Fatigue early warning method and device, equipment and storage medium

Abstract

The invention relates to a fatigue early warning method, device and equipment and a storage medium, and the method comprises the steps: obtaining initial physiological data of a driver and target physiological data collected at the current moment, the physiological data being composed of data collected by a plurality of physiological sensors; calling a pre-trained neural network model to enable the pre-trained neural network model to obtain a probability corresponding to each fatigue level by using the initial physiological data and the target physiological data; determining a target fatigue level of the driver according to the probability corresponding to each fatigue level; and according to the target fatigue level, determining a to-be-used fatigue warning mode so as to perform warning by adopting the fatigue warning mode. The method can improve the accuracy of the fatigue detection result.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a fatigue warning method, device, equipment and storage medium. Background Art

[0002] The driver is required to concentrate while driving. If the driver is under high mental stress for a long time, the brain will easily become fatigued, which can easily lead to traffic accidents. Therefore, it is crucial to ensure the driver's safe driving.

[0003] In order to ensure the driver's safe driving, the existing technology often uses heart rate data to determine whether the driver is driving fatigued, that is, by collecting current heart rate data and comparing it with preset normal heart rate data, the driver is judged whether he is awake or fatigued based on the comparison result.

[0004] However, when the heart rate sensor is interfered by external factors, the data collected by the heart rate sensor will be inaccurate, which will lead to inaccurate fatigue detection results. Summary of the invention

[0005] The present application provides a fatigue warning method, device, equipment and storage medium, which can improve the accuracy of fatigue detection results.

[0006] In a first aspect, the present application provides a fatigue warning method, the method comprising:

[0007] Acquire the driver's initial physiological data and target physiological data collected at the current moment, where the physiological data is composed of data collected by multiple physiological sensors;

[0008] Calling a pre-trained neural network model so that the pre-trained neural network model uses the initial physiological data and the target physiological data to obtain a probability corresponding to each fatigue level;

[0009] Determining a target fatigue level of the driver according to the probability corresponding to each fatigue level;

[0010] A fatigue warning method to be used is determined according to the target fatigue level, and the fatigue warning method is used for warning.

[0011] Optionally, the pre-trained neural network model uses the initial physiological data and the target physiological data to obtain the probability corresponding to each fatigue level, including:

[0012] Determining difference data according to the target physiological data and the initial physiological data;

[0013] Normalizing the difference data to obtain normalized data;

[0014] The pre-trained neural network model uses the normalized data to obtain the probability corresponding to each fatigue level.

[0015] Optionally, the method further includes:

[0016] acquiring a plurality of reference samples, each reference sample comprising reference physiological data;

[0017] For each reference physiological data in the reference sample, detecting whether the reference physiological data meets a preset condition;

[0018] determining the reference physiological data meeting the preset conditions as the target reference physiological data;

[0019] Determine the reference sample where the target reference physiological data is located as the target sample;

[0020] Combining the target samples to generate a sample set to be used;

[0021] Based on the sample set, the neural network model is trained to obtain a pre-trained neural network model.

[0022] Optionally, the detecting whether the reference physiological data meets a preset condition includes:

[0023] For each type of data to be tested in the reference physiological data, determine a filter for detecting the data to be tested according to the type of the data to be tested, and when the data to be tested is not filtered by the filter, determine that the data to be tested meets the preset condition; when the data to be tested is filtered by the filter, determine that the data to be tested does not meet the preset condition;

[0024] When there is no data to be measured that does not meet the preset condition in the reference physiological data, determining that the reference physiological data meets the preset condition;

[0025] When there is data to be measured that does not meet the preset condition in the reference physiological data, it is determined that the reference physiological data does not meet the preset condition.

[0026] Optionally, the detecting whether the reference physiological data meets a preset condition includes:

[0027] For each type of data to be tested in the reference physiological data, according to the type of the data to be tested and the correspondence between the preset type and the benchmark data, the benchmark data used to detect the data to be tested is determined; when the distance between the data to be tested and the benchmark data is less than or equal to the preset distance, it is determined that the data to be tested meets the preset condition; when the distance between the data to be tested and the benchmark data is greater than the preset distance, it is determined that the data to be tested does not meet the preset condition;

[0028] When there is no data to be measured that does not meet the preset condition in the reference physiological data, determining that the reference physiological data meets the preset condition;

[0029] When there is data to be measured that does not meet the preset condition in the reference physiological data, it is determined that the reference physiological data does not meet the preset condition.

[0030] Optionally, the neural network model includes three layers, namely an input layer, a hidden layer and an output layer, the input layer includes neurons corresponding to each type of data, the output layer includes neurons corresponding to each fatigue level, and the normalized data is input into a pre-trained neural network model to obtain the probability corresponding to each fatigue level, including:

[0031] For each type of data in the normalized data, input the data into a corresponding neuron in the input layer according to the type of the data, so as to input the normalized data into a pre-trained neural network model;

[0032] Using the hidden layer, performing hidden linear transformation and hidden nonlinear transformation on the input data to obtain first linear data;

[0033] Using each neuron in the output layer, output linear transformation and output nonlinear transformation are performed on the linearly processed data to obtain the probability corresponding to each fatigue level.

[0034] Optionally, the hidden layer includes a first hidden sublayer and a second hidden sublayer, and using the hidden layer to perform a hidden linear transformation and a hidden nonlinear transformation on the input data to obtain the first linear data includes:

[0035] Using the first transformation formula in the first hidden sublayer, performing hidden linear mapping on the input data, mapping the input data to a first feature space, and using the first activation function, performing hidden nonlinear transformation on the data mapped to the first feature space to obtain second linear data output by the first hidden sublayer;

[0036] Using the second transformation formula in the second hidden sublayer, the second linear data is hidden linearly mapped, the second linear data is mapped to a second feature space, and using a second activation function, the data mapped to the second feature space is hidden nonlinearly transformed to obtain the first linear data output by the second hidden sublayer.

