded to the upper terminal.

[0026] The configured UWB radar operating parameters include: center frequency 7.9872GHz, signal bandwidth 1GHz, pulse repetition frequency 2kHz, transmit power spectral density ≤-41.3dBm / MHz, which complies with the domestic management regulations for the 7.25GHz-8.75GHz UWB frequency band; the time division switching period of the dual transmit antennas is 500μs, and the synchronization sampling error of the four receive channels is ≤10ps.

[0027] In specific implementation, the processing of the four synchronous baseband I / Q signals, including channel consistency calibration, DC removal, static clutter suppression, and differential motion compensation, includes: The amplitude and phase of the four synchronous baseband I / Q signals are calibrated to eliminate hardware deviations between channels; By differentially processing the echo signals from the main and auxiliary transmitters, millimeter-level macro-displacement interference caused by human breathing and slight body swaying is removed, while sub-millimeter-level micro-displacement signals of vascular pulsation are preserved.

[0028] In specific implementation, the spatial spectrum estimation of the preprocessed multi-channel echo signal is used to locate the target blood vessel region. A continuous phase change sequence is extracted through multi-channel joint phase dewinding, which is then converted into a sub-millimeter-level micro-displacement pulse wave time-series signal corresponding to human blood vessel pulsation. The extraction of multi-channel pulse wave spatial feature parameters includes: Using the MIMO spatial spectrum estimation algorithm, the DOA direction of arrival of the preprocessed multi-channel echo signal is estimated to locate the target vascular region of the radial artery or precordial region and extract the corresponding effective echo signal. An improved branching method based on the phase consistency constraint of the preprocessed multi-channel echo signal is used to perform joint phase unwrapping on the multi-channel echo signal, eliminating phase jumps under low signal-to-noise ratio and obtaining a continuous phase change sequence without jumps. Based on the linear relationship between phase change and displacement, the phase change sequence is converted into a sub-millimeter-level micro-displacement time sequence signal corresponding to human blood vessel pulsation, i.e., a multi-channel pulse wave signal. Extract the time-domain features of a single-channel pulse wave, as well as spatial feature parameters including the pulse wave phase difference, amplitude ratio, pulse wave propagation time difference (PATD), and peak arrival time difference between the four receiving channels.

[0029] In specific implementation, the pre-calibrated multi-channel feature-blood pressure mapping model is a multi-feature fusion regression model based on the LightGBM algorithm; The input features of the model include single-channel temporal features and multi-channel spatial features. The model is trained by a population sample covering different ages, genders, body mass indexes, and exercise states. The label of the training sample is the true blood pressure value measured by a standard mercury sphygmomanometer. The model supports personalized calibration by users. By inputting at least three sets of standard blood pressure measurements and corresponding multi-channel pulse wave feature parameters, the model parameters are fine-tuned and optimized based on transfer learning to improve individual detection accuracy.

[0030] In specific implementation, the output and storage of blood pressure detection results includes: Wireless data transmission is achieved with mobile APP and host computer via Bluetooth communication unit; An alarm is triggered by an audible and visual alarm unit when a blood pressure value exceeds a preset normal range.

[0031] In a specific implementation, the method further includes: performing orthogonal demodulation on the effective echo signal to obtain a multi-channel phase sequence wrapped in the range of [-π,π].

[0032] In practical implementation, the phase change sequence is converted into a sub-millimeter-level micro-displacement time-series signal corresponding to human blood vessel pulsation, specifically calculated using the following formula:

[0033] Where Δd is the sub-millimeter-level micro-displacement time-series signal. The wavelength corresponding to the center frequency of the UWB signal. It is a continuous, non-jumping phase change sequence.

[0034] In practice, the sample data is preprocessed by min-max normalization to eliminate differences in feature dimensions, and the training set and test set are divided in an 8:2 ratio.

[0035] In practice, the model after pre-calibration and personalized fine-tuning is iteratively optimized, and the feature weights and iteration parameters of the model are dynamically adjusted based on multiple batches of measured data.

