A multi-sensor fusion intelligent riding protection system and method

Abstract

The present application relates to computer human-computer interaction, multi-modal sensing signal processing and intelligent sensing system technical field, disclose a kind of multi-sensor fusion intelligent riding protection system and method, comprising: posture inertia acquisition unit, physiological sign monitoring unit, rear radar detection unit, position positioning unit, communication gateway module and fusion arbitration master unit, master unit extracts acceleration high-frequency component to photoelectric pulse wave sequence is carried out artifact hedging, analyze heart rate and blood oxygen determination physiological health state;According to the phase of helmet self-vibration, the same frequency compensation is implemented to radar intermediate frequency phase, analyze approach speed;According to the posture data to generate fall flag, and output emergency alarm instruction when physiological health is abnormal;Gateway module calls voice interface through system bus to dial emergency telephone and send map rendering coordinates, the present application eliminates body motion artifact and radar self-vibration interference, eliminates fall and blind area false alarm, improves the reliability of rescue call for help.

Description

Technical Field

[0001] This invention relates to a multi-sensor fusion intelligent cycling protection system and method, belonging to the technical fields of computer human-computer interaction, multimodal sensor signal processing, and intelligent sensing systems. Background Technology

[0002] Intelligent cycling safety devices typically integrate inertial measurement sensors, photoplethysmography (PPG) sensors, millimeter-wave radar, and positioning and communication components into the helmet. These are used to acquire real-time information about the cyclist's posture, physiological state, and road targets behind them. They trigger audible and visual warnings and emergency calls when a fall or approaching vehicle is detected. In intelligent sensing systems and human-computer interaction applications, these sensors collect different types of data, requiring real-time acquisition, noise reduction, and fusion processing of various sensor signals to reliably achieve cycling posture perception, state judgment, and interactive control. However, in dynamic cycling scenarios such as bumpy roads, cornering, and collisions, these sensors are easily affected by mechanical vibrations and cycling motion. The unevenness of the helmet surface can cause high-frequency micro-vibrations, which are transmitted to the millimeter-wave radar antenna at the rear. This coherently modulates the phase of the swept frequency echo, causing micro-Doppler spectrum broadening of the intermediate frequency signal. This can lead the radar to misjudge the presence of a high-speed approaching target at the rear. At the same time, the body movement of the rider and the head displacement at the moment of impact can cause relative sliding between the photoelectric sensor and the skin surface, which can disrupt the ratio of red light to infrared light absorbance in the photoplethysmography signal. This can lead to non-physiological abrupt changes in the calculation of physiological indicators. In addition, the centrifugal acceleration generated by riding on a curve can change the projection distribution of the gravity vector on the triaxial orthogonal accelerometer, causing attitude measurement reference drift. This makes it difficult for conventional static threshold judgment to distinguish between normal tilting and bending and real weightlessness from a fall.

[0003] To address the aforementioned interference, simply increasing the sensor trigger threshold or extending the filtering smoothing window would reduce the system's sensitivity to real collisions and physiological anomalies, leading to delays in emergency rescue calls. Software high-pass filtering of only a single sensor channel is insufficient to simultaneously handle phase modulation interference in radar echoes and motion artifacts in photoelectric signals, easily causing algorithm deadlock and signal delays. At the software processing level, existing technologies typically attempt to eliminate interference through logical segmentation or estimating reference signals, but these also have limitations. For example, the Chinese invention patent with authorization publication number CN107595296B... The patent discloses a method and device for detecting blood oxygen saturation to eliminate motion interference. It calculates the ratio of light signal intensity during the low acceleration phase and uses this ratio to construct a reference signal for input to an adaptive filter. However, this technology implicitly relies on the assumption that the light intensity ratio during the resting phase remains constant during subsequent high-dynamic motion. Under continuous cycling conditions, muscle contraction and intense body movement can cause dynamic drift in subcutaneous hemodynamics and physiological baseline, resulting in a physical mismatch between the historical resting phase ratio and the real-time high-dynamic process. This, in turn, distorts the constructed reference signal and causes a secondary deviation in the calculation of physiological indicators.

[0004] Therefore, how to reduce the parasitic interference and motion artifacts on multiple sensors in high-dynamic cycling scenarios, and how to reliably process, jointly judge, and respond to multimodal sensing signals in human-computer interaction, has become the technical problem to be solved by this invention. Summary of the Invention

[0005] To address the problems in the background art, the technical solution of the present invention is as follows: A multi-sensor fusion intelligent cycling protection system, comprising:

[0006] Attitude inertial acquisition unit, used to acquire three-axis acceleration data and three-axis angular velocity data;

[0007] Physiological signs monitoring unit, used to acquire photoelectric pulse wave sequences;

[0008] The rear radar detection unit is used to acquire the intermediate frequency signal of the frequency-modulated continuous wave millimeter-wave radar;

[0009] Positioning unit, used to acquire satellite positioning coordinate data;

[0010] The integrated arbitration control unit is connected to the attitude inertial acquisition unit, physiological sign monitoring unit, rear radar detection unit, position positioning unit, and communication gateway module, respectively. It is used to convert satellite positioning coordinate data into map rendering coordinate data; extract high-frequency acceleration components from the three-axis acceleration data as a motion noise reference sequence; employ adaptive filtering to offset motion artifacts in the photoelectric pulse wave sequence; calculate a clean pulse wave signal; and analyze heart rate and blood oxygen saturation values ​​to determine physiological health status; extract the helmet's self-vibration phase based on the three-axis acceleration and three-axis angular velocity data; perform same-frequency phase compensation on the original phase of the intermediate frequency signal; and analyze the target approach velocity data; generate a fall action indicator signal based on the three-axis acceleration and three-axis angular velocity data; and output an emergency alarm command when the fall action indicator signal is valid and the physiological health status is abnormal.

[0011] The communication gateway module is used to call the voice call interface through the system bus when an emergency alarm command is received, dial the preset emergency contact number and send map rendering coordinate data.

[0012] Preferably, the fusion arbitration control unit is also used to calculate the cornering centrifugal acceleration vector by combining the riding speed and yaw rate in the triaxial angular velocity data obtained by the position positioning unit, and to subtract the cornering centrifugal acceleration vector from the triaxial acceleration data to restore the gravitational acceleration projection vector; the fusion arbitration control unit calculates the resultant value of the triaxial acceleration and the resultant value of the triaxial angular velocity using the formula for the resultant value of the triaxial acceleration and the formula for the resultant value of the triaxial angular velocity. When the resultant value of the triaxial acceleration is lower than the low-order weightlessness threshold or higher than the high-order impact threshold, and the resultant value of the triaxial angular velocity is higher than the rotational angular velocity threshold, a fall action indicator signal is generated, and the effective duration of the fall action indicator signal is set to 10s.

[0013] Preferably, the low-order weightlessness threshold is 0.4g, the high-order impact threshold is 1.6g, and the rotational angular velocity threshold is 250dps; the fusion arbitration main control unit is used to set the fall action flag signal to a high level when the sum of the three-axis acceleration is lower than 0.4g or higher than 1.6g, and the sum of the three-axis angular velocities is higher than 250dps.

[0014] Preferably, the determination of physiological health status includes a continuous delay verification window mechanism; the fusion arbitration main control unit compares the parsed heart rate value and blood oxygen saturation value with preset health thresholds respectively. When the heart rate value exceeds the preset upper limit threshold or the blood oxygen saturation value is lower than the preset lower limit threshold, a timer is started; if the value exceeds the preset health threshold and the duration exceeds the preset verification time, the fusion arbitration main control unit sets the physiological health status to an abnormal health state; if the value recovers to the preset health threshold range within the preset verification time, the timer is reset and the physiological health status remains normal; wherein the preset verification time is set to 4 seconds.

[0015] Preferably, the rear radar detection unit uses a 24GHz frequency-modulated continuous wave millimeter-wave radar; before analyzing the target's approach velocity data, the fusion arbitration main control unit extracts the vertical vibration acceleration and yaw rate of the attitude inertial acquisition unit, calculates the helmet's natural phase, performs same-frequency phase compensation on the original phase of the intermediate frequency signal, and eliminates the micro-Doppler spectrum broadening caused by helmet micro-vibration; the fusion arbitration main control unit performs spectrum analysis on the compensated pure intermediate frequency signal to analyze the distance data and approach velocity data of the rear target.

