Driver heart monitoring system based on non-contact electrocardio and visual pulse waves

Abstract

The invention discloses a system for realizing heart monitoring in a driving state. The system integrates two signal channels of a non-contact electrocardiogram (ECG) electrode and a visual pulse wave (PPG). The non-contact electrocardio-electrodes are integrated in a safety belt structure and located on the left shoulder and the belly of a driver respectively, and high-precision electrocardio-signals are collected through clothes. The steering wheel is made of weak conductive materials to form a right leg driving circuit. The camera is installed in front of a driver, recognizes the face in real time and extracts visual pulse wave signals. The multi-mode signal realizes cardiac rhythm synchronous analysis through a fusion algorithm, so that the signal stability and accuracy are enhanced, and the method is particularly suitable for long-term monitoring and potential abnormal cardiac rhythm recognition. The invention further provides a self-adaptive channel selection strategy: when the quality of the non-contact electrocardiosignal is poor, the visual PPG channel is automatically switched and used, so that the monitoring continuity and accuracy in the driving environment are guaranteed.

Description

[0001] The field

[0002] The present application belongs to the field of intelligent medical monitoring and ubiquitous computing, and particularly relates to a multi-modal physiological signal acquisition system for an intelligent safe driving scene. Specifically, the present application integrates non-contact electrocardiogram (ECG) monitoring and non-contact visual pulse wave (PPG) technology, and realizes high-precision heart rate monitoring through multi-modal signal fusion and quality assessment. BACKGROUND

[0003] The current driver heart rate monitoring technology mainly has the following technical bottlenecks: 1) The traditional contact ECG monitoring can provide clinical level RR interval data, but needs electrode attachment, which seriously affects the driving operation comfort; 2) The pure visual PPG scheme (such as CN119992258B) realizes non-contact monitoring, but its millisecond level heart rate variability (HRV) analysis precision is restricted by environmental light, head movement and other factors; 3) The monitoring technology based on millimeter wave radar indirectly obtains heart rate signal through body surface vibration, and the signal-to-noise ratio is significantly reduced in the vehicle driving vibration environment.

[0004] The existing technical solutions have obvious performance limitations in driving scene applications: single modal monitoring is difficult to balance measurement accuracy and use convenience, and lacks self-adaptive ability to complex driving environment. Especially in long-distance driving and other scenes that need continuous monitoring, the existing single modal system cannot guarantee the clinical level reliability and time resolution of heart rate data. Therefore, developing a new type of monitoring system that can integrate the advantages of multi-source physiological signals and has dynamic signal quality evaluation function has become a key technical challenge to improve the driving safety monitoring efficiency. SUMMARY

[0005] Based on the above background, the present application relates to a driver heart monitoring system based on non-contact ECG and visual PPG.

[0006] The specific scheme is as follows:

[0007] A driver heart monitoring system based on non-contact ECG and visual pulse wave, comprising a non-contact ECG electrode arranged in a safety belt, a steering wheel conductive unit with an ECG right leg driving circuit, a high impedance circuit unit, a camera installed in front of the driver, a frame synchronization module for non-contact ECG and camera synchronous acquisition, and an intelligent computing unit for heart rate calculation.

[0008] Further, the non-contact ECG electrode is combined with a high-impedance circuit unit for clothing-collected ECG signals, which is integrated in a safety belt structure and located at the left shoulder and abdomen of the driver, respectively, to collect high-precision ECG signals through clothing (the non-contact ECG electrode collects ECG signals through clothing in a non-contact manner); the high-impedance circuit unit is made of an ECG high-impedance analog front end made of a high-impedance voltage follower, or a charge amplifier, or an instrument amplifier with extremely high input impedance, which is inputted on the non-contact ECG electrode, and high-quality ECG signals are obtained through differential amplification and filtering.

[0009] Further, the steering wheel conductive unit constitutes an ECG right leg drive circuit, which reduces common-mode interference through negative feedback and improves the signal-to-noise ratio of the ECG signal, wherein the steering wheel as a whole is made of conductive fibers or conductive metals; when in the driving state, the driver's hands directly contact the steering wheel conductive unit to form a negative feedback loop, thereby obtaining high-quality non-contact ECG signals (the right leg drive electrode directly contacts the human hand, which can effectively avoid the problem of unstable voltage caused by impedance fluctuation of the negative feedback voltage, and significantly reduces the common-mode voltage through stable common-mode negative feedback).