[0036] Specifically, the pre-calibration process of the multi-channel feature-blood pressure mapping model includes: collecting a predetermined number of sample data, covering healthy and hypertensive individuals of different ages, genders, and body mass indexes, and covering resting and mild exercise states. Each sample includes a corresponding multi-channel pulse wave feature set and the true values ​​of systolic and diastolic blood pressure measured by a standard mercury sphygmomanometer; performing min-max normalization preprocessing on the sample data to eliminate differences in feature dimensions, and dividing the training and testing sets in an 8:2 ratio; constructing a multi-feature fusion regression model using the LightGBM algorithm, with the multi-channel pulse wave feature set as input and the true values ​​of systolic and diastolic blood pressure as output, and training the model; and using the test set to verify the accuracy of the trained model. The model is considered accurate when its measurement error conforms to ISO standards. When implementing the 81060-2:2018 international standard for electronic blood pressure monitors, the model pre-calibration is completed; personalized calibration is supported, allowing users to input at least three sets of standard blood pressure measurements and corresponding multi-channel pulse wave feature parameters, and fine-tuning the model based on transfer learning to improve individual detection accuracy; the pre-calibrated and personalized fine-tuned model is iteratively optimized, dynamically adjusting the model's feature weights and iteration parameters based on multiple batches of real-world data, further reducing measurement errors caused by individual differences and environmental interference, and ensuring the model's long-term detection stability; after model optimization, model deployment and adaptation are performed, and the model is lightweighted and trimmed according to the hardware resources of different application scenarios to ensure that the model can achieve real-time blood pressure inversion on low-power hardware, adapting to the needs of multiple application scenarios, including wearable, desktop, and wall-mounted monitors.

[0037] Figure 2 This is a schematic diagram of a non-contact blood pressure monitoring system based on MIMO UWB, according to a preferred embodiment of the present invention. Now refer to... Figure 2 This invention also provides a non-contact blood pressure detection system based on MIMO UWB, the system comprising: The MIMO UWB radar front-end module 21 is used to transmit ultra-wideband impulse pulse signals through dual transmitting antennas in time-division duplex mode, and synchronously collect echo signals reflected from the human body part to be tested through four receiving antennas, and output four synchronous baseband I / Q signals. The multi-channel synchronous preprocessing module 22 is used to perform channel consistency calibration, DC removal, static clutter suppression, and differential motion compensation on the four synchronous baseband I / Q signals to obtain the preprocessed multi-channel echo signal. The MIMO spatial domain pulse wave extraction module 23 is used to perform spatial spectrum estimation on the preprocessed multi-channel echo signal to locate the blood vessel region to be tested, extract the continuous phase change sequence through multi-channel joint phase dewinding, convert it into the sub-millimeter-level micro-displacement pulse wave time sequence signal corresponding to human blood vessel pulsation, and extract the multi-channel pulse wave spatial feature parameters. The multi-channel feature fusion blood pressure inversion module 24 is used to calculate the systolic and diastolic blood pressure values ​​of the human body based on the pre-calibrated multi-channel feature-blood pressure mapping model and the multi-channel pulse wave spatial feature parameters. The output interaction module 25 is used to output and store blood pressure detection results and realize communication interaction with the host terminal.

[0038] Specifically, the MIMO UWB radar front-end module 21 includes a UWB radar main control chip, two transmit links, four receive links, a two-transmit, four-receive microstrip antenna array, a dual-transmit timing control unit, and a four-channel synchronous sampling unit. The UWB radar main control chip operates in the 6GHz-9GHz frequency band, with a maximum signal bandwidth of 1GHz, complying with the domestic 7.25GHz-8.75GHz UWB spectrum management specifications. It integrates two fully coherent transmitters, four fully coherent receivers, and an ARM Cortex-M series main control processor. The dual-transmit timing control unit controls the two transmit antennas to alternately transmit ultra-wideband signals in time-division duplex mode. The main transmit antenna is used for primary detection of the target blood vessel area, while the auxiliary transmit antenna is used for differential calibration of overall human motion. The four-channel synchronous sampling unit is used to achieve synchronous sampling of the ADCs of the four receive links, with a sampling synchronization error ≤10ps, ensuring the phase consistency of the multi-channel signals.

[0039] Specifically, the two-transmitter, four-receiver microstrip antenna array adopts a linear arrangement structure. The two transmitting antennas are located at both ends of the array, and the four receiving antennas are evenly distributed between the two transmitting antennas. The spacing between antenna elements is half the wavelength corresponding to the center frequency. The antenna array operates in the 6GHz-9GHz frequency band, with an element gain ≥3.5dBi, a front-to-back ratio ≥18dB, a half-power beamwidth of 55°-65°, and a beam coverage overlap of ≥80% for the dual transmitting antennas.