[0037] In a second aspect, the present application provides a fatigue warning device, the device comprising:

[0038] An acquisition unit, used to acquire the driver's initial physiological data and target physiological data collected at the current moment, wherein the physiological data is composed of data collected by a variety of physiological sensors;

[0039] A calling unit, configured to call a pre-trained neural network model, so that the pre-trained neural network model uses the initial physiological data and the target physiological data to obtain a probability corresponding to each fatigue level;

[0040] A first determination unit, configured to determine a target fatigue level of the driver according to a probability corresponding to each fatigue level;

[0041] The second determining unit is used to determine a fatigue warning method to be used according to the target fatigue level, so as to use the fatigue warning method for warning.

[0042] Optionally, the calling unit is used to:

[0043] Determining difference data according to the target physiological data and the initial physiological data;

[0044] Normalizing the difference data to obtain normalized data;

[0045] The pre-trained neural network model uses the normalized data to obtain the probability corresponding to each fatigue level.

[0046] Optionally, the device further comprises a training unit, wherein the training unit is used to:

[0047] acquiring a plurality of reference samples, each reference sample comprising reference physiological data;

[0048] For each reference physiological data in the reference sample, detecting whether the reference physiological data meets a preset condition;

[0049] determining the reference physiological data meeting the preset conditions as the target reference physiological data;

[0050] Determine the reference sample where the target reference physiological data is located as the target sample;

[0051] Combining the target samples to generate a sample set to be used;

[0052] Based on the sample set, the neural network model is trained to obtain a pre-trained neural network model.

[0053] Optionally, the training unit is used to:

[0054] For each type of data to be tested in the reference physiological data, determine a filter for detecting the data to be tested according to the type of the data to be tested, and when the data to be tested is not filtered by the filter, determine that the data to be tested meets the preset condition; when the data to be tested is filtered by the filter, determine that the data to be tested does not meet the preset condition;

[0055] When there is no data to be measured that does not meet the preset condition in the reference physiological data, determining that the reference physiological data meets the preset condition;

[0056] When there is data to be measured that does not meet the preset condition in the reference physiological data, it is determined that the reference physiological data does not meet the preset condition.

[0057] Optionally, the training unit is used to:

[0058] For each type of data to be tested in the reference physiological data, according to the type of the data to be tested and the correspondence between the preset type and the benchmark data, the benchmark data used to detect the data to be tested is determined; when the distance between the data to be tested and the benchmark data is less than or equal to the preset distance, it is determined that the data to be tested meets the preset condition; when the distance between the data to be tested and the benchmark data is greater than the preset distance, it is determined that the data to be tested does not meet the preset condition;

[0059] When there is no data to be measured that does not meet the preset condition in the reference physiological data, determining that the reference physiological data meets the preset condition;

[0060] When there is data to be measured that does not meet the preset condition in the reference physiological data, it is determined that the reference physiological data does not meet the preset condition.

[0061] Optionally, the neural network model includes three layers, namely an input layer, a hidden layer and an output layer, the input layer includes neurons corresponding to each type of data, the output layer includes neurons corresponding to each fatigue level, and the calling unit is used to:

[0062] For each type of data in the normalized data, input the data into a corresponding neuron in the input layer according to the type of the data, so as to input the normalized data into a pre-trained neural network model;

[0063] Using the hidden layer, performing hidden linear transformation and hidden nonlinear transformation on the input data to obtain first linear data;

[0064] Using each neuron in the output layer, output linear transformation and output nonlinear transformation are performed on the linearly processed data to obtain the probability corresponding to each fatigue level.

[0065] Optionally, the hidden layer includes a first hidden sublayer and a second hidden sublayer, and the calling unit is used to:

[0066] Using the first transformation formula in the first hidden sublayer, performing hidden linear mapping on the input data, mapping the input data to a first feature space, and using the first activation function, performing hidden nonlinear transformation on the data mapped to the first feature space to obtain second linear data output by the first hidden sublayer;

[0067] Using the second transformation formula in the second hidden sublayer, the second linear data is hidden linearly mapped, the second linear data is mapped to a second feature space, and using a second activation function, the data mapped to the second feature space is hidden nonlinearly transformed to obtain the first linear data output by the second hidden sublayer.

[0068] In a third aspect, the present application provides a fatigue warning device, comprising: at least one communication interface; at least one bus connected to the at least one communication interface; at least one processor connected to the at least one bus; at least one memory connected to the at least one bus, wherein the processor is configured to:

[0069] Acquire the driver's initial physiological data and target physiological data collected at the current moment, where the physiological data is composed of data collected by multiple physiological sensors;

[0070] Calling a pre-trained neural network model so that the pre-trained neural network model uses the initial physiological data and the target physiological data to obtain a probability corresponding to each fatigue level;

[0071] Determining a target fatigue level of the driver according to the probability corresponding to each fatigue level;

[0072] A fatigue warning method to be used is determined according to the target fatigue level, and the fatigue warning method is used for warning.

[0073] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, and the computer program implements the above-mentioned fatigue warning method when executed by a processor.

[0074] Compared with the prior art, the above technical solution provided by the embodiment of the present application has the following advantages: In the embodiment of the present application, by obtaining the initial physiological data of the driver and the reference physiological data collected at the current moment. Then, the pre-trained neural network model can be called so that the pre-trained neural network model uses the initial physiological data and the target physiological data to obtain the probability corresponding to each fatigue level. According to the probability corresponding to each fatigue level, the target fatigue level of the driver is determined; according to the target fatigue level, the fatigue warning method to be used is determined, and the fatigue warning method is used for warning. Since the physiological data is composed of data collected by multiple physiological sensors, and the fatigue detection result is determined based on these physiological data, even if a physiological sensor is interfered with, resulting in deviations in the data collected by it, it will not affect the final fatigue detection result much, thereby improving the accuracy of the fatigue detection result. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0076] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0077] One or more embodiments are exemplarily described by pictures in the corresponding drawings, and these exemplified descriptions do not constitute limitations on the embodiments. Elements with the same reference numerals in the drawings represent similar elements, and unless otherwise stated, the figures in the drawings do not constitute proportional limitations.

