A method and apparatus for extracting samples of abnormal driving states.

Abstract

This application relates to the field of model training technology, specifically to a method and apparatus for extracting abnormal driving state samples. The method includes acquiring normal driving state samples and mixed driving state samples; extracting features from the normal driving state samples and mixed driving state samples respectively to obtain normal feature vectors and mixed feature vectors; determining a similarity score for each mixed driving state sample based on the distance between the mixed feature vector and the normal feature vector; determining a threshold at the boundary between normal and abnormal driving states based on the similarity score; and classifying the mixed driving state samples according to the threshold to obtain normal driving state samples and abnormal driving state samples. This application separates normal driving state samples and abnormal driving state samples from mixed driving state samples, achieving sample purification.

Description

Technical Field

[0001] This application relates to the field of model training technology, and more specifically, to a method and apparatus for extracting samples of abnormal driving states. Background Technology

[0002] Normal driving condition refers to a state in which a driver is in good physical, mental, and emotional condition, and is able to fully and effectively perform all driving tasks (perceiving the environment, making decisions, and precise operations). Abnormal driving condition refers to a state in which a driver's driving ability declines due to various internal and external factors, making it impossible to safely and effectively complete driving tasks. Identifying whether a driver's condition is normal during vehicle operation is a crucial task.

[0003] Sample purity is a crucial factor affecting the accuracy of identification results. Accurate identification of abnormal driving states requires a mixture of samples representing both normal and abnormal driving states. Current driver state databases typically label unclear normal driving states, abnormal driving states, and the transition between abnormal and focused driving states for safety reasons as abnormal driving state samples. This results in abnormal driving state samples being a mixture of various driving states with limited purity. Consequently, subsequent supervised identification based on driving state labels is affected by sample impurity, leading to low identification accuracy, significant discrepancies between the identification results and actual results, and ultimately hindering the practical application of the identification system.

[0004] In view of the above, this application is hereby submitted. Summary of the Invention

[0005] The purpose of this application is to provide a method and apparatus for extracting samples of abnormal driving states, so as to separate normal driving state samples and abnormal driving state samples from mixed driving state samples, thereby achieving sample purification.

[0006] To achieve the above objectives, this application adopts the following technical solution:

[0007] Firstly, this application provides a method for extracting samples of abnormal driving states, including:

[0008] Obtain samples of normal driving state and mixed driving state;

[0009] Feature extraction is performed on the normal driving state samples and the mixed driving state samples respectively to obtain normal feature vectors and mixed feature vectors;

[0010] Based on the distance between the hybrid feature vector and the normal feature vector, a similarity score is determined for each hybrid driving state sample;

[0011] Based on the similarity score, determine the threshold at the boundary between normal driving state and abnormal driving state;

[0012] The mixed driving state samples are classified according to the threshold to obtain normal driving state samples and abnormal driving state samples.

[0013] Secondly, this application provides an electronic device, comprising:

[0014] At least one processor, and a memory communicatively connected to at least one of the processors;

[0015] The memory stores instructions that can be executed by at least one of the processors, which enable the at least one processor to perform the above-described method for extracting abnormal driving state samples.

[0016] Compared with the prior art, the beneficial effects of this application are as follows:

[0017] This embodiment extracts features from normal driving state samples and mixed driving state samples respectively, obtaining normal feature vectors and mixed feature vectors. Based on the distance between the mixed feature vector and the normal feature vector, a similarity score is determined for each mixed driving state sample. Based on the similarity score, a threshold is determined at the boundary between normal and abnormal driving states for classification. This embodiment relies on an existing driver state database, using normal driving state samples as a reference standard, to extract pure abnormal driving state samples from the original, impure, mixed driving state samples that contain multiple driver states. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the specific embodiments of this application or the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0019] Figure 1 This is a flowchart of a method for extracting samples of abnormal driving states provided in an embodiment of this application;

[0020] Figure 2 This is a flowchart of another method for extracting abnormal driving state samples provided in an embodiment of this application;

[0021] Figure 3 This is a schematic diagram of the structure of the electronic device provided in this application. Detailed Implementation

[0022] The following description, in conjunction with the accompanying drawings, illustrates exemplary embodiments of this application, including various details to aid understanding. These should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this application. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0023] The present application will be further described in detail below with reference to the embodiments.

[0024] Figure 1 This is a flowchart illustrating a method for extracting abnormal driving state samples according to an embodiment of this application. This method can be executed by a computer program and integrated into an electronic device. This application is applicable to scenarios where normal driving state samples and abnormal driving state samples are extracted from mixed driving state samples containing various driving states.

[0025] like Figure 1 As shown, this embodiment provides a method for extracting samples of abnormal driving states, including the following steps:

[0026] S110. Obtain samples of normal driving state and mixed driving state.

[0027] The normal driving state sample consists of data collected from drivers during normal driving, including multidimensional physiological data and multidimensional driving behavior data. This data is labeled as "normal."