[0040] Specifically, the multi-channel synchronous preprocessing module 22 includes a channel consistency calibration unit, a DC removal unit, a second-order moving target display MTI filter unit, a differential motion compensation unit, and an adaptive bandpass filter unit connected in sequence. The channel consistency calibration unit is used to perform amplitude and phase calibration on the four received signals to eliminate hardware deviations between channels; the differential motion compensation unit is used to remove millimeter-level macro-displacement interference caused by human breathing and slight body swaying through differential processing of the echo signals from the main transmitter and auxiliary transmitter, while retaining the sub-millimeter-level micro-displacement signal of vascular pulsation. The passband range of the adaptive bandpass filter unit is 0.5Hz-20Hz, used to cover the effective frequency range of human heart rate and pulse waves.

[0041] Specifically, the multi-channel synchronous preprocessing includes: performing channel consistency calibration on the four I / Q signals, compensating for amplitude and phase deviations between channels using pre-stored calibration coefficients; performing DC removal processing on the calibrated multi-channel signals, removing the DC component caused by static environmental reflections by subtracting the signal mean; using a second-order moving target display MTI filter to remove static clutter and low-frequency environmental interference; performing differential processing on the echo signals from the main and auxiliary transmitters to remove millimeter-level macro-displacement interference caused by human respiration and slight body swaying, while retaining sub-millimeter-level micro-displacement signals of vascular pulsation; and using an adaptive Butterworth bandpass filter from 0.5Hz to 20Hz to suppress out-of-band high-frequency noise and retain effective pulse wave signal components.

[0042] Figure 3 This is a schematic diagram comparing the measured results of MIMO spatial domain pulse wave signals in a non-contact blood pressure detection method based on MIMO UWB according to a preferred embodiment of the present invention. (See now) Figure 3 The antenna array operates in the 6GHz-9GHz frequency band, with a unit gain ≥3.5dBi, a front-to-back ratio ≥18dB, a half-power beamwidth of 55°-65°, and a beam coverage overlap of ≥80% for the dual-transmit antennas.

[0043] The following are further examples: In a specific implementation, a desktop non-contact blood pressure detection system based on MIMO UWB radar is provided, including a MIMO UWB radar front-end module, a multi-channel synchronous preprocessing module, a MIMO spatial domain pulse wave extraction module, a multi-channel feature fusion blood pressure inversion module, and an output interaction module.

[0044] The MIMO UWB radar front-end module uses a UWB main control chip, with two fully coherent transmitters and four fully coherent receivers. It operates in the 6GHz-9GHz frequency band, with a maximum signal bandwidth of 1GHz and a center frequency of 7.9872GHz, complying with domestic 7.25GHz-8.75GHz UWB spectrum management standards. Its transmit power spectral density is -41.3dBm / MHz. The module incorporates a dual-transmit timing control unit and four synchronous sampling units. The dual-transmit antenna time-division switching period is 500μs, and the synchronous sampling error of the four receiving channels is ≤10ps. It is equipped with a two-transmit, four-receive microstrip antenna array, employing a linear arrangement. The two transmit antennas are located at both ends of the array, and the four receive antennas are evenly distributed in the middle. The antenna element spacing is 18.75mm, the antenna element gain is 3.8dBi, the front-to-back ratio is 20dB, the half-power beamwidth is 60°, and the dual-transmit beam overlap is 85%.

[0045] The multi-channel synchronous preprocessing module is connected to the MIMO UWB radar front-end module. It includes a channel consistency calibration unit, a DC removal unit, a second-order MTI filter unit, a differential motion compensation unit, and a 0.5Hz-20Hz adaptive Butterworth bandpass filter unit connected in sequence. The differential motion compensation unit removes macro-displacement interference caused by human breathing and body swaying within ±5cm range by differentiating the echoes of the main and auxiliary transmitters, while retaining the sub-millimeter-level micro-displacement signal of blood vessel pulsation.

[0046] The MIMO spatial domain pulse wave extraction module, connected to the multi-channel synchronous preprocessing module, includes a vascular region localization unit, a multi-channel joint phase unwinding unit, a micro-displacement conversion unit, and a multi-channel feature extraction unit. The vascular region localization unit uses the MUSIC algorithm for DOA estimation, achieving a localization accuracy of ≤3°, and can accurately locate the radial artery region of the wrist. The multi-channel joint phase unwinding unit employs an improved branching method based on four-channel signal phase consistency constraints, achieving a phase unwinding accuracy of ≥99.5% and enabling micro-displacement detection accuracy at the 0.05mm level. The multi-channel feature extraction unit can extract four single-channel temporal features and six multi-channel spatial features, forming a 10-dimensional feature set.