[0078] Figure 1 A schematic diagram of a fatigue warning method provided in an embodiment of the present application;

[0079] Figure 2 A schematic diagram of a process for determining a fatigue level probability provided in an embodiment of the present application;

[0080] Figure 3 A flowchart of a data processing method provided in an embodiment of the present application;

[0081] Figure 4 A flowchart of another data processing method provided in an embodiment of the present application;

[0082] Figure 5 A flowchart of another training method provided in an embodiment of the present application;

[0083] Figure 6 A flowchart of a data detection method provided in an embodiment of the present application;

[0084] Figure 7 A flowchart of another data detection method provided in an embodiment of the present application;

[0085] Figure 8 A schematic diagram of a process flow of a fatigue warning device provided in an embodiment of the present application;

[0086] Fig. 9 A schematic diagram of a fatigue warning device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0087] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0088] The disclosure below provides many different embodiments or examples to implement different structures of the present invention. In order to simplify the disclosure of the present invention, the parts and settings of specific examples are described below. Of course, they are only examples, and the purpose is not to limit the present invention. In addition, the present invention can repeat reference numbers and / or letters in different examples. This repetition is for the purpose of simplification and clarity, and does not itself indicate the relationship between the various embodiments and / or settings discussed.

[0089] When the driver is in a state of fatigue, physiological data such as heart rate, blood pressure, respiratory rate and body temperature will change to a certain extent. Specifically, when the driver is in a state of fatigue, the heart rate value usually increases, because the body needs more oxygen and energy to maintain basic functions, and the heart needs faster blood to meet the demand. Blood pressure values ​​usually increase as well, which is due to the body's stress response, increased activity of the sympathetic nervous system, leading to vasoconstriction and increased cardiac output. The respiratory rate usually increases as well, because the body needs more oxygen to maintain energy metabolism, and the respiratory system needs to work faster to meet the demand. Body temperature also rises slightly, because the body usually produces more heat. Based on the above, the present application divides the corresponding states into multiple fatigue levels according to the changes in physiological data, and then more accurately determines the driver's fatigue state based on the physiological data.

[0090] In order to collect the above physiological data, the smart seat includes a physiological signal sensor module, a data processing module and a wake-up warning module. The physiological signal sensor module includes a heart rate sensor, a blood pressure sensor, a breathing sensor and a body temperature sensor, wherein the heart rate sensor is used to detect the driver's heart rate value, the blood pressure sensor is used to detect the driver's blood pressure value, the breathing sensor is used to detect the driver's breathing frequency, and the body temperature sensor is used to detect the driver's body temperature value. In order to accurately reflect the driver's physiological signals, the arrangement of the sensors is particularly important. The heart rate sensor and the blood pressure sensor are embedded in the seat back, close to the driver's back position, and the heart rate and blood pressure are measured by the changes in blood flow in the back. The breathing sensor adopts a piezoelectric sensing method, and the breathing sensor is embedded in the seat back, close to the driver's chest position, so as to measure the breathing frequency. The body temperature sensor is embedded in the seat headrest, close to the driver's ear position, to monitor the driver's body temperature. Afterwards, the physiological signal sensor module can collect the driver's physiological data and send it to the data processing module deployed in the seat controller, so that the data processing module processes the received data, obtains the processing result, and sends the processing result to the wake-up warning module.

[0091] Based on the above, the present application embodiment provides a fatigue warning method, which can improve the accuracy of fatigue detection results. Figure 1 As shown, the specific steps include:

[0092] Step 101 : acquiring the driver's initial physiological data and target physiological data collected at the current moment.

[0093] Among them, physiological data is composed of data collected by various physiological sensors, such as heart rate values ​​collected by heart rate sensors, blood pressure values ​​collected by blood pressure sensors, respiratory frequency collected by respiratory sensors, and body temperature values ​​collected by body temperature sensors.

[0094] In this step, the initial physiological data is collected when the driver sits in the seat and the vehicle starts to drive. This is because when the vehicle starts to drive, the driver is generally sitting in the seat and is not in a fatigued state. Therefore, when the vehicle starts to drive, the driver's physiological data can be collected to obtain the driver's initial physiological data. In order to make the collected initial physiological data more accurately reflect the driver's physiological state, the physiological data can be collected once every period of time, and collected continuously for multiple times, and then the average value of these physiological data is determined, and the average value is determined as the initial physiological data. For example, the physiological signal sensor module collects the driver's physiological data five times in a row at an interval of one minute, and averages these physiological data to obtain average physiological data, and the average physiological data is determined as the initial physiological data.

[0095] After obtaining the initial physiological data, the physiological signal sensor module starts to periodically collect the physiological data of the driver and sends the target physiological data collected at the current moment to the data processing module. When the data processing module receives the data, it can execute step 102 based on the data.

[0096] Step 102, calling a pre-trained neural network model so that the pre-trained neural network model uses the initial physiological data and the target physiological data to obtain the probability corresponding to each fatigue level.

[0097] In this step, the smart seat also includes a data storage module, which stores a pre-trained neural network model. When the pre-trained neural network model is needed, the data processing module can call the pre-trained neural network model stored in the data storage module, and use the pre-trained neural network model to use the initial physiological data and the target physiological data to obtain the probability corresponding to each fatigue level.

[0098] Step 103: determining the target fatigue level of the driver according to the probability corresponding to each fatigue level.

[0099] In this step, the data processing module may determine the fatigue level with the greatest probability among the probabilities corresponding to each fatigue level, and determine the fatigue level as the target fatigue level of the driver.

[0100] Step 104 , determining a fatigue warning method to be used according to the target fatigue level, and using the fatigue warning method for warning.

[0101] In this step, the storage module also stores the correspondence between the fatigue level and the fatigue warning mode. The data processing module can obtain the correspondence in the storage module and determine the fatigue warning mode to be used according to the target fatigue level and the correspondence.

[0102] For example, when the fatigue level is mild fatigue, the fatigue warning method is to control the seat to vibrate slightly; when the fatigue level is moderate fatigue, the fatigue warning method is to control the seat to vibrate moderately and the car audio to emit a warning sound; when the fatigue level is severe fatigue, the fatigue warning method is to control the seat to vibrate strongly, the car audio to emit a warning sound, and the central control screen to display a warning message.