[0016] Preferably, the system also includes an alarm prompting component connected to the fusion arbitration main control unit; when the extracted distance data is less than the distance safety threshold of 5.0m and the approach speed data is greater than the speed safety threshold of 2m / s, the fusion arbitration main control unit drives the alarm prompting component to output an audible and visual alarm signal, and starts a 3s cooling timer after triggering the warning, suppressing repeated voice broadcasts during the 3s cooling timer operation.

[0017] Preferably, the communication gateway module runs on an embedded Linux system and integrates the DBus system bus and the ofono mobile communication framework. When the communication gateway module receives an emergency alarm command, it obtains the ofono management interface through the DBus system bus and scans for available modems. It then directly calls the dialing method of the voice call management interface to pass in the preset emergency contact phone number and establishes a system bus-level voice call channel.

[0018] Preferably, the location positioning unit extracts latitude and longitude in degree-division format from the satellite positioning coordinate data and converts it into decimal degree coordinates; the fusion arbitration main control unit converts the decimal degree coordinates in the WGS-84 Earth coordinate system into map rendering coordinate data in the GCJ-02 Mars coordinate system through a nonlinear coordinate offset algorithm, and synchronizes the map rendering coordinate data to the mobile terminal through a long connection network protocol.

[0019] Preferably, the physiological sign monitoring unit includes a multispectral photoplethysmography (PPG) sensor module that fits onto the rider's scalp. The PPG sensor module uses a 660nm red light emitter and a 940nm infrared light emitter to alternately irradiate subcutaneous microvessels, and a photodetector collects the reflected light intensity change sequence to output a photoelectric pulse wave sequence. The PPG sensor module is connected to the fusion arbitration main control unit via a universal asynchronous transceiver serial port.

[0020] A multi-sensor fusion intelligent cycling protection method, used to operate a multi-sensor fusion intelligent cycling protection system, includes the following steps:

[0021] Step S1: Acquire triaxial acceleration data and triaxial angular velocity data through the attitude inertial acquisition unit;

[0022] Step S2: Acquire photoelectric pulse wave sequence through physiological sign monitoring unit;

[0023] Step S3: Acquire the intermediate frequency signal of the frequency-modulated continuous wave millimeter-wave radar through the rear radar detection unit;

[0024] Step S4: Obtain satellite positioning coordinate data through the positioning unit;

[0025] Step S5: The satellite positioning coordinate data is converted into map rendering coordinate data by the fusion arbitration main control unit; the high-frequency acceleration component in the three-axis acceleration data is extracted as a motion noise reference sequence; adaptive filtering is used to offset the motion artifacts in the photoelectric pulse wave sequence, and a clean pulse wave signal is calculated; the heart rate and blood oxygen saturation values ​​are analyzed to determine the physiological health status; the helmet's self-vibration phase is extracted based on the three-axis acceleration data and three-axis angular velocity data, and the original phase of the intermediate frequency signal is compensated for with the same frequency; the target approach speed data is analyzed; a fall action indicator signal is generated based on the three-axis acceleration data and three-axis angular velocity data; when the fall action indicator signal is valid and the physiological health status is abnormal, an emergency alarm command is output.

[0026] Step S6: Upon receiving an emergency alarm command, the system bus is used via the communication gateway module to call the voice call interface, dial the preset emergency contact number, and send map rendering coordinate data.

[0027] Compared with the prior art, the beneficial effects of the present invention are:

[0028] 1. In multi-sensor fusion intelligent cycling protection, the high-frequency acceleration component obtained by the attitude inertial acquisition unit is used as a motion noise reference sequence to counteract body motion artifacts in the photoelectric pulse wave sequence. This can reduce the interference of road bumps, violent exercise and collision impact on photoelectric signals, and keep the ratio of red light to infrared light absorbance true and stable. This can avoid non-physiological abnormal changes in heart rate and blood oxygen saturation values ​​due to body motion, and improve the reliability of physiological health status assessment.

[0029] 2. By using the helmet's self-vibration phase to perform phase compensation on the intermediate frequency signal of the millimeter-wave radar, the interference caused by helmet micro-vibration on the radar echo phase can be suppressed, the micro-Doppler spectrum broadening caused by micro-vibration can be reduced, and the frequency shift caused by helmet self-vibration can be avoided as the approach of a target behind. This can reduce the false triggering of blind zone collision avoidance warning and the frequent activation of the cooling mechanism under bumpy conditions, and improve the continuity and stability of the target detection process behind.

[0030] 3. By compensating for centrifugal acceleration in corners based on riding speed and yaw rate, the impact of centrifugal acceleration on attitude measurement during fast cornering and large-angle leaning can be reduced, allowing the gravitational acceleration projection vector to more realistically reflect the riding posture. This reduces attitude measurement reference drift and avoids misjudging normal cornering or obstacle crossing as a fall. Furthermore, by combining collision actions with physiological abnormalities for joint judgment, the accuracy and stability of fall collision event identification under complex riding conditions can be improved. Attached Figure Description

[0031] Figure 1 This is a flowchart of the multimodal sensing data fusion processing of the present invention;

[0032] Figure 2 This is a state transition diagram for the dual-channel alarm arbitration of the present invention.

[0033] The objectives, features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0034] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0035] A multi-sensor fusion intelligent cycling protection system includes:

[0036] Attitude inertial acquisition unit, used to acquire three-axis acceleration data and three-axis angular velocity data;

[0037] Physiological signs monitoring unit, used to acquire photoelectric pulse wave sequences;

[0038] The rear radar detection unit is used to acquire the intermediate frequency signal of the frequency-modulated continuous wave millimeter-wave radar;

[0039] Positioning unit, used to acquire satellite positioning coordinate data;

[0040] The integrated arbitration control unit is connected to the attitude inertial acquisition unit, physiological sign monitoring unit, rear radar detection unit, position positioning unit, and communication gateway module, respectively. It is used to convert satellite positioning coordinate data into map rendering coordinate data; extract high-frequency acceleration components from the three-axis acceleration data as a motion noise reference sequence; employ adaptive filtering to offset motion artifacts in the photoelectric pulse wave sequence; calculate a clean pulse wave signal; and analyze heart rate and blood oxygen saturation values ​​to determine physiological health status; extract the helmet's self-vibration phase based on the three-axis acceleration and three-axis angular velocity data; perform same-frequency phase compensation on the original phase of the intermediate frequency signal; and analyze the target approach velocity data; generate a fall action indicator signal based on the three-axis acceleration and three-axis angular velocity data; and output an emergency alarm command when the fall action indicator signal is valid and the physiological health status is abnormal.

[0041] The communication gateway module is used to call the voice call interface through the system bus when an emergency alarm command is received, dial the preset emergency contact number and send map rendering coordinate data.

[0042] Preferably, the fusion arbitration control unit is also used to calculate the cornering centrifugal acceleration vector by combining the riding speed and yaw rate in the triaxial angular velocity data obtained by the position positioning unit, and to subtract the cornering centrifugal acceleration vector from the triaxial acceleration data to restore the gravitational acceleration projection vector; the fusion arbitration control unit calculates the resultant value of the triaxial acceleration and the resultant value of the triaxial angular velocity using the formula for the resultant value of the triaxial acceleration and the formula for the resultant value of the triaxial angular velocity. When the resultant value of the triaxial acceleration is lower than the low-order weightlessness threshold or higher than the high-order impact threshold, and the resultant value of the triaxial angular velocity is higher than the rotational angular velocity threshold, a fall action indicator signal is generated, and the effective duration of the fall action indicator signal is set to 10s.

[0043] Preferably, the low-order weightlessness threshold is 0.4g, the high-order impact threshold is 1.6g, and the rotational angular velocity threshold is 250dps; the fusion arbitration main control unit is used to set the fall action flag signal to a high level when the sum of the three-axis acceleration is lower than 0.4g or higher than 1.6g, and the sum of the three-axis angular velocities is higher than 250dps.