[0047] The multi-channel feature fusion blood pressure inversion module is connected to the MIMO spatial domain pulse wave extraction module. It incorporates a multi-feature fusion regression model based on the LightGBM algorithm. The model was trained using 1200 samples covering individuals aged 18-80 years with different ages, genders, and body mass indexes, covering both resting and mild exercise states. Model training and calibration sequentially complete sample collection (A1), preprocessing (A2), model training (A3), accuracy verification (A4), personalized calibration (A5), iterative optimization (A7), and deployment adaptation (A8). Step A7 iteratively adjusts model parameters using multiple batches of real-world data to further improve detection stability. Step A8 performs lightweight model adaptation for desktop scenarios to ensure real-time detection. The model test set shows a mean absolute error of 2.9 mmHg and a standard deviation of 3.0 mmHg for systolic blood pressure and a mean absolute error of 2.5 mmHg and a standard deviation of 2.4 mmHg for diastolic blood pressure, fully complying with the ISO81060-2:2018 standard. Personalized calibration is supported; users can input three sets of standard blood pressure values ​​to complete model fine-tuning.

[0048] The output interaction module, connected to the multi-channel feature fusion blood pressure inversion module, includes a 1.3-inch OLED display unit, a BLE5.0 Bluetooth communication unit, a buzzer-based audible and visual alarm unit, and an 8MB local Flash storage unit. The OLED display unit displays systolic blood pressure, diastolic blood pressure, heart rate, and measurement status in real time. The Bluetooth unit can upload data to a mobile APP to generate a health report. The audible and visual alarm unit triggers an alarm when blood pressure exceeds the normal range.

[0049] This embodiment provides a desktop non-contact blood pressure monitoring system based on MIMO UWB radar. The detection distance is 0.2m-1m, it supports measurement through 3 layers of thin cotton clothing, supports dynamic detection under slight human body shaking, and the single measurement time is ≤3s. It can realize 24-hour continuous dynamic blood pressure monitoring and is suitable for scenarios such as home health monitoring, office health monitoring, and smart elderly care.

[0050] In a specific implementation, a non-contact blood pressure detection method based on MIMO UWB radar is provided, including the following steps: The system is powered on and initialized, and the UWB radar operating parameters are configured as follows: center frequency 7.9872GHz, signal bandwidth 1GHz, pulse repetition frequency 2kHz, and transmit power spectral density -41.3dBm / MHz; the dual transmit antenna time division duplex timing is set with a switching period of 500μs; the synchronization calibration and channel consistency calibration of the four receive channels are completed, and the amplitude and phase deviations between channels are compensated. The user aligns the radial artery of their wrist with the radar antenna, maintaining a detection distance of 0.3m-0.5m. The system controls the two transmitting antennas to alternately transmit ultra-wideband impulse pulse signals in a time-division manner. The main transmitting antenna is aimed at the radial artery area for primary detection, while the auxiliary transmitting antenna monitors the overall movement of the human body. The four receiving antennas synchronously collect the reflected echoes, which are then down-converted by the radar chip and sampled by the ADC to output four synchronous baseband I / Q signals. Multi-channel synchronous preprocessing is performed on the four synchronous baseband I / Q signals: first, channel consistency calibration is completed by pre-stored calibration coefficients, then DC removal is performed to remove static environmental noise, low-frequency interference is suppressed by second-order MTI filtering, macro displacement interference from breathing and body swaying is removed by main and auxiliary transmitter echo differential processing, and finally, the preprocessed multi-channel echo signal is obtained by 0.5Hz-20Hz adaptive bandpass filtering. MIMO spatial domain processing was performed on the preprocessed multi-channel echo signal: the MUSIC algorithm was used to estimate the direction of arrival (DOA) to locate the area to be tested in the radial artery of the wrist, and the effective echo signal of the area was extracted; the effective signal was orthogonally demodulated to obtain the wrapped phase sequence, and the continuous phase change sequence was obtained by four-way joint phase dewinding, which was converted into a sub-millimeter-level micro-displacement pulse wave signal, and a 10-dimensional multi-channel pulse wave feature set was extracted; The extracted multi-channel feature set is input into the pre-calibrated LightGBM multi-feature fusion regression model, and the model calculates and outputs the systolic and diastolic blood pressure values ​​of the human body in real time. Blood pressure and heart rate values ​​are displayed in real time on an OLED screen, and the detection data is stored locally in Flash and uploaded to a mobile app via Bluetooth. When a systolic blood pressure ≥140mmHg or a diastolic blood pressure ≥90mmHg is detected, an audible and visual alarm is triggered to remind the user.

