Driving fatigue monitoring system and method based on flexible fabric electrode

Abstract

The invention relates to the technical field of intelligent wearing and health monitoring, in particular to a driving fatigue monitoring system based on a flexible fabric electrode, which comprises fatigue monitoring clothes, a data acquisition box, a trained fatigue monitoring model based on a multi-modal physiological signal and a terminal output platform, the fatigue monitoring garment comprises a garment body, a flexible fabric electrode, a flexible magnetic strip and magnetic attraction type conductive rubber, and the flexible fabric electrode is connected with the data acquisition box through the magnetic attraction type conductive rubber and the flexible magnetic strip in an attraction mode. The driving fatigue monitoring system has the advantage of non-sensitive wearing, a flexible fabric electrode seamless integration technology is adopted, and bio-compatible electrodes are embedded into the inner layer of the clothes through a sewing process; by means of the design, the problem that a traditional hard electrode mechanically compresses the skin is solved; and the flexible fabric electrode is connected with the data acquisition box in an automatic adsorption manner, so that the problem of interference on driving attention caused by manual cable plugging and pulling in a traditional scheme is avoided.

Description

Technical Field

[0001] This application relates to the field of smart wearable and health monitoring technology, and in particular to a driving fatigue monitoring system based on flexible fabric electrodes, and a monitoring method for the driving fatigue system based on flexible fabric electrodes. Background Technology

[0002] In recent years, multimodal fusion methods have improved monitoring accuracy by combining signals such as electrocardiogram, body temperature, and body movement. However, their hardware design still has drawbacks: rubber electrodes or metal patches compress the chest cavity, causing strong rejection by the driver; wired transmission requires manual plugging and unplugging of the device, which distracts the driver's attention; and rigidly packaged sensors produce motion artifacts when the vehicle makes sharp turns, with a signal distortion rate of up to 40%.

[0003] To address the aforementioned issues, there is an urgent need for an innovative solution that balances seamless wearability with precise monitoring. In this regard, the present invention proposes a driver fatigue monitoring system based on flexible fabric electrodes, and a monitoring method for the driver fatigue system based on flexible fabric electrodes. Summary of the Invention

[0004] To address the problems in the prior art, the present invention proposes the following technical solution.

[0005] On the one hand, a driving fatigue monitoring system based on flexible fabric electrodes is provided, including a fatigue monitoring garment, a data acquisition box, a trained fatigue monitoring model based on multimodal physiological signals, and a terminal output platform; The fatigue monitoring garment includes a garment, a flexible fabric electrode, a flexible magnetic strip, and a magnetically conductive rubber. The garment has a pocket at the chest position. The flexible fabric electrode is in close contact with the skin to collect electrocardiogram signals and is seamlessly integrated into the garment by sewing. The flexible magnetic strip is installed on the back of the flexible fabric electrode, and the magnetically conductive rubber is installed on the back of the data acquisition box. The data acquisition box is placed in a pocket. The magnetic conductive rubber on the back of the data acquisition box attracts the flexible magnetic strip on the back of the flexible fabric electrode. The flexible fabric electrode is connected to the data acquisition box. The data acquisition box realizes non-invasive acquisition of heart rate, heart rate variability, body temperature, and body movement physiological parameters, and sends them to the fatigue monitoring method and terminal output platform based on multimodal physiological signals. The trained fatigue monitoring model based on multimodal physiological signals is used to acquire multimodal physiological signal data from the fatigue monitoring garment and process the data to obtain fatigue monitoring results. The terminal output platform has functions such as user information input, multimodal physiological signal fusion analysis, dynamic fatigue assessment, and early warning push, which users can view and analyze in real time.

[0006] As a preferred embodiment of the above technical solution, the flexible fabric electrode is made by hot pressing a conductive layer woven from highly conductive silver-plated nylon fibers and a comfortable and breathable flexible polymer insulating base, and the flexible fabric electrode is seamlessly embedded into the inner layer of the garment using a high-precision overlock sewing process.

[0007] As a preferred embodiment of the above technical solution, the flexible fabric electrode is provided in three sets, two of which are fixed to the shoulders of the garment, and one set is fixed to the connection between the flexible fabric electrode on the shoulders and the pocket where the data acquisition box is located.