[0044] Preferably, the determination of physiological health status includes a continuous delay verification window mechanism; the fusion arbitration main control unit compares the parsed heart rate value and blood oxygen saturation value with preset health thresholds respectively. When the heart rate value exceeds the preset upper limit threshold or the blood oxygen saturation value is lower than the preset lower limit threshold, a timer is started; if the value exceeds the preset health threshold and the duration exceeds the preset verification time, the fusion arbitration main control unit sets the physiological health status to an abnormal health state; if the value recovers to the preset health threshold range within the preset verification time, the timer is reset and the physiological health status remains normal; wherein the preset verification time is set to 4 seconds.

[0045] Preferably, the rear radar detection unit uses a 24GHz frequency-modulated continuous wave millimeter-wave radar; before analyzing the target's approach velocity data, the fusion arbitration main control unit extracts the vertical vibration acceleration and yaw rate of the attitude inertial acquisition unit, calculates the helmet's natural phase, performs same-frequency phase compensation on the original phase of the intermediate frequency signal, and eliminates the micro-Doppler spectrum broadening caused by helmet micro-vibration; the fusion arbitration main control unit performs spectrum analysis on the compensated pure intermediate frequency signal to analyze the distance data and approach velocity data of the rear target.

[0046] Preferably, the system also includes an alarm prompting component connected to the fusion arbitration main control unit; when the extracted distance data is less than the distance safety threshold of 5.0m and the approach speed data is greater than the speed safety threshold of 2m / s, the fusion arbitration main control unit drives the alarm prompting component to output an audible and visual alarm signal, and starts a 3s cooling timer after triggering the warning, suppressing repeated voice broadcasts during the 3s cooling timer operation.

[0047] Preferably, the communication gateway module runs on an embedded Linux system and integrates the DBus system bus and the ofono mobile communication framework. When the communication gateway module receives an emergency alarm command, it obtains the ofono management interface through the DBus system bus and scans for available modems. It then directly calls the dialing method of the voice call management interface to pass in the preset emergency contact phone number and establishes a system bus-level voice call channel.

[0048] Preferably, the location positioning unit extracts latitude and longitude in degree-division format from the satellite positioning coordinate data and converts it into decimal degree coordinates; the fusion arbitration main control unit converts the decimal degree coordinates in the WGS-84 Earth coordinate system into map rendering coordinate data in the GCJ-02 Mars coordinate system through a nonlinear coordinate offset algorithm, and synchronizes the map rendering coordinate data to the mobile terminal through a long connection network protocol.

[0049] Preferably, the physiological sign monitoring unit includes a multispectral photoplethysmography (PPG) sensor module that fits onto the rider's scalp. The PPG sensor module uses a 660nm red light emitter and a 940nm infrared light emitter to alternately irradiate subcutaneous microvessels, and a photodetector collects the reflected light intensity change sequence to output a photoelectric pulse wave sequence. The PPG sensor module is connected to the fusion arbitration main control unit via a universal asynchronous transceiver serial port.

[0050] A multi-sensor fusion intelligent cycling protection method, used to operate a multi-sensor fusion intelligent cycling protection system, includes the following steps:

[0051] Step S1: Acquire triaxial acceleration data and triaxial angular velocity data through the attitude inertial acquisition unit;

[0052] Step S2: Acquire photoelectric pulse wave sequence through physiological sign monitoring unit;

[0053] Step S3: Acquire the intermediate frequency signal of the frequency-modulated continuous wave millimeter-wave radar through the rear radar detection unit;

[0054] Step S4: Obtain satellite positioning coordinate data through the positioning unit;

[0055] Step S5: The satellite positioning coordinate data is converted into map rendering coordinate data by the fusion arbitration main control unit; the high-frequency acceleration component in the three-axis acceleration data is extracted as a motion noise reference sequence; adaptive filtering is used to offset the motion artifacts in the photoelectric pulse wave sequence, and a clean pulse wave signal is calculated; the heart rate and blood oxygen saturation values ​​are analyzed to determine the physiological health status; the helmet's self-vibration phase is extracted based on the three-axis acceleration data and three-axis angular velocity data, and the original phase of the intermediate frequency signal is compensated for with the same frequency; the target approach speed data is analyzed; a fall action indicator signal is generated based on the three-axis acceleration data and three-axis angular velocity data; when the fall action indicator signal is valid and the physiological health status is abnormal, an emergency alarm command is output.

[0056] Step S6: Upon receiving an emergency alarm command, the system bus is used through the communication gateway module to call the voice call interface, dial the preset emergency contact number, and send map rendering coordinate data.

[0057] Example 1: During high-speed dynamic mountain road cycling, the system includes an attitude inertial acquisition unit, a physiological sign monitoring unit, a rear radar detection unit, a position positioning unit, a communication gateway module, and a fusion arbitration main control unit (hereinafter referred to as the main control unit). These units work collaboratively. The attitude inertial acquisition unit acquires raw triaxial acceleration and triaxial angular velocity data at a sampling rate of 100Hz using a six-axis inertial measurement sensor, and converts the data according to the sampling resolution to obtain triaxial acceleration and triaxial angular velocity. Road bumps and physical impacts cause high-frequency vibrations in the rider's head protection components. The multispectral photoplethysmography sensor module, which is in contact with the scalp, experiences relative displacement, introducing motion artifacts and disrupting the absorbance ratio of the 660nm red light to 940nm infrared light intensity sequence. At this time, the main control unit extracts the high-frequency acceleration components from 10Hz to 50Hz in the triaxial acceleration data as a motion noise reference sequence. The raw photoelectric pulse wave sequence acquired by the photoplethysmography sensor is applied. The normalized adaptive least mean square filter algorithm is used to dynamically update the weight vector. The noise estimate caused by body motion is offset in real time from the original photoelectric pulse wave sequence, and the pure pulse wave signal is reconstructed in situ. ,in, The sampling time is indicated; then the AC and DC components are extracted, and the true heart rate and blood oxygen saturation values ​​are analyzed. In the cross-scale mapping mechanism, the physical relationship between macroscopic rigid body acceleration and microscopic optical path changes is established through the skin elastic deformation and optical path absorptivity model. The helmet high-frequency acceleration directly determines the dynamic compression force between the photoelectric module and the skin contact surface. The main control unit first calculates the composite acceleration amplitude of the three-axis high-frequency acceleration according to the Euclidean norm. This composite acceleration amplitude reflects the macroscopic dynamic normal pressure change acting on the skin surface. Then, based on the pre-determined skin viscoelastic strain coefficient, the normal pressure change is linearly transformed into the microscopic compression depth change of the subcutaneous microvascular layer. Then, the compression depth is transformed into a scalar modulation noise estimate of the red light and infrared light absorptivity through the Lambert-Beer law. The adaptive filter weight vector is dynamically updated on this basis, transforming the macroscopic mechanical quantity into an optical path artifact estimate consistent with the physical dimensions of the single-channel photoelectric signal.

[0058] The continuous uneven road surface causes high-frequency self-vibration mechanical vibrations generated by the protective components to be transmitted to the 24GHz frequency-modulated continuous wave millimeter-wave radar antenna of the rear radar detection unit. This results in coherent phase modulation of the original phase of the received intermediate frequency signal with a sweep period of 1ms and a bandwidth of 250MHz, and induces micro-Doppler spectrum broadening. This can easily lead to misinterpretation of the self-vibration frequency shift as a high-speed approach of a vehicle behind. Therefore, the main control unit extracts the vertical vibration acceleration and yaw rate from the attitude inertial acquisition unit to calculate the helmet's self-vibration phase. Combined with the pre-calibrated attitude-radar mechanical coupling coefficient The original phase of the radar intermediate frequency signal Performing in-frequency phase compensation yields a clean intermediate frequency signal phase. The coherent phase modulation superposition condition of mechanical micro-vibration on the echo phase is eliminated; then, the spectrum analysis of the compensated pure intermediate frequency signal is performed and the difference frequency is extracted. Calculate the distance data of the target behind. Approach speed data When distance data Less than 5.0m from the safety threshold and approaching speed data When the speed exceeds the safety threshold of 2 m / s, the main control unit drives the alarm prompt component to output an audible and visual alarm signal, and simultaneously starts a 3-second cooling timer to suppress repeated voice broadcasts. The extraction of the helmet's self-vibration phase is based on the second integral of the helmet's rigid body kinematics. The main control unit performs a second time-domain integration on the vertical vibration acceleration to obtain the physical displacement of the antenna in the vertical direction. At the same time, the yaw rate is multiplied by the spatial geometric distance lever arm from the antenna mounting center to the helmet's rotation axis to calculate the tangential displacement along the normal direction of the radar antenna caused by the rotational angular motion. Subsequently, the vertical displacement and tangential displacement are vector-superimposed and multiplied by a phase shift conversion constant that matches the electromagnetic wave wavelength of the radar center to convert it into the electromagnetic wave physical phase change value of the antenna in the line-of-sight direction. Since there is a fixed mechanical rigidity transmission loss between the antenna and the inertial sensor, the main control unit uses the mechanical coupling coefficient of the attitude radar to weight and scale the physical phase change value to restore the parasitic phase change of the received intermediate frequency directly caused by the antenna's physical displacement. This is then directly subtracted from the original intermediate frequency phase received by the radar to eliminate the micro-Doppler spectrum broadening caused by the antenna's mechanical micro-vibration.