[0028] Mixed driving state samples are data collected when the driver is in a state that is not clearly normal, a state that is abnormal, or a state that the driver is in while transitioning between abnormal and focused driving states in order to ensure driving safety. These samples also include multidimensional physiological and behavioral data of the driver. Mixed driving state samples are unlabeled.

[0029] Both the normal driving state samples and the mixed driving state samples are multi-dimensional time-series data. The time length of a normal / abnormal driving state sample can be the same or different.

[0030] S120. Extract features from the normal driving state samples and the mixed driving state samples respectively to obtain normal feature vectors and mixed feature vectors.

[0031] This step performs the same feature extraction operation on both normal driving state samples and mixed driving state samples. Feature extraction is performed on normal driving state samples to obtain normal feature vectors, and feature extraction is performed on mixed driving state samples to obtain mixed feature vectors. This embodiment does not limit the feature extraction method; it can be the analytic hierarchy process or expert experience.

[0032] To facilitate the calculation of distances between vectors, normal feature vectors and abnormal feature vectors have the same dimension.

[0033] S130. Determine the similarity score for each mixed driving state sample based on the distance between the mixed feature vector and the normal feature vector.

[0034] For example, calculating the Euclidean distance, Manhattan distance, or cosine distance between the mixed feature vector and the normal feature vector. The closer the mixed feature vector is to the normal feature vector, the more likely it is to be a mixed feature vector; the farther the mixed feature vector is to the normal feature vector, the more likely it is to be a non-normal feature vector.

[0035] In this embodiment, a similarity score is used to evaluate the distance between the mixed feature vector and the normal feature vector; the greater the distance, the lower the similarity score.

[0036] S140. Based on the similarity score, determine the threshold at the boundary between normal driving state and abnormal driving state.

[0037] S150. Classify the mixed driving state samples according to the threshold to obtain normal driving state samples and abnormal driving state samples.

[0038] Each mixed driving state sample has a similarity score. If the similarity score is greater than or equal to the threshold, the mixed driving state sample is classified as a normal driving state sample; if the similarity score is less than the threshold, the mixed driving state sample is classified as an abnormal driving state sample. Thus, the normal / abnormal driving state sample is classified according to the threshold.

[0039] This embodiment extracts features from normal driving state samples and mixed driving state samples respectively, obtaining normal feature vectors and mixed feature vectors. Based on the distance between the mixed feature vector and the normal feature vector, a similarity score is determined for each mixed driving state sample. Based on the similarity score, a threshold is determined at the boundary between normal and abnormal driving states for classification. This embodiment relies on an existing driver state database, using normal driving state samples as a reference standard, to extract pure abnormal driving state samples from the original, impure, mixed driving state samples that contain multiple driver states.

[0040] Figure 2 This is a flowchart of another method for extracting samples of abnormal driving states provided in this application embodiment. Based on the above embodiments, this embodiment refines the process of obtaining normal / mixed feature vectors and the calculation process of similarity scores. See also... Figure 2 The method provided in this embodiment includes:

[0041] S210. Obtain samples of normal driving state and mixed driving state.

[0042] S220. Use the sliding window method to augment data for normal driving state samples and mixed driving state samples.

[0043] The samples from normal driving state and mixed driving state were uniformly cleaned to remove missing measurements. A sliding window method was used to divide the samples from normal driving state and mixed driving state into subsamples with window lengths of a to bs and a sliding step size of s, respectively, to augment the sample data. The window length was determined based on the possible duration of the abnormal driving state, with a recommended range of 5 to 40 seconds.

[0044] S230. Combining paired t-tests, effect sizes, and Spearman correlation coefficients, multidimensional feature indicators are initially selected. The vectors of multidimensional feature indicators for normal driving state samples and mixed driving state samples are respectively processed to be dimensionless. The vectors of the multidimensional feature indicators are dimensionality reduced using a neural network model to obtain normal feature vectors and mixed feature vectors.

[0045] The raw features were constructed using driver physiological data and driving behavior data, as shown in the table below:

[0046]

[0047] First, a paired t-test is used to analyze the significance of each original feature with the target state (i.e., normal driving state and abnormal driving state). Features with p-values ​​less than 0.05 are identified as candidate features, indicating that these features have statistically significant differences. The paired t-test method can be found in existing techniques and will not be elaborated here.

[0048] For each candidate feature, calculate the overall score for each candidate feature using the following formula:

[0049] ;

[0050] in, It is a comprehensive score. It is the p-value obtained from the paired t-test. It is the Cohen effect measure, which is obtained by dividing the difference between the means of the candidate feature in two states (i.e., normal driving state and abnormal driving state) by the pooled standard deviation.

[0051] Next, the top 10 candidate features with the highest overall scores are selected. For each selected candidate feature, the Spearman correlation coefficient is calculated between each pair of features. For highly correlated feature pairs, only the feature with the higher overall score is retained. The final result is a multidimensional feature index with high overall scores and no redundancy between features. For example, the following 10 vectors, which contain both physiological data and driving behavior data, are shown in the table below:

[0052]

[0053] In the table above, the lane keeping score is a metric used to quantitatively evaluate the performance of a vehicle's lane keeping assist system. The lane keeping score is typically based on the following core performance indicators, which are mainly measured using specialized equipment on standard test roads: lateral deviation distance, lane centering ability, and cornering stability. The specific score can be defined by the user. The 5th percentile for steering angle refers to the percentage of all recorded steering angle data that is less than or equal to this value.