[0008] As a preferred embodiment of the above technical solution, the data acquisition box includes a heart rate monitoring module, a temperature monitoring module, an acceleration monitoring module, a wireless transmission module, and a power supply module; The core component of the heart rate monitoring module is the BMD101 heart rate sensor, which has a built-in high-precision bioelectrical signal amplifier. It can capture weak electrocardiogram signals and execute the R-wave detection algorithm through the built-in hard-wired digital signal processor to output heart rate and heart rate variability parameters in real time. The core component of the temperature monitoring module is the NST1001 temperature sensor. The probe of the NST1001 temperature sensor is attached to the inside of the shell via thermally conductive silicone. The temperature value is calculated based on the number of pulses, which can indirectly sense the human body surface temperature and ensure real-time capture of body temperature fluctuations. The calculation formula is as follows: ; in, It's a temperature value. It is the number of pulses; The core component of the acceleration monitoring module is a six-axis inertial sensor, which is used to measure the acceleration and angular velocity of an object, identify the body motion characteristics during driving, execute motion compensation algorithms with a built-in digital motion processor to reduce interference signals caused by vehicle bumps, and finally output a 16-bit digital signal through the SPI interface. The core component of the wireless transmission module is Bluetooth Low Energy, which supports real-time transmission with the terminal output platform and can send raw physiological data to the fatigue monitoring model and the terminal output platform in real time. The core component of the power supply module is a wireless power supply unit, which includes a receiving coil, a rectifier circuit, and a voltage regulator chip. The receiving coil is vacuum-encapsulated with epoxy resin.

[0009] As a preferred embodiment of the above technical solution, the fatigue monitoring model based on multimodal physiological signals includes a feature extraction module, a global cross-modal interaction module, a prediction module, and a loss function. The fatigue monitoring model based on multimodal physiological signals obtains its results by inputting the acquired physiological signal data to be analyzed into the trained fatigue monitoring model based on multimodal physiological signals to obtain the fatigue monitoring results.

[0010] As a preferred embodiment of the above technical solution, the training method for the fatigue monitoring model based on multimodal physiological signals includes the following steps: A. Obtain multimodal physiological signal data from the fatigue monitoring garment and perform data preprocessing to obtain preprocessed heart rate data, body temperature data, and exercise data; B. Input the preprocessed heart rate data, body temperature data, and exercise data into the BiLSTM module respectively. First, execute the forward LSTM unit, then execute the backward LSTM unit to obtain heart rate features. Body temperature characteristics Motion characteristics ; C. Heart rate characteristics Body temperature characteristics Motion characteristics The global cross-modal interaction module is used to achieve comprehensive cross-modal interaction, specifically including the following steps: C1. Execute a linear projection layer to project the features to the same dimensions, resulting in... , , ; C2. Execute the global cross-attention submodule to achieve full-modal interaction, specifically including the following steps: C21. Perform average pooling operation to obtain... , , ; C22. Perform the Cartesian product operation to generate the attention tensor A, which is formulated as follows: ; in, Indicates the outer product; C3. Execute the cross-attention mechanism to calculate the weighted features for each modality, as formulated below: ; ; ; ; ; ; in, , , It is a weighted combination of physiological characteristics. It is a unified regularization coefficient; right , , Perform average pooling separately, and then compare each with the original projected features. , , Perform residual connection to obtain , , ; C4. Execute the parallel feedforward layer, , , The inputs are fed into two multilayer perceptrons, and the results are obtained respectively. , , Then, respectively with , , Perform residual joins to obtain... , , ; D. Execute the prediction module, , , Perform average pooling separately to obtain , , ,Will , , Perform the splicing operation to obtain ,Will The input is placed into a fully connected layer, and then the sigmoid function is used to predict the fatigue probability. E. The mean absolute error is used as the loss function to supervise the network's learning, as shown in the following formula: ; in, For the number of samples, It refers to the first The true label of each sample It refers to the first Predicted values ​​for each sample; During the training process, according to the formula Calculate the loss value, calculate the gradient of the loss function, backpropagate the error from the output layer back to each layer of the network, determine the gradient vector through backpropagation of the loss value, and update the model parameters of the fatigue monitoring model based on multimodal physiological signals. Determine if 100 epochs have been reached. If so, complete the training of the fatigue monitoring model based on multimodal physiological signals; otherwise, continue training.