[0059] When a rider makes a sharp turn, the lateral centrifugal acceleration alters the projection distribution of the gravitational acceleration vector onto the orthogonal triaxial accelerometer, causing a dynamic drift in the attitude measurement reference. The main control unit combines the riding speed of 8 m / s output by the position positioning unit with the yaw rate measured by the attitude inertial acquisition unit to calculate the cornering centrifugal acceleration vector. This centrifugal acceleration vector is subtracted from the original triaxial acceleration data to restore the gravitational acceleration projection vector in the orthogonal coordinate system. Finally, the corrected triaxial acceleration summation is calculated using the triaxial acceleration summation formula and the triaxial angular velocity summation formula. Combined value of triaxial angular velocity The resultant value of the three-axis acceleration is calculated by taking the square root of the sum of the squares of the three axial acceleration components after correction, and the resultant value of the three-axis angular velocity is calculated by taking the square root of the sum of the squares of the three axial angular velocity components. When the resultant value of the three-axis acceleration is lower than the low-order weightlessness threshold of 0.4g or higher than the high-order impact threshold of 1.6g, and the resultant value of the three-axis angular velocity is higher than the rotational angular velocity threshold of 250dps, a high-level fall action flag signal with a continuous effective time of 10s is generated to eliminate the interference of cornering and large-angle obstacle avoidance actions on the fall judgment criteria. In the subtraction calculation of the centrifugal acceleration vector in the corner, the main control unit first uses the pitch angle and roll angle obtained by the attitude inertial acquisition unit to construct the direction cosine transformation moment from the sensor coordinate system to the bicycle body navigation coordinate system in real time. Since the helmet yaw rate includes local high-frequency disturbances from the rider's head rotation, the main control unit uses low-pass filtering to remove the high-frequency head rotation and extracts the low-frequency vehicle turning angular velocity consistent with the vehicle's steering. Then, the forward riding speed along the vehicle's direction of travel measured by the positioning unit is multiplied with the low-frequency vehicle turning angular velocity to obtain the lateral centrifugal acceleration scalar in the navigation coordinate system. This centrifugal acceleration scalar is then projected back to the sensor's orthogonal three-axis coordinate system through the inverse matrix of the direction cosine transformation matrix to obtain the cornering centrifugal acceleration vector. Finally, this cornering centrifugal acceleration vector is subtracted from the original three-axis acceleration data one by one along the axis to complete the decoupling of the vehicle's attitude and head rotation and restore the gravitational acceleration projection vector.

[0060] The determination of physiological health status adopts a 4-second continuous delay verification window mechanism. When the analyzed heart rate value exceeds the preset upper heart rate threshold or the blood oxygen saturation value falls below the preset lower blood oxygen threshold, a timer is activated. Within the continuous 4-second verification period, if every frame of physiological data continuously exceeds the preset health threshold, the physiological health status is set to an abnormal health state. The health warning mechanism of the physiological sign monitoring unit operates independently. Once the physiological health status is determined to be abnormal, the main control unit immediately pushes a health alert pop-up to the mobile terminal's WeChat mini-program via a long-connection network protocol. The pop-up of this health alert pop-up is not affected by whether the rider has fallen. In the fall alarm call determination, the fall alarm dialing is controlled by physiological health monitoring. The main control unit performs a binary arbitration of fall action + health abnormality. Only when the fall action flag signal is valid and the physiological health status is abnormal will the main control unit output a call. If an emergency alarm command is issued, and only a fall is detected while the patient's physiological health is within the normal range, the output of the emergency alarm command and telephone dialing are suppressed to avoid false alarms. The communication gateway module runs on an embedded Linux system. When it receives an emergency alarm command, it directly obtains the management interface of the ofono mobile communication framework through the DBus system bus and scans for available modems. It then calls the dialing method of the voice call management interface to pass in the preset emergency contact phone number and establishes a system bus-level voice call channel. At the same time, the location unit extracts the latitude and longitude in degree format from the satellite positioning coordinate data and converts it into decimal coordinates in the WGS-84 Earth coordinate system. The main control unit then converts it into map rendering coordinate data in the GCJ-02 Mars coordinate system through a nonlinear coordinate offset algorithm and synchronizes it to the mobile terminal and the server backend via a long connection network protocol to complete the rescue response.

[0061] Example 2: A multi-sensor fusion intelligent cycling protection system was installed on a six-degree-of-freedom motion test bench to simulate real rugged mountain roads and high-frequency vibration conditions. The system includes an attitude inertial acquisition unit, a physiological sign monitoring unit, a rear radar detection unit, a position positioning unit, a communication gateway module, and a fusion arbitration main control unit. During data acquisition, the six-axis inertial measurement sensor outputs raw triaxial acceleration and triaxial angular velocity data at a sampling rate of 100Hz. The multispectral photoplethysmography sensor module outputs 660nm red light and 940nm infrared pulse wave sequences via a 115200bps serial port, and frequency-modulated continuous wave millimeter... The radar antenna outputs intermediate frequency signals via a 250MHz sweep bandwidth and a 1ms sweep period. The satellite positioning chip outputs latitude and longitude data in degree-division format at 1Hz, which is then converted into decimal coordinates by the position positioning unit. The test bench superimposes Gaussian white noise with a signal-to-noise ratio of 20dB and mechanical vibration harmonics at a frequency of 12Hz. When the self-oscillation frequency of the cycling protective helmet is in the range of 10Hz to 50Hz, the attitude inertial acquisition unit adopts a sampling rate of 100Hz to satisfy the Nyquist sampling law and leave more than twice the phase calculation margin. While extracting high-frequency noise, it also takes into account the microprocessor bus communication load to avoid data bus blockage.

[0062] In the photoelectric motion artifact offset test, the mechanical vibration harmonics caused micron-level relative sliding between the multispectral photoplethysmography sensor module and the head, resulting in the original photoelectric pulse wave sequence. When baseline drift occurs, the fusion arbitration control unit extracts high-frequency acceleration components from 10Hz to 50Hz to form a motion noise reference sequence. The normalized adaptive minimum mean square filter algorithm uses an update step size. and regularization constant Iterative update of weight vector ,according to Calculate and estimate noise Then by A pure pulse wave signal was obtained. The sample group of this invention uses the above-mentioned adaptive filtering, while the first sample group uses a fixed high-pass filtering algorithm without adaptive noise offsetting. The out-of-range control group sets the update step size to... The motion noise reference sequence consists of high-frequency acceleration components in the axial, lateral, and vertical directions at each sampling time. The adaptive weight vectors correspond to the normalized weighting coefficients of the three acceleration axes, respectively. The main control unit performs dot product multiplication and accumulation of the high-frequency acceleration components of each axis with the corresponding weighting coefficients to obtain the estimated noise in single-channel scalar form. Then, it subtracts the noise from the single-channel photoelectric pulse wave signal to keep the physical quantities and data dimensions involved in the subtraction consistent.