[0010] Further, the camera mounted in front of the driver collects the driver's face video, and the visual PPG signal is extracted through image processing; wherein the camera has a frequency of no less than 30 fps and a pixel resolution of more than 484x274 to ensure the time resolution of the visual PPG signal.

[0011] Further, the frame synchronization module adopts a unified clock control and combines hardware fast cache and synchronous sampling markers to ensure the consistency of the non-contact ECG signal and the video frame in the time dimension, thereby realizing the real-time synchronization of ECG and PPG.

[0012] Further, the intelligent computing unit includes an ECG signal quality evaluation module, a visual PPG signal quality evaluation module, and a heart rate fusion calculation module, which are used for quality judgment, fusion, and real-time heart rate output of the two signals.

[0013] Further, the extraction process of the visual PPG signal is as follows: first, face detection and ROI selection are performed, and a plurality of candidate ROIs in the detected driver face region are scored; for the pixels in the tracked ROI region of each frame, first, the RGB channel average is performed, and the weighted POS algorithm is used for orthogonal filtering processing of the RGB three-channel signal to obtain a pure PPG time domain signal; finally, the current heart rate value is calculated based on the signal.

[0014] Further, the electrocardiosignal quality evaluation module evaluates the signal quality of the non-contact electrocardio based on the frequency domain interference energy ratio, the time domain wave peak consistency index, focuses on the high frequency vibration noise of the driving environment of 80-100Hz, the power frequency interference of 45-55Hz, and the capacitive coupling type respiratory motion artifact interference of 0.1-0.5Hz; the time domain wave peak consistency adopts multi-QRS positioning, and the signal quality is evaluated by taking the consistency of the QRS number and QRS position parameters as an index.

[0015] Further, the visual PPG signal quality evaluation module evaluates the PPG signal quality based on the video micro-motion feature, the PPG energy spectrum distribution feature and the optical flow interference feature.

[0016] Further, when the non-contact electrocardio signal quality is better than the preset threshold, the heart rate calculated by the electrocardio signal is preferentially output; when the electrocardio signal quality is lower than the threshold and the visual PPG signal quality meets the requirement, the heart rate is calculated by switching to the visual PPG channel; if both the signals are lower than the threshold, an invalid prompt is output.

[0017] Compared with the prior art, the beneficial effects of the present application patent are:

[0018] 1. The non-contact electrode and capacitive coupling method are adopted to collect the non-contact ECG signal, and the electrocardio signal can be collected through clothes. The non-invasive and patch-free collection is realized, and the heart rate monitoring accuracy reaches the millisecond level.

[0019] 2. The non-contact electrocardio and RGB camera fusion method is adopted to extract the heart rate information. When the non-contact electrocardio signal quality is poor, the PPG signal is extracted by the RGB camera, and the skin color detection, ROI scoring, KCF tracking and POS algorithm are relied on to realize the high-robustness visual heart rate extraction.

[0020] 3. The signal quality evaluation and channel switching strategy are introduced to ensure the continuity and accuracy of the heart rate output. BRIEF DESCRIPTION OF DRAWINGS

[0021] Figure 1 It is a system schematic diagram of the present application;

[0022] Figure 2 It is a camera visual PPG signal extraction flowchart in the present application;

[0023] Figure 3 It is a signal quality evaluation and heart rate calculation flowchart in the present application. DETAILED DESCRIPTION

[0024] The present application will be described in detail below in combination with the drawings and implementation examples.

[0025] As Figure 1As shown, the present invention relates to a driver heart monitoring system based on non-contact electrocardiogram and visual PPG, comprising: non-contact electrocardiogram electrodes integrated in the seat belt, a right leg drive circuit disposed on the conductive layer of the steering wheel, a high-impedance front-end analog circuit module, a camera module installed in front of the driver, and an intelligent computing unit for multi-channel signal fusion and heart rate calculation.

[0026] In this embodiment, the non-contact ECG electrodes are made of conductive fabric and sewn inside the seatbelt on the driver's left shoulder and left abdomen, corresponding to the upper left and lower left positions of the heart, respectively. ECG signals are acquired using the principle of capacitive coupling. Since there is no direct contact with the skin, wearing comfort and acceptability are improved, making it particularly suitable for continuous monitoring needs in daily driving scenarios.