[0011] As a preferred embodiment of the above technical solution, the terminal output platform includes a user information management module, a multimodal physiological signal fusion analysis module, a dynamic fatigue assessment module, a multi-level early warning push module, and a driving report generation module; The user information management module is used for driver identity registration, personal health data entry, and historical record storage; The multimodal physiological signal fusion analysis module receives and visualizes heart rate, body temperature, and body movement data transmitted by the data acquisition box in real time, and performs signal quality assessment and anomaly detection. The dynamic fatigue assessment module calculates the probability of mild fatigue and the probability of severe fatigue in real time based on the trained fatigue monitoring model. The fatigue determination is based on a preset decision rule. If the total probability of mild fatigue plus severe fatigue exceeds a specific threshold of 0.6, it is determined to be a fatigue state. Furthermore, the specific fatigue level is determined based on the maximum probability. If the probability of mild fatigue is greater than the probability of severe fatigue, it is determined to be a mild fatigue state; otherwise, it is determined to be a severe fatigue state. The multi-level early warning push module triggers differentiated early warning strategies based on fatigue levels. When in a state of mild fatigue, the terminal outputs a pop-up reminder on the platform interface and vibrates. When in a state of severe fatigue, a full-screen red warning is issued, a continuous buzzer alarm is sounded, and location information is automatically pushed to preset emergency contacts. The driving report generation module automatically generates an analysis report that includes the time point of fatigue events, duration, and changes in physiological parameters, provides personalized suggestions, and supports historical data review.

[0012] Secondly, this application provides a monitoring method for a driver fatigue monitoring system based on flexible fabric electrodes, comprising the following steps: S1. Data Acquisition: The fatigue monitoring garment collects the user's physiological signal data in real time while driving through the data acquisition box, including heart rate signal, heart rate variability signal, body temperature signal, angular acceleration and linear acceleration data, and then transmits them to the trained fatigue monitoring model based on multimodal physiological signals and the terminal output platform. S2. Fatigue Status Monitoring: Using a trained fatigue monitoring model based on multimodal physiological signals, the system calculates the probability of mild fatigue and the probability of severe fatigue in real time, performs an assessment of the user's fatigue status, and generates fatigue prediction results. S3. Real-time alerts and data visualization: Users can use the terminal output platform to dynamically display heart rate curves and fatigue indexes. Differentiated alert strategies are triggered based on fatigue levels. When in a state of mild fatigue, a pop-up reminder and vibration will appear on the terminal output platform interface. When in a state of severe fatigue, a full-screen red alert will be displayed, a continuous buzzer alarm will sound, and location information will be automatically pushed to preset emergency contacts. S4. Fatigue Report Generation and Personalized Suggestions: Based on physiological signal data and fatigue state monitoring models, an analysis report is automatically generated, including fatigue event time points, duration, and changes in physiological parameters. Personalized suggestions are provided, and historical data backtracking is supported. This application provides a method for predicting and optimizing motor parameters based on finite element simulation and neural networks.

[0013] Advantages of this invention: 1. The driving fatigue monitoring system based on flexible fabric electrodes in this technical solution has the advantage of being wearable without being obtrusive. It adopts a seamless integration technology of flexible fabric electrodes, embedding biocompatible electrodes into the inner layer of clothing through a sewing process. This design solves the problem of mechanical pressure on the skin caused by traditional rigid electrodes, significantly reducing the risk of skin allergies. The magnetic conductive rubber on the back of the data acquisition box attracts the flexible magnetic strip on the back of the flexible fabric electrode, realizing the automatic adsorption connection between the flexible fabric electrode and the data acquisition box, avoiding the problem of manual plugging and unplugging of cables interfering with the driver's attention in traditional solutions.

[0014] 2. The driver fatigue monitoring system based on flexible fabric electrodes in this technical solution improves the accuracy of fatigue prediction by utilizing multimodal fusion. Its core algorithm uses a bidirectional long short-term memory network to extract spatiotemporal correlation features, which can capture the coupling relationship between heart rate variability and steering wheel operation during sharp turns. The improved Transformer architecture quantifies the multimodal interaction weights through a global cross-attention mechanism, accurately analyzing the impact of body temperature fluctuations on heart rate variability.