[0063] During continuous motion, at the 0.0s sampling time, the original photoelectric signal of the sample group of this invention is 1.25, and the motion noise reference sequence... for After the above processing, a pure signal of 1.08 was obtained, with a heart rate of 135 bpm and blood oxygen saturation of 88.4%, indicating a healthy physiological state. False; at the 1.0s sampling time, the original photoelectric signal is 1.31, and the motion noise reference sequence is... The pure signal was 1.06, the analyzed heart rate was 138 bpm, and the blood oxygen saturation was 87.8%. Keep False; at the 2.0s sampling time, the original photoelectric signal is 1.28, and the motion noise reference sequence is... The pure signal strength was 1.05, the analyzed heart rate was 140 bpm, and the blood oxygen saturation was 87.2%. Keep False; at the 3.0s sampling time, the original photoelectric signal is 1.34, and the motion noise reference sequence is... The pure signal strength was 1.04, the analyzed heart rate was 142 bpm, and the blood oxygen saturation was 86.5%. Still False; at the 4.0s sampling time, the original photoelectric signal is 1.36, and the motion noise reference sequence is... The pure signal was 1.03, the analyzed heart rate was 145 bpm, and the blood oxygen saturation was 85.9%. After continuous anomalies met the verification criteria... When True, under the same 4.0s input conditions, the original photoelectric signal of the comparison sample group 1 was 1.36, but without using a motion noise reference sequence, the pure signal after fixed high-pass filtering was 0.42 with amplitude distortion, the resolved heart rate was 182 bpm with a spurious mutation, and the resolved blood oxygen was 74.2% with a non-physiological drop. The result is True, leading to a misjudgment; the original photoelectric signal of the out-of-range control group is 1.36, and the motion noise reference sequence remains unchanged. Because the update step size exceeded the convergence stability limit, the pure signal was -0.85 and numerical divergence occurred. The heart rate analysis result was 0 and could not be analyzed, and the blood oxygen analysis result was 0.0%. If True, the corresponding algorithm fails.

[0064] The rear blind zone millimeter-wave radar anti-self-vibration interferometry test used a 24GHz frequency-modulated continuous wave millimeter-wave radar. High-frequency micro-vibration caused the original phase of the intermediate frequency signal to be lost. Coherent phase modulation is generated, causing micro-Doppler spectrum broadening. The fusion arbitration control unit obtains the helmet's natural phase based on triaxial angular velocity and vertical acceleration data. The mechanical coupling coefficient of attitude radar is adopted. Perform in-frequency phase compensation, and the phase of the compensated clean intermediate frequency signal is according to... The pure intermediate frequency signal is then subjected to spectrum analysis, and the difference frequency is extracted. Analyze distance data Approach speed data The present invention uses the same-frequency phase compensation in the sample group, while the comparison sample group two directly performs spectrum analysis on the uncompensated original intermediate frequency signal. The out-of-range control group sets the attitude radar mechanical coupling coefficient to... This results in overcompensation distortion, and warnings are issued based on a safe distance threshold of 5.0m and a safe speed threshold of 2m / s.

[0065] Under the condition of no vehicle behind and self-vibrating, the original radar phase of the sample group of this invention is 0.48 rad, the inertial phase is 0.35 rad, and after compensation, it is 0.20 rad. The spectrum broadening is 12 Hz, the resolvable distance is 18.5 m, and the approach speed is 0.02 m / s. The warning conditions are not met, so the alarm is suppressed. The second comparison sample group, under the same original radar phase of 0.48 rad, does not use the inertial phase. The compensated phase is maintained at 0.48 rad, the spectrum broadening is 145 Hz, the resolvable distance is 3.2 m, and the approach speed is 1.45 m / s. It outputs a false alarm. The out-of-range control group, under the original radar phase of 0.48 rad and the inertial phase of 0.35 rad, gets -0.08. With rad-compensated phase, a spectrum broadening of 98Hz, a resolution range of 4.1m, and an approach speed of 0.88m / s, false alarms were still generated. Under real-world vehicle approach conditions, the original radar phase of the sample group of this invention was 0.85rad, the inertial phase was 0.35rad, and after compensation it became 0.57rad, with a spectrum broadening of 15Hz, a resolution range of 4.8m, and an approach speed of 1.12m / s, meeting the warning conditions and triggering an alarm. In contrast, sample group two, which did not use inertial phase, had both the original radar phase and the compensated phase of 0.85rad, a spectrum broadening of 160Hz, a resolution range of 2.1m, and an approach speed of 2.85m / s, resulting in false lock-up cooling.

[0066] In the cornering and fall arbitration test, the rider passed through a corner with a radius of 2.36m at a speed of 8m / s, with a yaw rate of... The lateral centrifugal acceleration scalar is -193.980 dps, calculated from the riding speed and the vehicle's turning angular velocity. It weighs 2.763g and is superimposed on the orthogonal accelerometer. On the axis, the fusion arbitration control unit projects the lateral centrifugal acceleration scalar onto the sensor's orthogonal triaxial coordinate system to form a curve centrifugal acceleration vector, and subtracts it axis by axis from the original triaxial acceleration data. The resulting triaxial acceleration sum after sample correction in this invention is... The g was maintained at 1.0g, and no fall signal was generated; under simulated collision and ejection conditions, The magnitude is 1.8g, which is higher than the higher-order impact threshold of 1.6g. The sum of the three-axis angular velocities is... The value was 362.06 dps, which is higher than the rotational angular velocity threshold of 250 dps. The main control unit generated a high-level fall action signal lasting for 10 seconds. Its status is True. After the physiological sign monitoring unit continuously exceeds the preset health threshold within the 4-second continuous delay verification window, When True, a separate health alert pop-up appears on the mobile WeChat mini-program; during fall alarm dialing arbitration, the main control unit performs a combined fall and health determination, when... True and When True, the binary joint arbitration logic If the answer is 1, the communication gateway module, via the DBus system bus of the embedded Linux system, calls the VoiceCallManager interface and Dial method of the ofono mobile communication framework to send a dialing command to the modem. True but If it is False, then The value remains 0 and no dialing is triggered; compared to the third group which uses a static acceleration threshold judgment algorithm without centrifugal acceleration deduction, false alarms of falling occur when cornering; the out-of-range control group compresses the continuous verification window to 1 second, resulting in false judgments of physiological health status when the rider coughs transiently or goes over bumps.

[0067] Statistical analysis of continuous test results showed that the physiological heart rate analysis error of the present invention's sample group was 1.8 bpm, within 2 bpm, and the physiological blood oxygen analysis deviation was 0.8%, less than 1.2%. In contrast, the physiological heart rate analysis error of control group one was 18.5 bpm, and the physiological blood oxygen analysis deviation was 8.6%. The out-of-range control group did not generate corresponding heart rate analysis error and blood oxygen analysis deviation values ​​due to numerical divergence. These two values ​​were not recorded in control groups two and three. The radar blind zone false alarm rate of the present invention's sample group was 0.2%, compared to 34.5% in control group two and 21.8% in the out-of-range control group. This value was not recorded in control groups one and three. The curve fall false trigger rate of the present invention's sample group was 0.0%. The false alarm rate for the bumpy working condition in the present invention sample group was 0.1%, compared to 42.1% in the control group, while no such value was recorded in the other groups. Regarding the sensitivity for real collision recognition, the present invention sample group achieved 99.8%, compared to 91.2% in the control group, 88.5% in the control group, and both the control group and the control group achieved 99.5%. The emergency call bus connection delay in the present invention sample group was 138ms, while no corresponding value was recorded in the other groups. Under the same frequency phase compensation condition, the micro-Doppler spectrum broadening and narrowing was 82%. The above data were used to verify the engineering feasibility of the multi-source sensor signal offset decoupling and binary joint arbitration mechanism under the established test conditions.