[0103] In an embodiment of the present application, the initial physiological data of the driver and the reference physiological data collected at the current moment are obtained. Then, the pre-trained neural network model can be called, so that the pre-trained neural network model uses the initial physiological data and the target physiological data to obtain the probability corresponding to each fatigue level. According to the probability corresponding to each fatigue level, the target fatigue level of the driver is determined; according to the target fatigue level, the fatigue warning method to be used is determined, and the fatigue warning method is used for warning. Since the physiological data is composed of data collected by multiple physiological sensors, and the fatigue detection results are determined based on these physiological data, even if a physiological sensor is interfered with, resulting in deviations in the data collected by it, it will not affect the final fatigue detection results much, thereby improving the accuracy of the fatigue detection results.

[0104] In the embodiment of the present application, due to the differences between individual drivers, if the probability corresponding to each fatigue level is determined directly based on the target physiological data, the calculation result may be biased, and the driver may not be accurately warned. Therefore, in order to reduce the differences between individual drivers, the present application can calculate the difference data between the target physiological data and the initial physiological data, and calculate the probability corresponding to each fatigue level based on the difference data. In summary, the embodiment of the present application provides a method for determining the probability of fatigue level, such as Figure 2 As shown, the specific steps include:

[0105] Step 201, determining difference data according to target physiological data and initial physiological data.

[0106] In this step, since the physiological data includes heart rate value, blood pressure value, respiration value and body temperature value, the smart seat can calculate the difference between the target heart rate value in the target physiological data and the initial heart rate value in the initial physiological data, calculate the difference between the target blood pressure value in the target physiological data and the initial blood pressure value in the initial physiological data, calculate the difference between the target respiration value in the target physiological data and the initial respiration value in the initial physiological data, and calculate the difference between the target body temperature value in the target physiological data and the initial body temperature value in the initial physiological data. After obtaining these differences, the difference data between the target physiological data and the initial physiological data is obtained.

[0107] Step 202, normalize the difference data to obtain normalized data.

[0108] In this step, in order to ensure the standardization of the data, the smart seat can use a normalization algorithm to normalize the difference data to obtain normalized data.

[0109] Among them, the normalization algorithm is a data preprocessing technology, which aims to map the characteristic values ​​of the data to a specific interval, usually [0,1], so that different features are comparable, and it also helps to improve the performance and convergence speed of certain algorithms.

[0110] Step 203: input the normalized data into a pre-trained neural network model to obtain the probability corresponding to each fatigue level.

[0111] In this step, the normalized data is input into the pre-trained neural network model, and the probability corresponding to each fatigue level is output.

[0112] In the embodiment of the present application, the neural network model adopts a multi-layer perceptron (MLP) neural network structure. The neural network model of this structure includes three layers, namely an input layer, a hidden layer, and an output layer. The input layer includes neurons corresponding to each type of data, and the output layer includes neurons corresponding to each fatigue level. Based on the neural network model of the above structure, the normalized data is processed to obtain the probability corresponding to each fatigue level. Therefore, the embodiment of the present application provides a data processing method, such as Figure 3 As shown, the specific steps include:

[0113] Step 301, for each type of data in the normalized data, according to the type of data, the data is input into the corresponding neuron in the input layer, so as to input the normalized data into the pre-trained neural network model.

[0114] For example, the neurons corresponding to each type of data include neurons corresponding to heart rate, neurons corresponding to blood pressure, neurons corresponding to respiration, and neurons corresponding to body temperature. The heart rate value is input into the neuron corresponding to the heart rate, the blood pressure value is input into the neuron corresponding to the blood pressure, the respiratory frequency is input into the neuron corresponding to the respiration, and the body temperature value is input into the neuron corresponding to the body temperature.

[0115] Step 302: Use the hidden layer to perform hidden linear transformation and hidden nonlinear transformation on the input data to obtain first linear data.

[0116] In this step, after the normalized data is input into the input layer, the neurons in the input layer receive the data and input it into the hidden layer, so that the hidden layer performs hidden linear transformation and hidden nonlinear transformation on the data to obtain the first linear data.

[0117] Step 303: Use each neuron in the output layer to perform output linear transformation and output nonlinear transformation on the first linear data to obtain the probability corresponding to each fatigue level.

[0118] In this step, the output layer includes neurons corresponding to each fatigue level, and each neuron is provided with a linear transformation function and a nonlinear activation function, wherein the linear transformation function performs a linear transformation on the first linear data to obtain a linear transformation result, and the nonlinear activation function performs a nonlinear transformation on the linear change result to obtain the probability corresponding to each fatigue level. The linear transformation function can be z(j) is the linear transformation result of the jth neuron in the output layer, w(i,j) is the weight between the i-th neuron in the hidden layer and the j-th neuron in the output layer, x(i) is the first linear data corresponding to the i-th neuron in the last layer of the hidden layer, and b(j) is the bias term corresponding to the j-th neuron in the output layer. The minimum value of i is 1, and the maximum value is the number of neurons in the last layer of the hidden layer, m. The minimum value of j is 1, and the maximum value is the number of fatigue levels, k. The nonlinear activation function is the Softmax activation function, specifically z(n) is the linear transformation result of the nth neuron in the output layer, z(k) is the linear transformation result of the kth neuron in the output layer, σ(z) k is the probability of the fatigue level corresponding to the kth neuron.

[0119] In this step, the Softmax activation function can convert the linear combination of the output layer into a probability distribution, making the output of the model more intuitive and easier to interpret. The Softmax activation function can more accurately predict the probability of each category, thereby improving classification accuracy.

[0120] In the embodiment of the present application, the hidden layer includes a first hidden sublayer and a second hidden sublayer. The number of neurons included in the first hidden sublayer is greater than the number of neurons included in the second hidden sublayer. For example, the first hidden sublayer includes 128 neurons, and the second hidden sublayer includes 64 neurons. Therefore, the smart seat uses the hidden layer to perform hidden linear transformation and hidden nonlinear transformation on the input data to obtain first linear data. Therefore, the embodiment of the present application provides a data processing method, which is as follows: Figure 4 Specifically, it includes:

[0121] Step 401, using the first transformation formula in the first hidden sublayer, performing hidden linear mapping on the input data, mapping the input data to the first feature space, using the first activation function, performing hidden nonlinear transformation on the data mapped to the first feature space, and obtaining the second linear data output by the first hidden sublayer.