[0027] The ECG signals acquired by the electrodes are fed into a high-impedance front-end analog processing module. This module includes an analog front-end circuit with extremely high input impedance, consisting of a voltage follower or charge amplifier. The signals are then amplified first and then bandpass filtered by a high-impedance instrumentation amplifier to enhance the signal amplitude, suppress common-mode interference and low-frequency drift, and provide a stable and identifiable QRS complex structure for subsequent digital analysis modules.

[0028] The system controls both the ECG sampling module and the camera sampling module via a unified clock, with each module bearing a timestamp and written to a high-speed cache. The intelligent computing unit matches the timestamps to ensure the alignment of the ECG signal and the PPG signal on the time axis, enabling synchronous analysis.

[0029] like Figure 1 As shown in the steering wheel position, to improve the system's common-mode rejection ratio, the steering wheel surface is covered with a conductive material, allowing it to naturally contact the driver's hand or right leg while driving, forming a right leg drive circuit. This circuit couples common-mode interference signals back to the driver's body surface through negative feedback, suppressing external electromagnetic disturbances from interfering with the electrocardiogram signal.

[0030] The camera is installed on the dashboard or near the rearview mirror, facing the driver's face. The image acquisition frequency is ≥30 frames / second, and the resolution is not less than 484×274 pixels to ensure that sufficient time and space information is obtained to extract the visual PPG signal.

[0031] like Figure 2 As shown, the video images captured by the camera are processed by the intelligent computing unit. First, a face detection algorithm (such as MTCNN or RetinaFace based on deep convolutional neural networks) is called to locate the facial region. Then, facial regions such as the forehead and cheeks are selected as candidate ROIs, scored according to indicators such as brightness uniformity and texture consistency, and the optimal ROI is dynamically selected as the signal acquisition window.

[0032] For each candidate ROI, calculate the brightness uniformity U and texture consistency T, and select the optimal ROI based on a weighted scoring function:

[0033] Score = γ1·U + γ2·T

[0034] Where γ1 and γ2 are weighting parameters. The system dynamically selects the ROI with the highest score as the signal acquisition window.

[0035] To mitigate the impact of subtle head movements by the driver, the Kernelized Correlation Filters (KCF) algorithm is employed for inter-frame tracking of selected regions of interest (ROIs) to achieve stable facial region locking. In each frame, the RGB pixels of the ROI region are extracted and averaged across channels to obtain the following: Where N is the number of pixels within the ROI, R i (t),G i (t),B i (t) represents the red, green, and blue components of a pixel, respectively. The three-channel signals are combined into a matrix: Next, the weighted Plane-Orthogonal-to-Skin (POS) algorithm is applied to orthogonally project and weighted fuse the RGB three-channel signals, thereby removing common-mode interference such as lighting changes and extracting the original pulse wave signal corresponding to PPG motion. POS (t)=w1·(r(t)-g(t))+w2·(r(t)+g(t)-2b(t)), where w1 and w2 are channel weighting coefficients used to suppress common-mode interference from illumination. The algorithm incorporates overexposure detection, applying low weights to bright areas to improve illumination robustness. For pixels within the ROI with brightness I… i (t) exceeds the preset threshold I th Therefore, a low weight is applied when calculating the channel mean:

[0036] Where η < 1, it is used to reduce the interference weight of overexposed pixels, thereby improving illumination robustness. Finally, S is obtained after the above processing. POS (t) represents the raw PPG time-domain signal related to the heartbeat, which is used for subsequent heart rate and signal quality assessment.

[0037] The obtained PPG signal enters the frequency domain analysis process. The system uses a Hanning window for time-domain smoothing and performs a Fast Fourier Transform (FFT) to extract the dominant frequency component from the spectrum and convert it into heart rate (unit: BPM). To enhance accuracy, the system can combine autocorrelation spectrum analysis and historical heart rate trend data to filter and correct abnormal spectral peaks.

[0038] The non-contact ECG signal quality assessment module evaluates signal quality based on metrics such as frequency domain interference energy ratio and time domain peak consistency. It pays particular attention to high-frequency vibration noise in the driving environment (80–100 Hz), power frequency interference (45–55 Hz), and capacitively coupled respiratory motion artifacts (0.1–0.5 Hz). Time domain peak consistency is assessed using multi-QRS localization, with the number of QRSs and the consistency of their position parameters serving as indicators for signal quality evaluation.