[0068] Example 3: Before performing co-frequency phase compensation and motion artifact cancellation in a multi-sensor fusion intelligent cycling protection system, the main control unit first completes static and dynamic offline calibration to determine the mechanical coupling coefficient of the attitude radar. Update step size of the normalized adaptive minimum mean square filter algorithm With regularization constant The calibration values ​​were determined by fixing a protective helmet equipped with a 24GHz frequency-modulated continuous wave millimeter-wave radar antenna on a six-degree-of-freedom vibration table in a test lane with no approaching targets behind it. The vibration table was driven to apply sweep frequency vertical vibration and angular vibration in the 10Hz to 50Hz frequency band. The attitude inertial acquisition unit extracted the high-frequency component of acceleration and the vertical vibration component at a sampling rate of 100Hz and calculated the helmet's natural phase. Simultaneously, the original phase of the mixed intermediate frequency signal is acquired by millimeter-wave radar. The main control unit calculates the extreme values ​​of the cross-correlation coefficient between the phase of the original intermediate frequency signal and the phase of the helmet's natural vibration in the time domain, and uses the least squares method to linearly fit the phase shift ratio under multiple vibration amplitudes to solve for the mechanical coupling coefficient of the attitude radar under the protective helmet structure. In the noise calibration test of the multispectral photoplethysmography sensor module, raw photoelectric pulse wave sequences of riders were collected under steady-state riding conditions on flat roads and under bumpy conditions. Motion noise reference sequence constructed by combining high-frequency components of triaxial acceleration Under the constraints that the mean square error of the filter convergence is less than 0.005 and the convergence time is less than 120ms, the update step size is determined. With the regularization constant used to prevent the denominator from being zero The calibration parameters mentioned above are written into the non-volatile memory of the main control unit during system factory initialization and loaded into the static memory area during each system power-on self-test. They serve as input parameters for adaptive filtering and phase compensation operations. The calibration dataset consists of 100 discrete signal matrices collected under different riding bumps and vibration spectra. When the update step size is set to 0.01, the adaptive filtering algorithm achieves a balance between the mean square error convergence rate and steady-state mismatch. When the update step size is greater than 0.05, the iteration diverges, and when it is less than 0.002, it cannot keep up with transient body motion changes. The regularization constant is set to 0.001 to keep the denominator value stable at transient points where the acceleration is close to zero. The mechanical coupling coefficient of the attitude radar, 0.80, represents the mechanical amplitude attenuation rate when the high-frequency vibration of the helmet shell is transmitted to the millimeter-wave radar antenna substrate. It is pre-calibrated and stored in non-volatile memory for the standardized loading of algorithm parameters.

[0069] In determining the warning signal of a fall At that time, the resultant value of the triaxial acceleration The corresponding low-order weightlessness threshold of 0.4g and high-order impact threshold of 1.6g, as well as the combined triaxial angular velocity. The corresponding rotational angular velocity threshold of 250 dps is determined based on the dynamic data envelopment (DVE) under riding conditions. This threshold is applied to riding on flat roads, over obstacles, during sudden braking, and in steep cornering situations. The main control unit then considers the riding speed... 8.0 m / s and yaw rate Calculate the centrifugal acceleration vector for the curve at -193.980 dps. ,in, This represents the curve centrifugal acceleration vector projected onto the sensor's orthogonal triaxial coordinate system; subtracted from the original triaxial acceleration data. Subsequently, under normal riding conditions, the sum of the gravitational acceleration projection vector ranged from 0.7g to 1.3g, and the sum of the three-axis angular velocities was below 180 dps. In the test of the artificial dummy's unbalanced rollover and frontal collision, the sum of the three-axis acceleration after deducting the centrifugal acceleration vector in the curve dropped below 0.4g at the moment of weightlessness and exceeded 1.6g at the moment the head protection component hit the ground. At the same time, the sum of the three-axis angular velocities exceeded 250 dps. When the judgment conditions were met, the main control unit issued a fall action indicator signal. Set to high level and start a hold timer for 10 seconds. If no manual cancellation command is received from the rider within the 10-second hold time, then hold. The high-level state prevents the judgment from flickering due to rider posture adjustments. The physical boundaries of the above thresholds are determined based on the human head dynamic tolerance limit and the human biomechanical free fall model. When the human body loses balance and falls, the head support force disappears rapidly, and the projected value of gravitational acceleration drops below 0.4g within 0.15s after the start of the fall. When it is below 0.4g, the human body is judged to be in an irreversible weightlessness and overturning state. When the protective helmet hits the ground, the impact load on the contact surface causes the triaxial acceleration to rise to above 1.6g. Above 1.6g corresponds to the helmet withstanding a strong physical impact sufficient to cause injury. The rotational angular velocity threshold of 250dps corresponds to the critical uncontrolled angular velocity of the head rotating in space. When it is above this value, the rider is judged to have involuntary rollover. The fall action is determined by the three boundaries of low-order weightlessness, high-order impact, and rotational angular velocity.

[0070] Physiological health status A 4-second continuous delay verification window mechanism is employed to eliminate pulse wave waveform distortion and non-physiological outliers caused by transient body movements. The multispectral photoplethysmography sensor module transmits heart rate data to the main control unit at a serial port baud rate of 115200bps. With blood oxygen saturation value The frame data includes preset upper heart rate thresholds and lower blood oxygen thresholds, which support different working modes and dynamic customization. Specifically, these include normal cycling mode and sport cycling mode: in normal cycling mode, the preset upper heart rate threshold is set to 120 bpm by default and the preset lower blood oxygen threshold is set to 90% by default; in intense sport cycling mode, the preset upper heart rate threshold adaptively switches to 160 bpm and the preset lower blood oxygen threshold switches to 88%. Furthermore, both the preset upper heart rate threshold and the preset lower blood oxygen threshold in normal cycling mode and sport cycling mode can be customized by users via a WeChat mini-program on their mobile devices, and dynamically synchronized to the main control unit via network protocol for real-time configuration. The system is designed to accommodate the personalized monitoring needs of riders with different physical conditions. When the analyzed heart rate exceeds the preset upper limit threshold (e.g., 120 bpm in normal mode or 160 bpm in exercise mode) or the blood oxygen saturation falls below the preset lower limit threshold (e.g., 90% in normal mode or 88% in exercise mode), the main control unit starts a hardware timer and compares the received data frame by frame within a 4-second verification window. If any frame of data recovers to within the preset health threshold range set for the current working mode within the 4-second verification period (i.e., heart rate not higher than the upper limit threshold of the corresponding mode and blood oxygen saturation not lower than the lower limit threshold of the corresponding mode), the timer is reset and the starting point of the verification window is reset, and the physiological health status is maintained. When all sampled frames exceed the preset health threshold within 4 consecutive seconds, the main control unit sets the physiological health status to [condition]. Furthermore, the main control unit independently drives the mobile terminal's WeChat mini-program to display a health alert, and the triggering of this health alert is independent of whether the rider has fallen; simultaneously, The determination result is independently transmitted to the main control unit as the enabling condition for controlling the emergency call dialing in case of a fall.

[0071] The safe distance threshold of 5.0m and the safe speed threshold of 2m / s for the rear radar detection unit are set based on the engineering and physical relationship between braking distance and human reaction time delay. When the frequency-modulated continuous wave millimeter-wave radar performs spectral analysis on the compensated clean intermediate frequency signal, the distance data is extracted... Less than 5.0m and close to speed data When the speed exceeds 2 m / s, the main control unit drives the alarm prompt component to output an audible and visual alarm signal and simultaneously starts a 3-second cooling timer. During the cooling timer's operation, the millimeter-wave radar continues to acquire intermediate frequency signals and calculate the target distance and approach speed, while suppressing the audio amplifier chip from repeatedly broadcasting the voice alarm. The communication gateway module runs on an embedded Linux system. The system uses a fall + health binary joint judgment mechanism to control the fall alarm dialing: the health monitoring module has an independent early warning path, when... A health alert pop-up appears independently on the WeChat mini-program on the mobile terminal; while the emergency dialing alarm is regulated by physiological health status, and only when the binary joint arbitration logic outputs... At this time, the gateway daemon initiates an interface query to org.ofono.Manager via the DBus system bus and scans for available mobile modem nodes. If the primary modem node's response timeout exceeds 200ms, the daemon automatically switches to the backup Bluetooth gateway channel, calls the Dial method of the org.ofono.VoiceCallManager interface, passes in the preset emergency contact phone number to establish a voice call, and simultaneously, the location unit extracts latitude and longitude in degree format from the satellite positioning coordinate data and converts it to decimal degrees coordinates in the WGS-84 Earth coordinate system. The main control unit then converts the decimal degrees... The coordinates are converted into map rendering coordinate data in the GCJ-02 Mars coordinate system. After being packetized, the data is synchronized to the mobile terminal and the server backend via the WebSocket long connection protocol to complete the delivery of rescue instructions and location information. During the 3-second cooldown timer, only the repeated long voice broadcasts are blocked, while the visual warning lights and high-frequency buzzers continue to flash and sound in real time. If the speed of the target approaching from behind suddenly increases to more than 1.5 m / s or the distance suddenly shrinks to within 2.0 m, the main control unit will forcibly reset and interrupt the cooldown timer, and re-trigger the extremely high-priority voice alarm. This reduces the repetitive high-frequency voice broadcasts while retaining real-time early warning response under extremely dangerous conditions.