[0015] 3. The driver fatigue monitoring system based on flexible fabric electrodes in this technical solution has strong environmental adaptability and robustness. The adaptive bonding characteristics of the flexible electrodes greatly improve signal stability. Compared with solutions that rely on optical sensing, this system is not affected by lighting conditions and successfully solves the signal failure problem in tunnels and at night. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of a driver fatigue monitoring system based on flexible fabric electrodes in one embodiment. Figure 2 This is a schematic diagram of the data acquisition box structure in the embodiment; Figure 3 This is a flowchart illustrating a monitoring method for a driver fatigue monitoring system based on flexible fabric electrodes, as described in this embodiment. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below in conjunction with the embodiments. Specific Implementation like Figure 1 , Figure 2 As shown, a driver fatigue monitoring system based on flexible fabric electrodes includes a fatigue monitoring garment, a data acquisition box, a trained fatigue monitoring model based on multimodal physiological signals, and a terminal output platform. The fatigue monitoring garment includes a garment, a flexible fabric electrode, a flexible magnetic strip, and a magnetically conductive rubber. The garment has a pocket at the chest area. The flexible fabric electrode is in close contact with the skin to collect electrocardiogram signals and is seamlessly integrated into the garment by sewing, eliminating the stiff feel of traditional electrodes and ensuring a comfortable and burden-free wearing experience during driving. The flexible magnetic strip is installed on the back of the flexible fabric electrode. In this embodiment, the back of the flexible fabric electrode is the side away from the human body. The magnetically conductive rubber is installed on the back of the data acquisition box. The data acquisition box is placed in a pocket. The magnetic conductive rubber on the back of the data acquisition box attracts the flexible magnetic strip on the back of the flexible fabric electrode. The flexible fabric electrode is connected to the data acquisition box. The data acquisition box realizes non-invasive acquisition of heart rate, heart rate variability, body temperature, and body movement physiological parameters, and sends them to the fatigue monitoring method and terminal output platform based on multimodal physiological signals. The data acquisition box is small in size and waterproof and sweatproof. More specifically, the data acquisition box includes a heart rate monitoring module, a temperature monitoring module, an acceleration monitoring module, a wireless transmission module, and a power supply module; The core component of the heart rate monitoring module is the BMD101 heart rate sensor, which has a built-in high-precision bioelectrical signal amplifier. It can capture weak electrocardiogram signals and execute the R-wave detection algorithm through the built-in hard-wired digital signal processor to output heart rate and heart rate variability parameters in real time. The core component of the temperature monitoring module is the NST1001 temperature sensor. The probe of the NST1001 temperature sensor is attached to the inside of the shell via thermally conductive silicone. The temperature value is calculated based on the number of pulses, which can indirectly sense the human body surface temperature and ensure real-time capture of body temperature fluctuations. The calculation formula is as follows: ; in, It's a temperature value. It is the number of pulses; The core component of the acceleration monitoring module is a six-axis inertial sensor, which is used to measure the acceleration and angular velocity of an object, identify the body motion characteristics during driving, execute motion compensation algorithms with a built-in digital motion processor to reduce interference signals caused by vehicle bumps, and finally output a 16-bit digital signal through the SPI interface. The core component of the wireless transmission module is Bluetooth Low Energy, which supports real-time transmission with the terminal output platform and can send raw physiological data to the fatigue monitoring model and the terminal output platform in real time. The core component of the power supply module is a wireless power supply unit, including a receiving coil, a rectifier circuit and a voltage regulator chip. The receiving coil is vacuum-encapsulated with epoxy resin, which can isolate the coil and its interface from intrusion, and is waterproof and sweatproof. The trained fatigue monitoring model based on multimodal physiological signals is used to acquire multimodal physiological signal data from the fatigue monitoring garment and process the data to obtain fatigue monitoring results. The terminal output platform has functions for user information input, multimodal physiological signal fusion analysis, dynamic fatigue assessment, and early warning push, which users can view and analyze in real time. More specifically, the terminal output platform includes a user information management module, a multimodal physiological signal fusion analysis module, a dynamic fatigue assessment module, a multi-level early warning push module, and a driving report generation module; The user information management module is used for driver identity registration, personal health data entry, and historical record storage; The multimodal physiological signal fusion analysis module receives and visualizes heart rate, body temperature, and body movement data transmitted by the data acquisition box in real time, and performs signal quality assessment and anomaly detection. The dynamic fatigue assessment module calculates the probability of mild fatigue and the probability of severe fatigue in real time based on the trained fatigue monitoring model. The fatigue determination is based on a preset decision rule. If the total probability of mild fatigue plus severe fatigue exceeds a specific threshold of 0.6, it is determined to be a fatigue state. Furthermore, the specific fatigue level is determined based on the maximum probability. If the probability of mild fatigue is greater than the probability of severe fatigue, it is determined to be a mild fatigue state; otherwise, it is determined to be a severe fatigue state. The multi-level early warning push module triggers differentiated early warning strategies based on fatigue levels. When in a state of mild fatigue, the terminal outputs a pop-up reminder on the platform interface and vibrates. When in a state of severe fatigue, a full-screen red warning is issued, a continuous buzzer alarm is sounded, and location information is automatically pushed to preset emergency contacts. The driving report generation module automatically generates an analysis report that includes the time point of fatigue events, duration, and changes in physiological parameters, provides personalized suggestions, and supports historical data review.