[0122] In this step, the first hidden sublayer includes 128 neurons, and each neuron is set with a first transformation formula and a first activation function. Using the first transformation formula, the input data is subjected to hidden linear mapping, and the input data is mapped to the first feature space. The first transformation formula is z(j) is the linear transformation result of the jth neuron in the first hidden sublayer, w(i,j) is the weight between the i-th neuron in the input layer and the j-th neuron in the first hidden sublayer, x(i) is the input data corresponding to the i-th neuron in the input layer, and b(j) is the bias item corresponding to the j-th neuron in the first hidden sublayer. i ranges from 1 to n, n is the number of neurons in the input layer, and j ranges from 1 to 128. Then the first activation function is used to set negative values ​​to zero and positive values ​​remain unchanged. The first activation function is the ReLU activation function, specifically a(j) = max(0,z(j)), z(j) is the linear transformation result of the j-th neuron, and a(j) is the second linear data output by the j-th neuron.

[0123] The calculation of the ReLU activation function in the above process is simple and does not involve complex exponential operations. Using the ReLU activation function can significantly increase the training speed. The ReLU activation function introduces nonlinearity, allowing the neural network to learn and represent complex nonlinear relationships, thereby enhancing the expressive power of the model. The ReLU activation function can effectively alleviate the gradient vanishing problem, making the gradient update of the deep neural network more stable, thereby improving the training effect of the model.

[0124] It should be noted that during the training process, the parameters adjusted include weights and bias items, so that after the training is completed, the weights and bias items in the pre-trained neural network model will no longer change and become fixed values. Based on this, the required data processing process can be completed according to the above formula.

[0125] Step 402, using the second transformation formula in the second hidden sublayer, performing hidden linear mapping on the second linear data, mapping the second linear data to the second feature space, using the second activation function, performing hidden nonlinear transformation on the data mapped to the second feature space, and obtaining the first linear data output by the second hidden sublayer.

[0126] The first transformation formula and the second transformation formula may be the same or different. The first activation function and the second activation function may be the same or different. This step is illustrated by taking the first transformation formula and the second transformation formula as being the same, and the first activation function and the second activation function as being the same.

[0127] In this step, the second hidden sublayer includes 64 neurons, and each neuron is provided with a second transformation formula and a second activation function. Using the second transformation formula, the input data is subjected to hidden linear mapping, and the input data is mapped to the second feature space. The second transformation formula is z(j) is the linear transformation result of the jth neuron in the second hidden sublayer, w(i,j) is the weight between the i-th neuron in the first hidden sublayer and the j-th neuron in the second hidden sublayer, x(i) is the second linear data output by the i-th neuron in the first hidden sublayer, b(j) is the bias item corresponding to the j-th neuron in the second hidden sublayer, where the value range of i is 1 to 128 and the value range of j is 1 to 64. Then the first activation function is used to set negative values ​​to zero and positive values ​​remain unchanged. The first activation function is a(j) = max(0, z(j)), z(j) is the result of the linear transformation of the j-th neuron, and a(j) is the first linear data output by the j-th neuron.

[0128] In an embodiment of the present application, since multiple layers of hidden sublayers can capture the complex features and nonlinear relationships of data, the present application sets two layers of hidden sublayers, which can better extract and represent the features of input data. Each hidden sublayer has a different number of neurons. The first hidden sublayer has 128 neurons, which can provide a higher capacity to capture the complex features of input data. The second hidden sublayer has 64 neurons. The layer-by-layer decreasing design helps the model to gradually extract and compress features. The design of hidden sublayers in this neural network can improve the expressiveness and generalization ability of the model while reducing the risk of overfitting. This design improves training efficiency while ensuring model performance.

[0129] In the embodiment of the present application, after using the pre-trained neural network model, it is necessary to first train the created neural network model to obtain the pre-trained neural network model. Therefore, the embodiment of the present application provides a training method, such as Figure 5 As shown, the specific steps include:

[0130] Step 501, obtaining multiple reference samples.

[0131] The reference sample includes not only reference physiological data, but also reference labeling probabilities corresponding to each fatigue level. The reference labeling probabilities are determined by technicians based on the driver's status when the reference physiological data are collected.

[0132] In this step, the physiological data of different drivers at different times are collected, and the driver's state at the time of data collection is recorded to obtain multiple reference physiological data and the state corresponding to each reference physiological data. Afterwards, the technician can mark the corresponding state of each reference physiological data to obtain the reference marking probability corresponding to each reference physiological data. Afterwards, the reference physiological data and its corresponding reference marking probability are combined to obtain its corresponding reference sample.

[0133] Step 502 : for each reference physiological data in the reference sample, detecting whether the reference physiological data meets a preset condition.

[0134] In this step, when the physiological data is abnormal, the physiological data does not meet the preset condition, and when the physiological data is normal, the physiological data meets the preset condition. Therefore, whether the reference physiological data in each reference sample is normal is detected to detect whether the reference physiological data meets the preset condition.

[0135] Step 503: Determine the reference physiological data that meets the preset conditions as the target reference physiological data.

[0136] In this step, when a certain reference physiological data meets a preset condition, the reference physiological data is determined as the target reference physiological data.

[0137] Step 504: determine the reference sample where the target reference physiological data is located as the target sample.

[0138] In this step, when the reference physiological data in a reference sample is the target reference physiological data, it means that the reference sample meets the conditions, and training can be performed based on the reference sample, and the reference sample where the target reference physiological data is located is determined as the target sample.

[0139] Step 505: combine the target samples to generate a sample set to be used.

[0140] In this step, a sample set to be used is generated based on the target sample.

[0141] Step 506: Train the neural network model based on the sample set to obtain a pre-trained neural network model.

[0142] In this step, a sample is randomly selected from the sample set to train the neural network model to obtain a trained neural network model, and it is determined whether the trained neural network model meets the conditions. If it meets the conditions, the trained neural network model is determined as a pre-trained neural network model. If it does not meet the conditions, the parameters in the trained neural network model are adjusted, and the above training process is repeated based on the trained neural network model until the trained neural network model meets the conditions.