[0039] The specific implementation of the frequency domain interference energy ratio includes: calculating the power spectral density P of the original non-contact ECG signal x. x (f), where f is the frequency; by considering P x (f) Integrate the energy E in the 45Hz–55Hz frequency band to obtain the energy of this frequency band. 45-55 Non-contact ECG is highly susceptible to interference in the 45-55Hz frequency range, with interference energy in this band being [missing value]. By analyzing P x (f) Integrating over the respiratory signal frequency band (typically 0.1–0.5 Hz) yields the respiratory signal energy E_breath, i.e. E_breath can effectively capture respiratory coupling interference in seatbelt-style non-contact ECG. Through P x (f) Integrate across the entire signal frequency band to obtain the total energy E_total, i.e. And high-frequency vibration energy of 80-100Hz in the driving environment, i.e. Calculate the spectral ratio The specific implementation of time-domain peak consistency involves using five or more QRS wave monitors to monitor the R-wave position simultaneously. The time-domain peak consistency index is defined as:

[0040]

[0041] Where N matched(k，w) It is a consistent heart rate (within 200ms), while N matched (k, w) represents the number of heartbeats detected by the nth QRS detection algorithm. These parameters are input into a pre-trained signal quality model for signal quality assessment. The non-contact ECG signal quality assessment model is a pre-trained model using a training dataset. Specifically, this model can employ machine learning algorithms to achieve non-linear, efficient, and lightweight signal quality assessment. The results can include quality levels, such as three levels (good, medium, and poor) or two levels (good and poor).

[0042] The visual PPG signal quality assessment module is based on video micro-motion characteristics (SQI). motion PPG energy spectrum distribution characteristics (SQI)spec and optical flow interference characteristics SQI flow The PPG signal quality was evaluated. Three characteristics were defined as follows: Where, σ motion σ is defined as very subtle changes in pixel brightness within the facial region of interest (ROI) of a video frame. These changes may originate from: ambient light variations / minor facial jitter (useless noise). motion Defined as:

[0043] I i (t) refers to the brightness of the i-th pixel in the ROI region at time t. This refers to the average value of all pixels within the ROI, which is the average brightness. N is the sum of all pixels. E 0.7-3Hz The useful energy spectrum defined as the visual PPG in the 0.7–3 Hz range, E 0.1-10Hz This refers to the total energy obtained by integrating the signal across the entire frequency band. ΔF is defined as the optical flow vector field within each frame of the facial ROI calculated using an optical flow algorithm. Then, the mean of the optical flow is calculated to measure the overall motion amplitude: N refers to the number of pixels.

[0044] Similarly, the aforementioned visual PPG signal quality parameters are input into a pre-trained signal quality model for signal quality evaluation. The non-contact ECG signal quality evaluation model is a pre-trained model using a training dataset. Specifically, this model can employ machine learning algorithms to achieve non-linear, efficient, and lightweight signal quality evaluation, with results including quality levels such as good, medium, and poor, or good and poor.

[0045] When the ECG signal does not meet preset conditions, the system automatically switches to the visual PPG signal channel for heart rate calculation. Furthermore, the system features a dual-channel fusion mechanism. When both the non-contact ECG and visual PPG signals are valid simultaneously, and the difference in output heart rate values ​​is within 5 BPM, the system can weight the output heart rate based on the quality indicators of the two signals, improving overall robustness and continuity. Fusion strategies may include confidence-weighted averaging, maximum confidence optimization, or Kalman filtering fusion.

[0046] The final heart rate output can be displayed on the vehicle's infotainment system or uploaded wirelessly (e.g., via Bluetooth or 4G module) to a remote medical platform or mobile app. All data supports long-term trend recording, abnormal alerts, and remote doctor intervention.

[0047] In summary, the multimodal driver heart monitoring system proposed in this invention has good practicality, reliability and deployability, and is especially suitable for continuous monitoring and alarm of arrhythmia in autonomous driving and assisted driving scenarios.

[0048] The technical means disclosed in this invention are not limited to those disclosed in the above embodiments, but also include technical solutions composed of any combination of the above technical features. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of this invention, and these improvements and modifications are also considered within the scope of protection of this invention.

Claims

1. A driver cardiac monitoring system based on non-contact electrocardiogram and visual pulse wave, characterized in that: The non-contact electrocardio electrode is arranged in a safety belt, the steering wheel conductive unit has a right leg driving circuit of electrocardio, the high impedance circuit unit, the camera installed in front of the driver, the frame synchronization module for synchronous acquisition of the non-contact electrocardio and the camera, and the intelligent calculation unit for heart rate calculation.