[0019] like Figure 3 As shown, a monitoring method for a driver fatigue monitoring system based on flexible fabric electrodes includes the following steps: S1. Data Acquisition: The fatigue monitoring garment collects the user's physiological signal data in real time while driving through the data acquisition box, including heart rate signal, heart rate variability signal, body temperature signal, angular acceleration and linear acceleration data, and then transmits them to the trained fatigue monitoring model based on multimodal physiological signals and the terminal output platform. S2. Fatigue Status Monitoring: Using a trained fatigue monitoring model based on multimodal physiological signals, the system calculates the probability of mild fatigue and the probability of severe fatigue in real time, performs an assessment of the user's fatigue status, and generates fatigue prediction results. S3. Real-time alerts and data visualization: Users can use the terminal output platform to dynamically display heart rate curves and fatigue indexes. Differentiated alert strategies are triggered based on fatigue levels. When in a state of mild fatigue, a pop-up reminder and vibration will appear on the terminal output platform interface. When in a state of severe fatigue, a full-screen red alert will be displayed, a continuous buzzer alarm will sound, and location information will be automatically pushed to preset emergency contacts. S4. Fatigue Report Generation and Personalized Suggestions: Based on physiological signal data and fatigue state monitoring models, automatically generate analysis reports that include fatigue event time points, duration, and changes in physiological parameters, provide personalized suggestions, and support historical data review.

[0020] This technical solution presents a driver fatigue monitoring system based on flexible fabric electrodes. The fatigue monitoring garment collects physiological signals via flexible fabric electrodes; a data acquisition box enables signal acquisition and transmission; a trained fatigue monitoring model based on multimodal physiological signals is used to acquire and process multimodal physiological signal data from the fatigue monitoring garment to obtain fatigue monitoring results; and a terminal output platform provides user interaction, real-time analysis, dynamic assessment, and multi-level early warning functions. This invention, through the combination of flexible wearable technology and an artificial intelligence model, eliminates the rigid feel of traditional electrodes, ensuring a comfortable and burden-free wearing experience during driving, and achieving non-invasive physiological signal acquisition, real-time monitoring of driver fatigue, accurate assessment, and proactive early warning.

[0021] This technical solution presents a driver fatigue monitoring system based on flexible fabric electrodes, which has the advantage of being wearable without being obtrusive. It adopts a seamless integration technology for flexible fabric electrodes, embedding biocompatible electrodes into the inner layer of clothing through a sewing process. This design solves the problem of mechanical pressure on the skin caused by traditional rigid electrodes, significantly reducing the risk of skin allergies. The magnetic conductive rubber on the back of the data acquisition box attracts the flexible magnetic strip on the back of the flexible fabric electrode, realizing the automatic adsorption connection between the flexible fabric electrode and the data acquisition box, avoiding the problem of manual plugging and unplugging of cables in traditional solutions that interfere with the driver's attention.

[0022] To achieve seamless integration of the flexible fabric electrode into clothing, eliminate the stiff feel of traditional electrodes, and ensure a comfortable and burden-free wearing experience during driving, the flexible fabric electrode has been further optimized. The flexible fabric electrode is made of a conductive layer woven from highly conductive silver-plated nylon fibers and a comfortable and breathable flexible polymer insulating base, which is then composited through a hot-pressing process. A high-precision overlock sewing process is used to seamlessly embed the flexible fabric electrode into the inner layer of the clothing.

[0023] Meanwhile, to ensure stable acquisition of human physiological signal data, the flexible fabric electrode is provided in three sets, two of which are fixed to the shoulders of the clothing, and one set is fixed to the connection between the flexible fabric electrode on the shoulders and the pocket where the data acquisition box is located.