[0143] In the above process, the physiological data in the sample is input into the neural network model so that the neural network model outputs the predicted probability corresponding to each fatigue level, and the predicted probability corresponding to each fatigue level is compared with the reference labeled probability to determine whether the trained neural network model meets the conditions. The specific method can be to input the predicted probability corresponding to each fatigue level and the reference labeled probability into the loss function so that the loss function outputs a loss value, and to determine whether the trained neural network model meets the conditions by using the loss value, and if it does not meet the conditions, adjust the parameters in the model according to the loss value and the optimization algorithm.

[0144] In this step, the loss function selects the cross-entropy loss, which guides the model training by measuring the difference between the model's predicted value and the true value. By minimizing the cross-entropy loss, the model can more accurately predict the probability of each category. The optimization algorithm selects the SGD optimization algorithm, which calculates the gradient and updates the parameters on each batch, making it more efficient when processing large-scale data sets, reducing memory requirements and computing time.

[0145] In addition, the learning rate during training can be set to 0.001, which provides an appropriate update pace while ensuring model convergence, improving training efficiency and stability. The batch size is set to 32 to strike a balance between the stability of gradient estimation and the update frequency, ensuring that the model can stably update parameters while maintaining high training efficiency. The number of iterations is set to 100 to ensure that the model has enough time to learn data features and avoid underfitting, while preventing overfitting by monitoring verification performance.

[0146] Before starting training, you can also divide the sample set into training data set, validation data set, and test data set. Use the training data set to train the model, and after each training, use the validation data set to evaluate the performance of the model. Load the best model saved during the training process and use the test data set to predict the model.

[0147] In the embodiment of the present application, the smart seat stores the data range corresponding to each type of data. For each type of data, the data range corresponding to the data is determined according to the type of the data. Then, when the data is within the data range, it is determined that the data meets the preset condition. When the data is not within the data range, it is determined that the data does not meet the preset condition. Therefore, the embodiment of the present application provides a data detection method. Figure 6 As shown, the specific steps include:

[0148] Step 601, for each type of data to be tested in the reference physiological data, determine the filter to be used to detect the data to be tested according to the type of the data to be tested. When the data to be tested is not filtered by the filter, it is determined that the data to be tested meets the preset conditions. When the data to be tested is filtered by the filter, it is determined that the data to be tested does not meet the preset conditions.

[0149] In this step, when the data to be measured is the heart rate value, blood pressure value and body temperature value, a low-pass filter is used to detect these data, and when the data to be measured is the respiratory rate, a medium-pass filter is used to detect these data. When the data to be measured is not filtered by the corresponding filter, it is determined that the data to be measured meets the preset conditions, and when the data to be measured is filtered by the corresponding filter, it is determined that the data to be measured does not meet the preset conditions.

[0150] The reason why a low-pass filter is used to detect heart rate, blood pressure and body temperature is that under normal circumstances, heart rate, blood pressure and body temperature are relatively small values, and a low-pass filter is needed to effectively remove larger values ​​and retain relatively small values, thereby improving the smoothness and accuracy of the data. The reason why a mid-pass filter is used to detect the respiratory rate is that under normal circumstances, the respiratory rate is relatively close to the middle value, and a mid-pass filter is needed to effectively remove larger and smaller values, thereby improving the smoothness and accuracy of the data.

[0151] Step 602 : when there is no data to be measured that does not meet the preset condition in the reference physiological data, it is determined that the reference physiological data meets the preset condition.

[0152] Step 603: When there is data to be measured that does not meet the preset condition in the reference physiological data, it is determined that the reference physiological data does not meet the preset condition.

[0153] In the embodiment of the present application, the smart seat stores the reference data and preset distance corresponding to each data. For each data, the distance between the data and its corresponding reference data is determined. When the distance is less than or equal to the preset distance, the data is determined to meet the preset condition. When the distance is greater than the preset distance, the data is determined not to meet the preset condition. Therefore, the embodiment of the present application provides a data detection method. Figure 7 As shown, the specific steps include:

[0154] Step 701, for each type of data to be measured in the reference physiological data, determine the benchmark data corresponding to the type according to the type to which the data to be measured belongs and the correspondence between the preset type and the benchmark data. When the distance between the data to be measured and the benchmark data is less than or equal to the preset distance, determine that the data to be measured meets the preset conditions; when the distance between the data to be measured and the benchmark data is greater than the preset distance, determine that the data to be measured does not meet the preset conditions.

[0155] Among them, the benchmark data corresponding to each type of data can be set by the technician based on experience, or can be pre-set by the technician based on the average value of this type of data in the reference sample set. For example, the reference sample includes 10 samples, and each reference sample includes heart rate value, blood pressure value, respiratory rate and body temperature value. The average value corresponding to these 10 heart rate values ​​is calculated, and the average value is determined as the benchmark data corresponding to the heart rate data; the average value corresponding to these 10 blood pressure values ​​is calculated, and the average value is determined as the benchmark data corresponding to the blood pressure data; the average value corresponding to these 10 respiratory rates is calculated, and the average value is determined as the benchmark data corresponding to the respiratory data; the average value corresponding to these 10 body temperature values ​​is calculated, and the average value is determined as the benchmark data corresponding to the body temperature data.

[0156] After determining the reference data corresponding to each type of data, a corresponding relationship between the type of each type of data and its corresponding reference data is established, and the corresponding relationship is stored in the smart seat.

[0157] In this step, for each type of data to be measured in the reference physiological data, the benchmark data corresponding to the type is determined according to the type to which the data to be measured belongs and the correspondence between the preset type and the benchmark data, and then the difference between the data to be measured and the benchmark data is determined, and the absolute value of the difference is obtained. When the absolute value of the difference is less than or equal to the preset distance, it is determined that the data to be measured meets the preset conditions; when the distance between the data to be measured and the benchmark data is greater than the preset distance, it is determined that the data to be measured does not meet the preset conditions.

[0158] The preset distance is set by technicians based on experience, and is generally between 0 and 3.

[0159] Step 702 : when there is no data to be measured that does not meet the preset condition in the reference physiological data, it is determined that the reference physiological data meets the preset condition.