2. The driver cardiac monitoring system based on non-contact electrocardiogram and visual pulse wave according to claim 1, characterized in that: The non-contact electrocardio electrode is combined with the high impedance circuit unit to collect electrocardio signals through clothes, is integrated in a safety belt structure, is located at the left shoulder and the abdomen of the driver respectively, and collects high-precision electrocardio signals through clothes; the high impedance circuit unit is made of a high impedance voltage follower, or a charge amplifier, or an electrocardio high impedance analog front end of an instrument amplifier with extremely high input impedance, inputs the non-contact electrocardio electrode, and obtains high-quality electrocardio signals through differential amplification and filtering.

3. The driver cardiac monitoring system based on non-contact electrocardiogram and visual pulse wave according to claim 1, characterized in that: The steering wheel conductive unit constitutes a non-contact electrocardio right leg driving circuit, reduces common mode interference through negative feedback, and improves the signal-to-noise ratio of electrocardio signals, wherein the steering wheel is made of conductive fibers or conductive metals; when the driver's hands are in direct contact with the steering wheel conductive unit to form a negative feedback loop in the driving state, high-quality non-contact electrocardio signals are obtained.

4. The driver cardiac monitoring system based on non-contact electrocardiogram and visual pulse wave according to claim 1, characterized in that: The camera installed in front of the driver collects the video of the face of the driver, extracts visual PPG signals through image processing; wherein the camera has a frequency of no less than 30 fps and a pixel resolution of more than 484*274, so as to ensure the time resolution of the visual PPG signals.

5. The driver cardiac monitoring system based on non-contact electrocardiogram and visual pulse wave according to claim 1, characterized in that: The frame synchronization module is controlled by a unified clock, combines hardware fast cache and synchronous sampling marks, ensures the consistency of the non-contact electrocardio signals and the video frames in the time dimension, and realizes the real-time synchronization of electrocardio and visual pulse waves.

6. The driver cardiac monitoring system based on non-contact electrocardiogram and visual pulse wave according to claim 1, characterized in that: The intelligent calculation unit includes an electrocardio signal quality evaluation module, a visual PPG signal quality evaluation module, and a heart rate fusion calculation module, and is used for quality determination, fusion and real-time heart rate output of two signals.

7. The driver cardiac monitoring system based on non-contact electrocardiogram and visual pulse wave according to claim 6, characterized in that: The extraction process of the visual pulse wave signal is as follows: first, face detection and ROI selection are performed, a plurality of candidate ROIs are scored in the detected face region of the driver, for the pixels in each frame ROI region tracked, first, the RGB channel average is performed, and the weighted POS algorithm is used to perform orthogonal filtering processing on the RGB three-channel signals to obtain pure PPG time domain signals; finally, the current heart rate value is calculated based on the signals. Specifically, the following is performed: Firstly, a face detection algorithm is called to locate the face region; then, the face region is selected as a candidate ROI, the brightness uniformity and texture consistency indexes are scored, and the optimal ROI is dynamically selected as a signal acquisition window; The brightness uniformity U and the texture consistency T are calculated for each candidate ROI, and the optimal ROI is selected based on a weighted scoring function: Score = γ1·U + γ2·T Wherein γ1, γ2 are weight parameters; the ROI with the highest score is dynamically selected as a signal acquisition window. To alleviate the impact of the driver's slight head movement, the KCF algorithm is used to track the selected ROI between frames, and the face region is stably locked. In each frame of image, the RGB pixels of the ROI region are extracted and averaged by channel to obtain: where N is the number of pixels in the ROI, R i (t), G i (t), and B i (t) are the red, green, and blue components of the pixel, respectively; and the three channels are combined into a matrix: Then, the weighted POS algorithm is applied to the RGB three-channel signal for orthogonal projection and weighted fusion, so as to remove common-mode interference such as illumination change and extract the PPG dynamic pulse wave original signal; S POS (t) = w1- (r(t) - g(t)) + w2- (r(t) + g(t) - 2b(t)) where w1, w2 are channel weighting coefficients, used to suppress the illumination common mode interference; combined with overexposure detection, low weight is applied to highlight regions to improve illumination robustness, for pixels within ROI if intensity I i (t) exceeds a preset threshold I th low weight is applied in channel mean calculation: Wherein η<1, used to reduce the interference weight of overexposed pixels, so as to improve the illumination robustness; Finally, the S POS (t) is the raw PPG time-domain signal related to heartbeats, which is used for subsequent heart rate and signal quality assessment. The obtained PPG signal enters the frequency domain analysis process, uses Hanning window for time domain smoothing and performs fast Fourier transform FFT, extracts the main frequency component in the frequency spectrum and converts it into heart rate; In order to enhance the accuracy, the autocorrelation spectrum analysis and historical heart rate trend data are combined to filter and correct the abnormal spectrum main peak.