[0051] The pre-calibration process of the multi-channel feature-blood pressure mapping model includes: Data from 1200 samples were collected, covering healthy individuals and individuals with essential hypertension aged 18-80 years, with 50% males and 50% females, and a body mass index of 18.5-32. Data from resting state and mild exercise state after slow walking were collected for each sample. Each sample included a 10-dimensional multi-channel pulse wave feature set and the true blood pressure value measured by medical staff using a standard mercury sphygmomanometer. The sample data is processed by min-max normalization, which maps all features to the [0,1] interval to eliminate the difference in units. The training set and the test set are randomly divided in an 8:2 ratio. A LightGBM multi-feature fusion regression model was constructed with a learning rate of 0.05, a tree depth of 6, and 250 iterations. The model was trained with a 10-dimensional multi-channel feature set as input and the true values ​​of systolic and diastolic blood pressure as outputs. The model's accuracy was verified using a test set. The verification results showed that the mean absolute error of systolic blood pressure was 2.9 mmHg, with a standard deviation of 3.0 mmHg; and the mean absolute error of diastolic blood pressure was 2.5 mmHg, with a standard deviation of 2.4 mmHg. This fully complies with the ISO 81060-2:2018 international standard, and the model pre-calibration is complete. It supports personalized calibration for users. Users can use a standard cuff blood pressure monitor to measure three sets of blood pressure values ​​at rest and simultaneously collect three sets of corresponding multi-channel pulse wave feature sets. The pre-trained model is fine-tuned through transfer learning, and optimization can be completed in just 50 iterations, further reducing measurement errors caused by individual differences.

[0052] This embodiment provides a desktop-based non-contact blood pressure detection method based on MIMO UWB radar, which realizes cuffless and non-contact blood pressure detection. The measurement process is comfortable, supports dynamic scene detection, and the detection accuracy reaches medical-grade standards. The solution is highly scalable and can be adapted to various hardware forms and application scenarios.

[0053] In summary, the present invention provides a non-contact blood pressure detection method and system based on MIMO UWB. The method includes: transmitting ultra-wideband impulse pulse signals via time-division duplex transmission using dual transmitting antennas; synchronously acquiring echo signals reflected from the target body part via four receiving antennas; and outputting four synchronous baseband I / Q signals. The four synchronous baseband I / Q signals are then processed with channel consistency calibration, DC removal, static clutter suppression, and differential motion compensation to obtain pre-processed multi-channel echo signals. Spatial spectrum estimation is performed on the pre-processed multi-channel echo signals to locate the target blood vessel region. A continuous phase change sequence is extracted through multi-channel joint phase unwrapping and converted into a sub-millimeter-level micro-displacement pulse wave time-series signal corresponding to human vascular pulsation. Multi-channel pulse wave spatial feature parameters are extracted. Based on a pre-calibrated multi-channel feature-blood pressure mapping model, the systolic and diastolic blood pressure values ​​are calculated according to the multi-channel pulse wave spatial feature parameters. The blood pressure detection results are output and stored, enabling communication and interaction with a host terminal. High-precision micro-displacement monitoring is achieved through multi-channel joint signal processing. Furthermore, a multi-channel synchronous preprocessing and MIMO spatial domain signal processing link was designed. Macro displacement interference was removed by dual-transmission differential motion compensation, the target blood vessel region was located by MIMO spatial spectrum estimation, and sub-millimeter-level micro-displacement detection under low signal-to-noise ratio was achieved by multi-channel joint phase unwrapping. The pulse wave signal can be stably extracted under slight human body shaking and non-resting state, and the anti-interference capability has achieved a qualitative breakthrough. Furthermore, the dual transmitting antennas enable time-division beam scanning and differential motion calibration, and the four receiving channels enable spatial coherent accumulation, resulting in a 60% increase in beam coverage and a signal-to-noise ratio improvement of over 15dB. Furthermore, spatial features such as multi-channel phase difference and conduction time difference are extracted and combined with single-channel temporal features to construct a multi-channel feature set. A multi-feature fusion model is built based on the LightGBM algorithm, which significantly improves the model's generalization ability. At the same time, a personalized calibration process based on transfer learning and iterative optimization and deployment adaptation steps are designed. The detection accuracy meets the ISO 81060-2:2018 medical grade standard. The average absolute error of systolic blood pressure on the test set is ≤2.9mmHg, and the average absolute error of diastolic blood pressure is ≤2.5mmHg, which is significantly better than the existing technical solutions and is suitable for multi-scenario deployment. Furthermore, the detection distance covers 0.1m-1.5m, supports detection through up to 3 layers of thin clothing, and can be adapted to various scenarios such as wrist-worn, desktop, wall-mounted, and bedside monitoring. The hardware is highly customizable, does not require binding to a specific chip, and has extremely strong practical application value.