[0160] In this step, when all the data to be measured in the parameter physiological data meet the preset condition, it is determined that the reference physiological data meets the preset condition.

[0161] Step 703: When there is data to be measured that does not meet the preset condition in the reference physiological data, it is determined that the reference physiological data does not meet the preset condition.

[0162] In this step, when a certain data to be measured in the parameter physiological data does not meet the preset condition, it is determined that the reference physiological data where the data to be measured is located does not meet the preset condition.

[0163] like Figure 8 As shown, the embodiment of the present application provides a fatigue warning device, which corresponds to the method embodiment and specifically includes:

[0164] An acquisition unit 801 is used to acquire the driver's initial physiological data and target physiological data collected at the current moment, where the physiological data is composed of data collected by multiple physiological sensors;

[0165] A calling unit 802 is used to call a pre-trained neural network model so that the pre-trained neural network model uses the initial physiological data and the target physiological data to obtain a probability corresponding to each fatigue level;

[0166] A first determination unit 803, configured to determine a target fatigue level of the driver according to a probability corresponding to each fatigue level;

[0167] The second determining unit 804 is configured to determine a fatigue warning method to be used according to the target fatigue level, so as to use the fatigue warning method for warning.

[0168] Optionally, the calling unit 802 is used to:

[0169] Determining difference data according to the target physiological data and the initial physiological data;

[0170] Normalizing the difference data to obtain normalized data;

[0171] The pre-trained neural network model uses the normalized data to obtain the probability corresponding to each fatigue level.

[0172] Optionally, the apparatus further includes a training unit 805, wherein the training unit 805 is configured to:

[0173] acquiring a plurality of reference samples, each reference sample comprising reference physiological data;

[0174] For each reference physiological data in the reference sample, detecting whether the reference physiological data meets a preset condition;

[0175] determining the reference physiological data meeting the preset conditions as the target reference physiological data;

[0176] Determine the reference sample where the target reference physiological data is located as the target sample;

[0177] Combining the target samples to generate a sample set to be used;

[0178] Based on the sample set, the neural network model is trained to obtain a pre-trained neural network model.

[0179] Optionally, the training unit 805 is used to:

[0180] For each type of data to be tested in the reference physiological data, determine a filter for detecting the data to be tested according to the type of the data to be tested, and when the data to be tested is not filtered by the filter, determine that the data to be tested meets the preset condition; when the data to be tested is filtered by the filter, determine that the data to be tested does not meet the preset condition;

[0181] When there is no data to be measured that does not meet the preset condition in the reference physiological data, determining that the reference physiological data meets the preset condition;

[0182] When there is data to be measured that does not meet the preset condition in the reference physiological data, it is determined that the reference physiological data does not meet the preset condition.

[0183] Optionally, the training unit 805 is used to:

[0184] For each type of data to be tested in the reference physiological data, according to the type of the data to be tested and the correspondence between the preset type and the benchmark data, the benchmark data used to detect the data to be tested is determined; when the distance between the data to be tested and the benchmark data is less than or equal to the preset distance, it is determined that the data to be tested meets the preset condition; when the distance between the data to be tested and the benchmark data is greater than the preset distance, it is determined that the data to be tested does not meet the preset condition;

[0185] When there is no data to be measured that does not meet the preset condition in the reference physiological data, determining that the reference physiological data meets the preset condition;

[0186] When there is data to be measured that does not meet the preset condition in the reference physiological data, it is determined that the reference physiological data does not meet the preset condition.

[0187] Optionally, the neural network model includes three layers, namely an input layer, a hidden layer and an output layer. The input layer includes neurons corresponding to each type of data, and the output layer includes neurons corresponding to each fatigue level. The calling unit 802 is used to:

[0188] For each type of data in the normalized data, input the data into a corresponding neuron in the input layer according to the type of the data, so as to input the normalized data into a pre-trained neural network model;

[0189] Using the hidden layer, performing hidden linear transformation and hidden nonlinear transformation on the input data to obtain first linear data;

[0190] Using each neuron in the output layer, output linear transformation and output nonlinear transformation are performed on the linearly processed data to obtain the probability corresponding to each fatigue level.

[0191] Optionally, the hidden layer includes a first hidden sublayer and a second hidden sublayer, and the calling unit 802 is used to:

[0192] Using the first transformation formula in the first hidden sublayer, performing hidden linear mapping on the input data, mapping the input data to a first feature space, and using the first activation function, performing hidden nonlinear transformation on the data mapped to the first feature space to obtain second linear data output by the first hidden sublayer;

[0193] Using the second transformation formula in the second hidden sublayer, the second linear data is hidden linearly mapped, the second linear data is mapped to a second feature space, and using a second activation function, the data mapped to the second feature space is hidden nonlinearly transformed to obtain the first linear data output by the second hidden sublayer.

[0194] like Fig. 9 As shown, the embodiment of the present application provides a fatigue warning device, including a processor 901, a communication interface 902, a memory 903 and a communication bus 904, wherein the processor 901, the communication interface 902, and the memory 903 communicate with each other through the communication bus 904.

[0195] Memory 903, used for storing computer programs;

[0196] In one embodiment of the present application, the processor 901 is used to execute the program stored in the memory 903 to implement the fatigue warning method provided by any of the above method embodiments, including:

[0197] Acquire the driver's initial physiological data and target physiological data collected at the current moment, where the physiological data is composed of data collected by multiple physiological sensors;

[0198] Calling a pre-trained neural network model so that the pre-trained neural network model uses the initial physiological data and the target physiological data to obtain a probability corresponding to each fatigue level;

[0199] Determining a target fatigue level of the driver according to the probability corresponding to each fatigue level;

[0200] A fatigue warning method to be used is determined according to the target fatigue level, and the fatigue warning method is used for warning.

[0201] An embodiment of the present application further provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps performed by the fatigue warning method provided in any of the aforementioned method embodiments are implemented.