8. The driver cardiac monitoring system based on non-contact electrocardiogram and visual pulse wave according to claim 6, characterized in that: The electrocardiosignal quality evaluation module evaluates the signal quality of the non-contact electrocardio based on the frequency domain interference energy ratio and the time domain wave peak consistency index, focuses on the high frequency vibration noise of 80-100Hz in the driving environment, the power frequency interference of 45-55Hz, and the capacitive coupling type respiratory motion artifact interference of 0.1-0.5Hz; The time domain wave peak consistency uses multiple QRS positioning, and the QRS number and QRS position parameter consistency are used as indexes for signal quality evaluation; Specifically: The frequency domain interference energy ratio includes: calculating the power spectral density P of the original non-contact ECG signal x. x (f), where f is the frequency; by considering P x (f) Integrate the energy E in the 45Hz–55Hz frequency band to obtain the energy of this frequency band. 45-55 Non-contact ECG is highly susceptible to interference in the 45-55Hz frequency range, with interference energy in this band being [missing value]. By analyzing P x (f) Integrating over the respiratory signal frequency band yields the respiratory signal energy E_breath, i.e. E_breath can effectively capture respiratory coupling interference in seatbelt-style non-contact ECG; by analyzing P x (f) Integrate across the entire signal frequency band to obtain the total energy E_total, i.e. And high-frequency vibration energy of 80-100Hz in the driving environment, i.e. Calculate the spectral ratio: The time domain wave peak consistency includes: using more than 5 QRS wave monitors to monitor the R wave position, and detecting the R wave position at the same time; The time domain wave peak consistency index is defined as: wherein N matched(k，w) is the consistent heart rate, and N matched (k, w) is the heart rate detected by the nth QRS detection algorithm; the above parameters are input into a pre-trained signal quality model for signal quality assessment; wherein the non-contact electrocardiogram signal quality assessment model is a signal quality assessment model pre-trained using a training data set, specifically, the signal quality assessment model uses a machine learning algorithm to achieve non-linear, efficient and lightweight signal quality assessment, and the result includes a quality level.

9. The driver cardiac monitoring system based on non-contact electrocardiogram and visual pulse wave according to claim 6, characterized in that: The visual PPG signal quality evaluation module evaluates the PPG signal quality based on a video micro-motion feature, a PPG energy spectrum distribution feature and an optical flow interference feature; specifically, the PPG signal quality is evaluated based on a video micro-motion feature SQI motion , a PPG energy spectrum distribution feature SQI spec and an optical flow interference feature SQI flow . Three features are defined as: where σ motion is defined as the very slight pixel intensity variation within the face ROI; the formula is as follows: where I i (t) refers to the brightness of the i-th pixel in the ROI region at time t; refers to the average value of all the ROI pixels, i.e. the average brightness; E 0.7-3Hz defined as the useful energy spectrum of 0.7-3 Hz of visual PPG, E 0.1-10Hz refers to the total energy obtained by integrating the full frequency band of the signal; ΔF is defined as the optical flow vector field of each frame in the face ROI calculated using the optical flow algorithm Then the average of the optical flow is calculated to measure the overall motion amplitude: N refers to the number of pixels; The above visual PPG signal quality parameters are input into the pre-trained signal quality model for signal quality evaluation; wherein the non-contact electrocardio signal quality evaluation model is a signal quality evaluation model pre-trained using a training data set, specifically, the signal quality evaluation model uses a machine learning algorithm to realize nonlinear, efficient and lightweight signal quality evaluation, and the result includes quality level.

10. The driver cardiac monitoring system based on non-contact electrocardiogram and visual pulse wave according to claim 1, characterized in that: When the non-contact electrocardio signal quality is better than the preset threshold, the heart rate calculated by the electrocardio signal is preferentially output; when the electrocardio signal quality is lower than the threshold and the visual PPG signal quality meets the requirements, the heart rate calculated by the visual PPG channel is switched to; if both signals are lower than the threshold, an invalid prompt is output.

Images