[0072] Example 4: This example combines Figures 1 to 2This paper describes a multi-sensor fusion intelligent cycling protection system and method, such as... Figure 1 As shown, step S1 acquires triaxial acceleration and triaxial angular velocity data through the attitude inertial acquisition unit; step S2 acquires photoelectric pulse wave sequences through the physiological sign monitoring unit; step S3 acquires intermediate frequency signals from frequency-modulated continuous wave millimeter-wave radar through the rear radar detection unit; step S4 obtains satellite positioning coordinate data through the position positioning unit; step S5 converts the satellite positioning coordinate data into map rendering coordinate data through the fusion arbitration main control unit, extracts the high-frequency acceleration components from the triaxial acceleration data as a motion noise reference sequence, uses adaptive filtering to offset body motion artifacts in the photoelectric pulse wave sequence, and calculates the pure pulse wave. The system analyzes the heart rate and blood oxygen saturation values ​​to determine the physiological health status. It extracts the helmet's self-vibration phase based on triaxial acceleration and triaxial angular velocity data, performs same-frequency phase compensation on the original phase of the intermediate frequency signal, analyzes the target approach velocity data, and generates a fall action indicator signal based on the triaxial acceleration and triaxial angular velocity data. When the fall action indicator signal is valid and the physiological health status is abnormal, an emergency alarm command is output. In step S6, when the emergency alarm command is received, the system uses the communication gateway module to call the voice call interface through the system bus, dials the preset emergency contact number, and sends map rendering coordinate data.

[0073] like Figure 2As shown, under normal conditions, when the heart rate exceeds the preset upper limit threshold or the blood oxygen saturation falls below the preset lower limit threshold, a timer is activated, entering a continuous delay verification window mechanism to determine the physiological health status. If the values ​​recover to the preset health threshold range within the preset verification time, the timer resets and returns to the normal state. If the duration exceeds the preset verification time, the physiological health status is set to an abnormal health state. The physiological sign monitoring unit has an independent abnormal health alert mechanism. When an abnormal health state is determined, regardless of whether a fall has occurred, a health alert pop-up is independently pushed to the WeChat mini-program on the mobile terminal. Simultaneously, the abnormal health state serves as an enable input for a fall + health binary arbitration, enabling the fall emergency alarm dialing to be adjusted by physiological health monitoring. When the fall action indicator signal is valid and the physiological health status is abnormal, an emergency alarm command is output, which then enters the system bus-level voice communication channel, dials the preset emergency contact number, and sends map rendering coordinate data. If the fall action indicator signal is valid but the physiological health status remains normal, it is determined to be a non-injury minor fall, and no emergency alarm command is output or voice call is triggered. At the same time, under normal conditions, when the distance data is less than the distance safety threshold and the approach speed data is greater than the speed safety threshold, an audible and visual alarm signal is output and the system enters the cooling timer operation period. During this period, repeated voice broadcasts are suppressed. If a sudden increase in approach speed or a sudden decrease in distance occurs, the system is forcibly reset and the cooling timer is interrupted, triggering a very high-priority voice alarm and outputting an audible and visual alarm signal.

[0074] Example 5: During the initial setup and first deployment of the protective helmet, the attitude inertial acquisition unit, physiological sign monitoring unit, rear radar detection unit, and fusion arbitration main control unit initiate a power-on self-test calibration procedure. When the helmet wearing detection sensor detects head contact and the riding speed is at a static 0 m / s, the fusion arbitration main control unit opens a 5-second installation attitude zero-bias calibration window, continuously reads the raw triaxial acceleration data and triaxial angular velocity data at a sampling rate of 100Hz, and calculates the mean value of the static gravitational acceleration projection to determine the zero point of the helmet's tilt angle relative to the orthogonal coordinate system. Subsequently, the alarm prompt component is driven to generate a weak mechanical pulse excitation, and the attitude inertial acquisition unit simultaneously acquires the vertical vibration acceleration waveform. The helmet's self-vibration transfer function under the current wearing tightness is calculated through the pulse response attenuation rate, and the mechanical coupling coefficient of the attitude radar is finely adjusted in situ. The floating-point storage value enables the helmet's self-oscillation phase. The estimation deviation is reduced to within the preset tolerance threshold, and the initial baseline alignment between the multimodal sensing physical layer interface and the wearer is completed.

[0075] Under long-term cycling and variable temperature and humidity conditions, the fusion arbitration main control unit adopts a baseline drift adaptive reconstruction mechanism based on a sliding time window. The system extracts historical millimeter-wave radar difference frequency signals and photoelectric pulse wave sequences in non-high-frequency jitter intervals at 10-minute intervals, and reduces the weight of historical data by combining an exponential attenuation factor to dynamically update the radar noise floor baseline envelope and photoelectric absorptivity DC component under the current environment. When the light-emitting diode of the multispectral photoplethysmography sensor module decreases in luminous efficiency due to long-term operation, resulting in a decrease of more than 15% in the DC component of the received reflected light intensity, the main control unit adaptively increases the pulse width of the light-emitting driving current to restore the intensity amplitude of 660nm red light and 940nm infrared light to the rated dynamic range. When the millimeter-wave echo difference frequency signal is attenuated due to rain or dust on the radar antenna surface, the main control unit linearly increases the difference frequency energy detection threshold according to the attenuation ratio. Under the premise of maintaining the distance safety threshold of 5m and the speed safety threshold of 2m / s unchanged, it avoids criterion drift and failure caused by sensor device aging and environmental medium adhesion.

[0076] Example 6: When the positioning unit is in an outdoor riding environment with the protective helmet, it extracts the latitude and longitude in degree format from the NMEA-0183 protocol text data output by the satellite positioning chip at a refresh rate of 1Hz, and converts it into decimal coordinates. and The fusion arbitration control unit employs a nonlinear coordinate offset algorithm to convert WGS-84 Earth coordinates to GCJ-02 Mars coordinates. and In the nonlinear coordinate offset algorithm, the Earth's semi-major axis is used as the reference. With eccentricity Given physical constants, calculate the radii of curvature of the meridian and the latitudinal circle, as well as the latitude and longitude offsets; when decimal coordinates are input... and At that time, the map rendering coordinate data output by the nonlinear coordinate offset algorithm is and The fusion arbitration control unit encapsulates the map rendering coordinate data into JSON format data frames and sends them to the WeChat mini-program mobile terminal via the WebSocket long connection network protocol; the mobile terminal calls the map interface to... and The endpoint coordinates are used for path planning and map rendering to eliminate map offset errors caused by coordinate system mismatch.

[0077] In the nonlinear coordinate offset calculation process, the fusion arbitration control unit first performs polynomial sine and cosine fitting expansion based on trigonometric functions on the input longitude and latitude decimal degrees to obtain preliminary geometric offset parameters in the longitude and latitude directions. Then, based on the current latitude radian value, it calculates the meridian curvature radius and the ramusoidal curvature radius under the first eccentricity. The preliminary geometric offset parameters are divided by the corresponding curvature radius and multiplied by the semicircular radian constant to convert the distance length order into a longitude and latitude change in degrees. Finally, the longitude and latitude changes are superimposed on the original WGS-84 Earth coordinate system longitude and latitude values ​​to obtain map rendering coordinate data in the GCJ-02 Mars coordinate system.