[0202] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0203] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a general hardware platform, and of course, by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the relevant technology can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0204] It should be understood that the terms used herein are only for the purpose of describing specific example embodiments and are not intended to be limiting. Unless the context clearly indicates otherwise, the singular forms "one", "an" and "said" as used herein may also be meant to include plural forms. The terms "include", "comprise", "contain", and "have" are inclusive, and therefore specify the existence of stated features, steps, operations, elements and / or parts, but do not exclude the existence or addition of one or more other features, steps, operations, elements, parts, and / or combinations thereof. The method steps, processes, and operations described herein are not interpreted as necessarily requiring them to be performed in the specific order described or illustrated, unless the execution order is clearly indicated. It should also be understood that additional or alternative steps may be used.

[0205] The foregoing is merely a specific embodiment of the present invention, which enables those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A fatigue warning method, characterized in that: The method comprises: Acquire the driver's initial physiological data and target physiological data collected at the current moment, where the physiological data is composed of data collected by multiple physiological sensors; Calling a pre-trained neural network model so that the pre-trained neural network model uses the initial physiological data and the target physiological data to obtain a probability corresponding to each fatigue level; Determining a target fatigue level of the driver according to the probability corresponding to each fatigue level; A fatigue warning method to be used is determined according to the target fatigue level, and the fatigue warning method is used for warning.

2. The method according to claim 1, characterized in that: The pre-trained neural network model uses the initial physiological data and the target physiological data to obtain the probability corresponding to each fatigue level, including: Determining difference data according to the target physiological data and the initial physiological data; Normalizing the difference data to obtain normalized data; The pre-trained neural network model uses the normalized data to obtain the probability corresponding to each fatigue level.

3. The method according to claim 1, characterized in that: The method further comprises: acquiring a plurality of reference samples, each reference sample comprising reference physiological data; For each reference physiological data in the reference sample, detecting whether the reference physiological data meets a preset condition; determining the reference physiological data meeting the preset conditions as the target reference physiological data; Determine the reference sample where the target reference physiological data is located as the target sample; Combining the target samples to generate a sample set to be used; Based on the sample set, the neural network model is trained to obtain a pre-trained neural network model.

4. The method according to claim 3, characterized in that: The detecting whether the reference physiological data meets a preset condition includes: For each type of data to be tested in the reference physiological data, determine a filter for detecting the data to be tested according to the type of the data to be tested, and when the data to be tested is not filtered by the filter, determine that the data to be tested meets the preset condition; when the data to be tested is filtered by the filter, determine that the data to be tested does not meet the preset condition; When there is no to-be-tested data that does not meet the preset condition in the reference physiological data, determining that the reference physiological data meets the preset condition; When there is data to be measured that does not meet the preset condition in the reference physiological data, it is determined that the reference physiological data does not meet the preset condition.

5. The method according to claim 3, characterized in that: The detecting whether the reference physiological data meets a preset condition includes: For each type of data to be tested in the reference physiological data, determine the reference data used to detect the data to be tested according to the type to which the data to be tested belongs and the correspondence between the preset type and the reference data, and when the distance between the data to be tested and the reference data is less than or equal to the preset distance, determine that the data to be tested meets the preset condition; when the distance between the data to be tested and the reference data is greater than the preset distance, determine that the data to be tested does not meet the preset condition; When there is no to-be-tested data that does not meet the preset condition in the reference physiological data, determining that the reference physiological data meets the preset condition; When there is data to be measured that does not meet the preset condition in the reference physiological data, it is determined that the reference physiological data does not meet the preset condition.

6. The method according to claim 2, characterized in that: The neural network model includes three layers, namely an input layer, a hidden layer and an output layer. The input layer includes neurons corresponding to each type of data, and the output layer includes neurons corresponding to each fatigue level. The normalized data is input into the pre-trained neural network model to obtain the probability corresponding to each fatigue level, including: For each type of data in the normalized data, input the data into a corresponding neuron in the input layer according to the type of the data, so as to input the normalized data into a pre-trained neural network model; Using the hidden layer, performing hidden linear transformation and hidden nonlinear transformation on the input data to obtain first linear data; Using each neuron in the output layer, output linear transformation and output nonlinear transformation are performed on the linearly processed data to obtain the probability corresponding to each fatigue level.

7. The method according to claim 6, characterized in that: The hidden layer includes a first hidden sublayer and a second hidden sublayer, and the hidden layer is used to perform a hidden linear transformation and a hidden nonlinear transformation on the input data to obtain first linear data, including: Using the first transformation formula in the first hidden sublayer, performing hidden linear mapping on the input data, mapping the input data to a first feature space, and using the first activation function, performing hidden nonlinear transformation on the data mapped to the first feature space to obtain second linear data output by the first hidden sublayer; Using the second transformation formula in the second hidden sublayer, the second linear data is hidden linearly mapped, the second linear data is mapped to a second feature space, and using a second activation function, the data mapped to the second feature space is hidden nonlinearly transformed to obtain the first linear data output by the second hidden sublayer.

8. A fatigue warning device, characterized in that: The device comprises: An acquisition unit, used to acquire the driver's initial physiological data and target physiological data collected at the current moment, wherein the physiological data is composed of data collected by a variety of physiological sensors; A calling unit, configured to call a pre-trained neural network model, so that the pre-trained neural network model uses the initial physiological data and the target physiological data to obtain a probability corresponding to each fatigue level; A first determination unit, configured to determine a target fatigue level of the driver according to a probability corresponding to each fatigue level; The second determining unit is used to determine a fatigue warning method to be used according to the target fatigue level, so as to use the fatigue warning method for warning.

9. A fatigue warning device, characterized in that: include: at least one communication interface; at least one bus connected to the at least one communication interface; at least one processor connected to the at least one bus; at least one memory connected to the at least one bus, wherein the processor is configured to: Acquire the driver's initial physiological data and target physiological data collected at the current moment, where the physiological data is composed of data collected by multiple physiological sensors; Calling a pre-trained neural network model so that the pre-trained neural network model uses the initial physiological data and the target physiological data to obtain a probability corresponding to each fatigue level; Determining a target fatigue level of the driver according to the probability corresponding to each fatigue level; A fatigue warning method to be used is determined according to the target fatigue level, and the fatigue warning method is used for warning.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the fatigue warning method according to any one of claims 1 to 7 is implemented.

Images