[0024] Existing systems also face three major challenges at the algorithm level: single-segmentation multimodal features neglect the spatiotemporal correlation of physiological parameters; weak model generalization ability considering individual differences in heart rate baselines; and cloud transmission latency exceeding 500ms, which cannot meet the real-time warning requirements of high-speed scenarios. Therefore, this technical solution further optimizes a driver fatigue monitoring system based on flexible fabric electrodes, as follows: The fatigue monitoring model based on multimodal physiological signals includes a feature extraction module, a global cross-modal interaction module, a prediction module, and a loss function. It uses bidirectional LSTM to extract features of each modality, then uses Transformer to perform cross-modal interaction, and then uses the loss function to supervise training to analyze whether the driver is fatigued. The fatigue monitoring model based on multimodal physiological signals obtains its results by inputting the acquired physiological signal data to be analyzed into the trained fatigue monitoring model based on multimodal physiological signals to obtain the fatigue monitoring results.

[0025] The training method for the fatigue monitoring model based on multimodal physiological signals includes the following steps: A. Obtain multimodal physiological signal data from the fatigue monitoring garment and perform data preprocessing to obtain preprocessed heart rate data, body temperature data, and exercise data; B. Input the preprocessed heart rate data, body temperature data, and exercise data into the BiLSTM module respectively. First, execute the forward LSTM unit, then execute the backward LSTM unit to obtain heart rate features. Body temperature characteristics Motion characteristics ; C. Heart rate characteristics Body temperature characteristics Motion characteristics The global cross-modal interaction module is used to achieve comprehensive cross-modal interaction, specifically including the following steps: C1. Execute a linear projection layer to project the features to the same dimensions, resulting in... , , ; C2. Execute the global cross-attention submodule to achieve full-modal interaction, specifically including the following steps: C21. Perform average pooling operation to obtain... , , ; C22. Perform the Cartesian product operation to generate the attention tensor A, which is formulated as follows: ; in, Indicates the outer product; C3. Execute the cross-attention mechanism to calculate the weighted features for each modality, as formulated below: ; ; ; ; ; ; in, , , It is a weighted combination of physiological characteristics. It is a unified regularization coefficient; right , , Perform average pooling separately, and then compare each with the original projected features. , , Perform residual connection to obtain , , ; C4. Execute the parallel feedforward layer, , , The inputs are fed into two multilayer perceptrons, and the results are obtained respectively. , , Then, respectively with , , Perform residual joins to obtain... , , ; D. Execute the prediction module, , , Perform average pooling separately to obtain , , ,Will , , Perform the splicing operation to obtain ,Will The input is placed into a fully connected layer, and then the sigmoid function is used to predict the fatigue probability. E. The mean absolute error is used as the loss function to supervise the network's learning, as shown in the following formula: ; in, For the number of samples, It refers to the first The true label of each sample It refers to the first Predicted values ​​for each sample; During the training process, according to the formula Calculate the loss value, calculate the gradient of the loss function, backpropagate the error from the output layer back to each layer of the network, determine the gradient vector through backpropagation of the loss value, and update the model parameters of the fatigue monitoring model based on multimodal physiological signals. Determine if 100 epochs have been reached. If so, complete the training of the fatigue monitoring model based on multimodal physiological signals; otherwise, continue training.

[0026] The driving fatigue monitoring system based on flexible fabric electrodes in this technical solution improves the accuracy of fatigue prediction by utilizing multimodal fusion. Its core algorithm uses a bidirectional long short-term memory network to extract spatiotemporal correlation features, which can capture the coupling relationship between heart rate variability and steering wheel operation during sharp turns. The improved Transformer architecture quantifies the multimodal interaction weights through a global cross-attention mechanism, and accurately analyzes the impact of body temperature fluctuations on heart rate variability.

[0027] The fatigue monitoring model based on multimodal physiological signals in this technical solution, combined with the bidirectional LSTM-Transformer cross-modal interactive architecture, breaks through the accuracy bottleneck of traditional algorithms and provides a brand-new solution for driving safety.

[0028] This invention collects multimodal physiological signals from drivers using flexible fabric electrodes integrated into fatigue monitoring clothing. These signals encompass heart rate variability, body temperature, and body movement data, comprehensively capturing multidimensional fatigue indicators related to cardiovascular status, metabolic changes, and movement characteristics. After signal preprocessing, the data is transmitted to the fatigue monitoring model via a data acquisition box, and dynamic fatigue assessment is performed through a terminal output platform. This process ensures both real-time data collection and accurate assessment. Based on the physiological characteristics of fatigue development, the system quantifies and classifies fatigue states by probability, and provides intuitive risk trend warnings through a multi-level early warning mechanism. Furthermore, the fatigue monitoring model, trained on a large amount of driving data, can adaptively learn individual physiological characteristics and periodically optimize model weight parameters to ensure the accuracy and generalization ability of the assessment. Combined with real-time fatigue assessment results, the terminal output platform can generate personalized driving reports and rest suggestions for drivers, helping them to rationally plan their trips and recovery measures, thereby effectively improving driving safety.