[0078] The communication gateway module runs on an embedded Linux system and integrates the DBus system bus and the ofono mobile communication framework. In the fall alarm and emergency call control, the fall call dialing is regulated by physiological health monitoring. When the physiological health status... The health monitoring module independently pops up a health alert pop-up on the WeChat mini-program, while the voice emergency call dialing is based on a combined fall and health assessment, only responding when the integrated arbitration control unit determines a fall action signal. And physical health status When an emergency alarm command is simultaneously fulfilled and output, the daemon process connects to the DBus system bus to obtain the org.ofono.Manager service interface and scans the mounted modem node paths. After finding the active modem path, it obtains its org.ofono.VoiceCallManager interface, directly calls the Dial method, and passes in the preset emergency contact phone number 15259016816. The ofono framework sends the ATD15259016816 dialing command to the underlying modem to establish a system bus-level voice call channel. This dialing process is directly completed by the DBus system bus, the ofono mobile communication framework, and the underlying modem. The emergency call response latency is kept within 138ms of the DBus bus communication cycle.

[0079] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0080] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A multi-sensor fusion intelligent cycling protection system, characterized in that, include: Attitude inertial acquisition unit, used to acquire three-axis acceleration data and three-axis angular velocity data; Physiological signs monitoring unit, used to acquire photoelectric pulse wave sequences; The rear radar detection unit is used to acquire the intermediate frequency signal of the frequency-modulated continuous wave millimeter-wave radar; Positioning unit, used to acquire satellite positioning coordinate data; The integrated arbitration control unit is connected to the attitude inertial acquisition unit, physiological sign monitoring unit, rear radar detection unit, position positioning unit, and communication gateway module, respectively. It is used to convert satellite positioning coordinate data into map rendering coordinate data; extract the high-frequency acceleration component from the three-axis acceleration data as a motion noise reference sequence; use adaptive filtering to offset body motion artifacts in the photoelectric pulse wave sequence; calculate the pure pulse wave signal; and analyze the heart rate and blood oxygen saturation values ​​to determine the physiological health status. The helmet's natural phase is extracted based on triaxial acceleration and triaxial angular velocity data. The original phase of the intermediate frequency signal is compensated for with the same frequency, and the target approach velocity data is analyzed. A fall action indicator signal is generated based on triaxial acceleration data and triaxial angular velocity data. When the fall action indicator signal is valid and the physiological health status is abnormal, an emergency alarm command is output. The communication gateway module is used to call the voice call interface through the system bus when an emergency alarm command is received, dial the preset emergency contact number and send map rendering coordinate data.

2. The multi-sensor fusion intelligent cycling protection system according to claim 1, characterized in that, The fusion arbitration control unit is also used to calculate the cornering centrifugal acceleration vector by combining the riding speed obtained by the positioning unit with the yaw rate in the three-axis angular velocity data, and to subtract the cornering centrifugal acceleration vector from the three-axis acceleration data to restore the gravitational acceleration projection vector. The fusion arbitration control unit calculates the result using the formulas for the sum of three-axis acceleration and the sum of three-axis angular velocity. When the sum of three-axis acceleration is lower than the low-order weightlessness threshold or higher than the high-order impact threshold, and the sum of three-axis angular velocity is higher than the rotational angular velocity threshold, a fall action indicator signal is generated, and the effective duration of the fall action indicator signal is set to 10 seconds.

3. The multi-sensor fusion intelligent cycling protection system according to claim 2, characterized in that, The low-order weightlessness threshold is 0.4g, the high-order impact threshold is 1.6g, and the rotational angular velocity threshold is 250dps. The fusion arbitration main control unit is used to set the fall action flag signal to a high level when the sum of the three-axis acceleration is lower than 0.4g or higher than 1.6g, and the sum of the three-axis angular velocities is higher than 250dps.

4. The multi-sensor fusion intelligent cycling protection system according to claim 1, characterized in that, The determination of physiological health status includes a continuous delay verification window mechanism; The integrated arbitration control unit compares the analyzed heart rate and blood oxygen saturation values ​​with preset health thresholds. When the heart rate value exceeds the preset upper limit threshold or the blood oxygen saturation value is lower than the preset lower limit threshold, a timer is started. If the value exceeds the preset health threshold and the duration exceeds the preset verification time, the fusion arbitration control unit will set the physiological health status to an abnormal health status. If the value recovers to the preset health threshold range within the preset verification time, the timer is reset and the physiological health status remains normal. The preset verification time is set to 4 seconds.

5. The multi-sensor fusion intelligent cycling protection system according to claim 1, characterized in that, The rear radar detection unit uses a 24GHz frequency-modulated continuous wave millimeter-wave radar. Before analyzing the target's approach velocity data, the fusion arbitration main control unit extracts the vertical vibration acceleration and yaw rate of the attitude inertial acquisition unit, calculates the helmet's natural phase, and performs same-frequency phase compensation on the original phase of the intermediate frequency signal to eliminate the micro-Doppler spectrum broadening caused by helmet micro-vibration. The fusion arbitration control unit performs spectrum analysis on the compensated clean intermediate frequency signal to extract the distance and approach speed data of the target behind.

6. The multi-sensor fusion intelligent cycling protection system according to claim 5, characterized in that, The system also includes an alarm prompting component connected to the fusion arbitration main control unit; when the extracted distance data is less than the distance safety threshold of 5.0m and the approach speed data is greater than the speed safety threshold of 2m / s, the fusion arbitration main control unit drives the alarm prompting component to output an audible and visual alarm signal, and starts a 3s cooling timer after triggering the warning, suppressing repeated voice broadcasts during the 3s cooling timer operation.

7. The multi-sensor fusion intelligent cycling protection system according to claim 1, characterized in that, The communication gateway module runs on an embedded Linux system and integrates the DBus system bus and the ofono mobile communication framework. When the communication gateway module receives an emergency alarm command, it obtains the ofono management interface through the DBus system bus and scans for available modems. It then directly calls the dialing method of the voice call management interface to pass in the preset emergency contact phone number and establishes a system bus-level voice call channel.

8. The multi-sensor fusion intelligent cycling protection system according to claim 1, characterized in that, The location unit extracts latitude and longitude in degree-division format from the satellite positioning coordinate data and converts it into decimal coordinates. The fusion arbitration control unit uses a nonlinear coordinate offset algorithm to convert the decimal coordinates in the WGS-84 Earth coordinate system into map rendering coordinate data in the GCJ-02 Mars coordinate system, and synchronizes the map rendering coordinate data to the mobile terminal via a long connection network protocol.

9. The multi-sensor fusion intelligent cycling protection system according to claim 1, characterized in that, The physiological sign monitoring unit includes a multispectral photoplethysmography (PPG) sensor module that fits onto the rider's scalp. The PPG sensor module uses a 660nm red light emitter and a 940nm infrared light emitter to alternately irradiate subcutaneous microvessels, and a photodetector collects the reflected light intensity change sequence to output a photoelectric pulse wave sequence. The PPG sensor module is connected to the fusion arbitration main control unit via a universal asynchronous transceiver serial port.

10. A multi-sensor fusion intelligent cycling protection method, used to operate the multi-sensor fusion intelligent cycling protection system of claim 1, characterized in that, Includes the following steps: Step S1: Acquire triaxial acceleration data and triaxial angular velocity data through the attitude inertial acquisition unit; Step S2: Acquire photoelectric pulse wave sequence through physiological sign monitoring unit; Step S3: Acquire the intermediate frequency signal of the frequency-modulated continuous wave millimeter-wave radar through the rear radar detection unit; Step S4: Obtain satellite positioning coordinate data through the positioning unit; Step S5: The satellite positioning coordinate data is converted into map rendering coordinate data by the fusion arbitration main control unit; the high-frequency acceleration component in the three-axis acceleration data is extracted as the motion noise reference sequence; adaptive filtering is used to offset the body motion artifacts in the photoelectric pulse wave sequence; the pure pulse wave signal is calculated; and the heart rate value and blood oxygen saturation value are analyzed to determine the physiological health status. The helmet's natural phase is extracted based on triaxial acceleration and triaxial angular velocity data. The original phase of the intermediate frequency signal is compensated for with the same frequency, and the target approach velocity data is analyzed. A fall action indicator signal is generated based on triaxial acceleration data and triaxial angular velocity data. When the fall action indicator signal is valid and the physiological health status is abnormal, an emergency alarm command is output. Step S6: Upon receiving an emergency alarm command, the system bus is used through the communication gateway module to call the voice call interface, dial the preset emergency contact number, and send map rendering coordinate data.

Images