[0054] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; 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 or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A non-contact blood pressure detection method based on MIMO UWB, characterized in that, The method includes: Ultra-wideband impulse pulse signals are transmitted in time-division duplex mode using dual transmitting antennas, and echo signals reflected from the human body part under test are collected synchronously through four receiving antennas, outputting four synchronous baseband I / Q signals. The four synchronous baseband I / Q signals are subjected to channel consistency calibration, DC removal, static clutter suppression, and differential motion compensation to obtain preprocessed multi-channel echo signals. Spatial spectrum estimation is performed on the preprocessed multi-channel echo signal to locate the blood vessel region to be tested. Continuous phase change sequence is extracted by multi-channel joint phase dewinding and converted into sub-millimeter-level micro-displacement pulse wave time sequence signal corresponding to human blood vessel pulsation. Multi-channel pulse wave spatial feature parameters are then extracted. Based on the pre-calibrated multi-channel feature-blood pressure mapping model, the systolic and diastolic blood pressure values ​​of the human body are calculated according to the spatial feature parameters of the multi-channel pulse wave. Output and store blood pressure test results to enable communication and interaction with the host terminal; The process involves spatial spectrum estimation of the preprocessed multi-channel echo signal to locate the target blood vessel region, extraction of continuous phase change sequences through multi-channel joint phase dewinding, conversion into sub-millimeter-level micro-displacement pulse wave time-series signals corresponding to human blood vessel pulsation, and extraction of multi-channel pulse wave spatial feature parameters, including: Using the MIMO spatial spectrum estimation algorithm, the DOA direction of arrival of the preprocessed multi-channel echo signal is estimated to locate the target vascular region of the radial artery or precordial region and extract the corresponding effective echo signal. An improved branching method based on the phase consistency constraint of the preprocessed multi-channel echo signal is used to perform joint phase unwrapping on the multi-channel echo signal, eliminating phase jumps under low signal-to-noise ratio and obtaining a continuous phase change sequence without jumps. Based on the linear relationship between phase change and displacement, the phase change sequence is converted into a sub-millimeter-level micro-displacement time sequence signal corresponding to human blood vessel pulsation, i.e., a multi-channel pulse wave signal. Extract the time-domain features of a single-channel pulse wave, as well as spatial feature parameters including the pulse wave phase difference, amplitude ratio, pulse wave propagation time difference (PATD), and peak arrival time difference between the four receiving channels.

2. The non-contact blood pressure detection method based on MIMO UWB according to claim 1, characterized in that, The processing of the four synchronous baseband I / Q signals, including channel consistency calibration, DC removal, static clutter suppression, and differential motion compensation, includes: The amplitude and phase of the four synchronous baseband I / Q signals are calibrated to eliminate hardware deviations between channels; By differentially processing the echo signals from the main and auxiliary transmitters, millimeter-level macro-displacement interference caused by human breathing and slight body swaying is removed, while sub-millimeter-level micro-displacement signals of vascular pulsation are preserved.

3. The non-contact blood pressure detection method based on MIMO UWB according to claim 1, characterized in that, The pre-calibrated multi-channel feature-blood pressure mapping model is a multi-feature fusion regression model based on the LightGBM algorithm; The input features of the model include single-channel temporal features and multi-channel spatial features. The model is trained by a population sample covering different ages, genders, body mass indexes, and exercise states. The label of the training sample is the true blood pressure value measured by a standard mercury sphygmomanometer. The model supports personalized calibration by users. By inputting at least three sets of standard blood pressure measurements and corresponding multi-channel pulse wave feature parameters, the model parameters are fine-tuned and optimized based on transfer learning to improve individual detection accuracy.