[0029] The above are all preferred embodiments of this application, and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.

Claims

1. A flexible fabric electrode based driver fatigue monitoring system, characterized in that, The fatigue monitoring clothes, the data acquisition box, the fatigue monitoring model based on multi-modal physiological signals and the terminal output platform are included. The fatigue monitoring clothes include clothes, flexible fabric electrodes, flexible magnetic strips and magnetically attracted conductive rubber. The flexible fabric electrodes are tightly attached to the skin to collect electrocardio signals, and are seamlessly integrated into the clothes by sewing. The flexible magnetic strips are installed on the back of the flexible fabric electrodes. The magnetically attracted conductive rubber is installed on the back of the data acquisition box.

2. The flexible fabric electrode based driver fatigue monitoring system as claimed in claim 1, wherein, The data acquisition box is placed in the pocket, and the magnetically attracted conductive rubber on the back of the data acquisition box and the flexible magnetic strips on the back of the flexible fabric electrodes are attracted to each other.

3. A flexible fabric electrode based driver fatigue monitoring system as claimed in claim 2, wherein, The flexible fabric electrodes are connected to the data acquisition box.

4. The flexible fabric electrode based driver fatigue monitoring system as claimed in claim 1, wherein, The data acquisition box realizes non-invasive collection of heart rate, heart rate variability, body temperature and body movement physiological parameters, and sends them to the fatigue monitoring model based on multi-modal physiological signals and the terminal output platform. The fatigue monitoring model based on multi-modal physiological signals is used to obtain multi-modal physiological signal data from the fatigue monitoring clothes and process the data to obtain fatigue monitoring results. The terminal output platform has user information input, multi-modal physiological signal fusion analysis, dynamic fatigue evaluation and early warning push functions, and allows users to view and analyze in real time. ； wherein is a temperature value, is the number of pulses; The flexible fabric electrodes are composed of a conductive layer of silver-plated nylon fiber with high conductivity and a comfortable and breathable flexible polymer insulating base through a hot pressing process. The flexible fabric electrodes are seamlessly embedded in the inner layer of the clothes by high-precision lockstitch sewing process. The flexible fabric electrodes are provided with three groups, two of which are fixed on the shoulders of the clothes, and one is fixed on the connection between the flexible fabric electrodes on the shoulders and the pocket of the data acquisition box. The data acquisition box includes a heart rate monitoring module, a temperature monitoring module, an acceleration monitoring module, a wireless transmission module and a power supply module. The core element of the heart rate monitoring module is a BMD101 heart rate sensor with a built-in high-precision bioelectric signal amplifier that can capture weak electrocardio signals. The core element of the temperature monitoring module is an NST1001 temperature sensor. The core element of the acceleration monitoring module is a six-axis inertial sensor that measures the acceleration and angular velocity of an object, identifies the body movement characteristics during driving, and outputs a 16-bit digital signal through an SPI interface after executing a motion compensation algorithm to reduce interference signals caused by vehicle bumps. The core element of the wireless transmission module is a low-power Bluetooth that supports real-time transmission with the terminal output platform and can send raw physiological data to the fatigue monitoring model and the terminal output platform in real time. The core element of the power supply module is a wireless power supply unit that includes a receiving coil, a rectifier circuit and a voltage stabilizing chip, and the receiving coil is vacuum filled with epoxy resin.

5. The flexible fabric electrode based driver fatigue monitoring system as claimed in claim 1, wherein, The fatigue monitoring model based on the multi-modal physiological signal comprises a feature extraction module, a global cross-modal interaction module, a prediction module and a loss function; The fatigue monitoring model based on the multi-modal physiological signal is trained by the following steps:

6. A flexible fabric electrode based driver fatigue monitoring system as claimed in claim 5, wherein, A. Obtain multi-modal physiological signal data from the fatigue monitoring clothes and perform data preprocessing to obtain preprocessed heart rate data, body temperature data and motion data; C2. Perform a global cross-attention submodule to realize interaction among all modalities, specifically comprising the following steps: B, input the preprocessed heart rate data, body temperature data, and motion data into the BiLSTM module, first execute the forward LSTM unit, and then execute the backward LSTM unit to obtain heart rate features , body temperature features , and motion features ; C. heart rate features D. body temperature features E. motion features The global cross-modal interaction module is input, which is used to realize comprehensive cross-modal interaction, and specifically includes the following steps: C1, performing a linear projection layer, projecting the features into the same dimension, obtaining , , ; C22. Perform a Cartesian product operation to generate an attention tensor A, which is formulated as follows: C21, performing an average pooling operation, to obtain , , ; C3. Perform a cross-attention mechanism to calculate the weighted features of each modality, which is formulated as follows: ； wherein denotes the outer product; E. Use the mean absolute error as a loss function to supervise the learning of the network, which is formulated as follows: ； ； ； ； ； ； wherein, , , is a weighted combination of physiological features, is a uniform regularization coefficient; , , are respectively averaged-pooled, and then respectively connected with the original projection features , , to obtain , , ;​ C4, performing a parallel feedforward layer, to obtain , , input two multi-layer perceptrons, respectively, to obtain , , , respectively, and then respectively perform residual connection with , , to obtain , , ; D, performing a prediction module, will , , respectively average pooling, respectively, get , , , will , , perform splicing operation, get , will input into the full connection layer, and then use the sigmoid function to predict the fatigue probability; Determine whether 100 epochs are reached, if yes, the training of the fatigue monitoring model based on the multi-modal physiological signal is completed, otherwise, continue training. ； wherein, is the number of samples, refers to the true label of the th sample, refers to the predicted value of the th sample; In the training process, the loss value is calculated according to the formula The gradient of the loss function is calculated, the error is back-propagated from the output layer to each layer of the network, the gradient vector is determined through back-propagation of the loss value, and the model parameters of the fatigue monitoring model based on the multi-modal physiological signals are updated; The terminal output platform comprises a user information management module, a multi-modal physiological signal fusion analysis module, a dynamic fatigue evaluation module, a multi-level early warning pushing module and a driving report generation module; 7. The flexible fabric electrode based driver fatigue monitoring system as claimed in claim 1, wherein, The user information management module is used for driver identity registration, personal health data entry and historical record storage; The multi-modal physiological signal fusion analysis module receives and visualizes heart rate, body temperature and body motion data transmitted by the data acquisition box in real time, and performs signal quality evaluation and abnormality detection; The dynamic fatigue evaluation module calculates the probability of mild fatigue and the probability of severe fatigue in real time based on the trained fatigue monitoring model, and the fatigue determination is based on a preset decision rule. If the total probability of mild fatigue and severe fatigue exceeds a certain threshold: 0.6, it is determined to be in a fatigue state; and further according to the maximum probability, the specific fatigue level is determined. If the probability of mild fatigue is greater than the probability of severe fatigue, it is determined to be in a mild fatigue state; otherwise, it is determined to be in a severe fatigue state; The multi-level early warning pushing module triggers differentiated early warning strategies according to the fatigue level. When in a mild fatigue state, the terminal output platform interface pop-up window reminds and vibrates. When in a severe fatigue state, a full-screen red warning and continuous buzzing alarm are triggered, and positioning information is automatically pushed to the preset emergency contact person; The driving report generation module automatically generates an analysis report containing the fatigue event time point, duration and physiological parameter changes, provides personalized suggestions, and supports historical data backtracking. The method comprises the following steps:

8. A method of monitoring a driver fatigue monitoring system based on a flexible fabric electrode according to any one of claims 1-7, characterized in that, S1. Data acquisition: the fatigue monitoring clothes acquire physiological signal data of a user during driving in real time through a data acquisition box, including heart rate signal, heart rate variability signal, body temperature signal, angular acceleration and linear acceleration data, and then transmit the data to the fatigue monitoring model based on the multi-modal physiological signal and the terminal output platform; ​ S2, fatigue state monitoring: using the trained fatigue monitoring model based on multi-modal physiological signals, real-time calculation of mild fatigue probability and severe fatigue probability, evaluation of user fatigue state, generation of fatigue prediction results; S3, real-time warning and data visualization: the user uses the terminal output platform to dynamically display the heart rate curve and fatigue index, and triggers differentiated warning strategies according to the fatigue level. When in a mild fatigue state, the terminal output platform interface pop-up window reminds and vibrates. When in a severe fatigue state, a full-screen red warning, continuous bee buzzing alarm, and automatic push of positioning information to the preset emergency contact person are triggered; S4, fatigue report generation and personalized suggestions: based on physiological signal data and fatigue state monitoring model, automatically generate analysis report containing fatigue event time point, duration and physiological parameter change, provide personalized suggestions, and support historical data backtracking.

Images