4. The non-contact blood pressure detection method based on MIMO UWB according to claim 1, characterized in that, The output and storage of blood pressure test results includes: Wireless data transmission is conducted with mobile APP and host computer via Bluetooth communication unit; An alarm is triggered by an audible and visual alarm unit when a blood pressure value exceeds a preset normal range.

5. The non-contact blood pressure detection method based on MIMO UWB according to claim 1, characterized in that, The method further includes: performing orthogonal demodulation on the effective echo signal to obtain a multi-channel phase sequence wrapped in the range of [-π,π].

6. The non-contact blood pressure detection method based on MIMO UWB according to claim 5, characterized in that, The phase change sequence is converted into a sub-millimeter-level micro-displacement time-series signal corresponding to human blood vessel pulsation, specifically calculated using the following formula: Wherein, Δd is the sub-millimeter-level micro-displacement time-series signal. The wavelength corresponding to the center frequency of the UWB signal. It is a continuous, non-jumping phase change sequence.

7. The non-contact blood pressure detection method based on MIMO UWB according to claim 3, characterized in that, The sample data were preprocessed using min-max normalization to eliminate differences in feature dimensions, and the training and test sets were divided in an 8:2 ratio.

8. The non-contact blood pressure detection method based on MIMO UWB according to claim 7, characterized in that, The model is iteratively optimized after pre-calibration and personalized fine-tuning. Based on multiple batches of measured data, the feature weights and iteration parameters of the model are dynamically adjusted.

9. A non-contact blood pressure detection system based on MIMO UWB, characterized in that, The system includes: The MIMO UWB radar front-end module is used to transmit ultra-wideband impulse pulse signals in time-division duplex mode through dual transmitting antennas, and synchronously collect echo signals reflected from the human body part to be tested through four receiving antennas, and output four synchronous baseband I / Q signals. The multi-channel synchronous preprocessing module is used to perform channel consistency calibration, DC removal, static clutter suppression, and differential motion compensation on the four synchronous baseband I / Q signals to obtain the preprocessed multi-channel echo signals. The MIMO spatial domain pulse wave extraction module is used to perform spatial spectrum estimation on the preprocessed multi-channel echo signal to locate the target blood vessel region. It extracts a continuous phase change sequence through multi-channel joint phase dewinding, converts it into a sub-millimeter-level micro-displacement pulse wave time sequence signal corresponding to human blood vessel pulsation, and extracts multi-channel pulse wave spatial feature parameters. The multi-channel feature fusion blood pressure inversion module is used to calculate the systolic and diastolic blood pressure values ​​of the human body based on the pre-calibrated multi-channel feature-blood pressure mapping model and the spatial feature parameters of the multi-channel pulse wave. The output interaction module is used to output and store blood pressure test results and realize communication interaction with the host terminal. The process involves spatial spectrum estimation of the preprocessed multi-channel echo signal to locate the target blood vessel region, extraction of continuous phase change sequences through multi-channel joint phase dewinding, conversion into sub-millimeter-level micro-displacement pulse wave time-series signals corresponding to human blood vessel pulsation, and extraction of multi-channel pulse wave spatial feature parameters, including: Using the MIMO spatial spectrum estimation algorithm, the DOA direction of arrival of the preprocessed multi-channel echo signal is estimated to locate the target vascular region of the radial artery or precordial region and extract the corresponding effective echo signal. An improved branching method based on the phase consistency constraint of the preprocessed multi-channel echo signal is used to perform joint phase unwrapping on the multi-channel echo signal, eliminating phase jumps under low signal-to-noise ratio and obtaining a continuous phase change sequence without jumps. Based on the linear relationship between phase change and displacement, the phase change sequence is converted into a sub-millimeter-level micro-displacement time sequence signal corresponding to human blood vessel pulsation, i.e., a multi-channel pulse wave signal. Extract the time-domain features of a single-channel pulse wave, as well as spatial feature parameters including the pulse wave phase difference, amplitude ratio, pulse wave propagation time difference (PATD), and peak arrival time difference between the four receiving channels.

Images