[0054] The current feature index vectors have different dimensions, so it is necessary to make the feature index vectors dimensionless to make the vectors of different features comparable.

[0055] ;

[0056] Where X is the dimensionless eigenvector. The original feature vector, Let be the mean of this eigenvector. is the standard deviation of the eigenvector.

[0057] The following section uses a neural network model to further reduce the dimensionality of the feature index vectors. This requires pre-training the neural network model, such as a convolutional neural network. Principal component analysis is performed on the multidimensional feature indices of the normal driving state samples to obtain the first three principal components. These three principal components are then used as the supervision target (i.e., training labels), and the normal driving state samples are used as input to train the neural network model.

[0058] Principal Component Analysis (PCA) is a commonly used method for dimensionality reduction and feature extraction. Its core idea is to map the original high-dimensional feature vectors to a set of mutually orthogonal (uncorrelated) new variables: principal components, through a linear transformation. Existing techniques for PCA methods are available and will not be elaborated upon here. The weight parameters of each layer in the neural network model (such as W and b in fully connected layers) are randomly initialized. The mean squared error (MSE) is used as the loss function. The neural network model is trained to minimize the difference between the output of the neural network model and the PCA result, enabling the neural network model to learn both the linear information compression characteristics of PCA and to uncover more complex relationships between features through nonlinear activation layers.

[0059] ;

[0060] Where M is the number of samples in normal driving conditions. X i It is a vector of dimensionless feature indices. This represents the output of the neural network model. This represents the principal components obtained from principal component analysis.

[0061] After the neural network model is trained, the vector of feature indicators obtained from normal driving samples is input into the neural network model to obtain the normal feature vector; the vector of feature indicators obtained from mixed driving samples is input into the neural network model to obtain the mixed feature vector.

[0062] S240. Calculate the center vector of the normal driving state sample based on the normal feature vectors of all normal driving state samples; obtain the covariance matrix representing the collaborative changes of the normal driving state samples based on the normal feature vector and the center vector of each normal driving state sample; calculate the initial distance between each mixed feature vector and the normal feature vector based on the center vector and the covariance matrix; normalize the initial distance to obtain a similarity score.

[0063] The mean coordinates (i.e., the center vector) of the normal feature vector obtained through dimensionality reduction are calculated in multidimensional space and used as a reference for normal driving conditions. The center vector can be obtained according to the following formula. .

[0064] ;

[0065] Where M is the number of samples in normal driving conditions. Let represent the normal feature vector of the i-th normal driving state sample.

[0066] Since the importance and correlation of multiple normal feature vectors are not the same in the normal driving state sample, it is necessary to calculate the covariance matrix of the normal feature vectors in order to quantify the fluctuation degree of each normal feature vector under normal driving state and the cooperative change trend between normal feature vectors.

[0067] ;

[0068] in, is the covariance matrix. When i=j, it represents the importance of the variation of the normal eigenvectors; when i≠j, it represents the correlation between the normal eigenvectors. To correct the distance between the mixed driving state samples and the center of the normal driving state samples, the initial distance between each mixed eigenvector and the normal eigenvector is calculated:

[0069] ;

[0070] in, It is the hybrid feature vector of the i-th hybrid driving state sample. Let be the covariance matrix mentioned above. Let be the initial distance between the mixed feature vector and the normal feature vector of the i-th mixed driving state sample.

[0071] The initial distance index is then converted into a similarity score between 0 and 1 using the following formula. , where 1 indicates complete similarity and 0 indicates dissimilarity.

[0072] ;

[0073] in, is the similarity score between the i-th mixed driving state sample and the normal feature vector, and N is the number of all mixed driving state samples. max is the maximum value.

[0074] S250. Based on the similarity score, determine the threshold at the boundary between normal driving state and abnormal driving state.

[0075] Optionally, define the groups to which the mixed driving state samples are divided: normal driving state sample group and abnormal driving state sample group. To distinguish between normal and abnormal driving state samples from the mixed driving state samples, thresholds can be determined from two aspects: firstly, maximizing the sample difference between the two groups (i.e., between groups), meaning the mean distance between the two groups is large; secondly, ensuring that the samples within a group are as similar as possible, meaning small internal fluctuations. Based on this, an objective function is constructed by maximizing the difference in similarity scores between groups and minimizing the difference in similarity scores within groups. The objective function B is:

[0076] ;

[0077] ;

[0078] ;

[0079] In the formula, , These represent the sample size of the normal driving state sample group and the abnormal driving state sample group, respectively. , These are the average similarity scores for the normal driving state sample group and the abnormal driving state sample group, respectively. The average similarity score for all samples in the mixed driving state. , Let S be the variance of the similarity scores between the normal driving state sample group and the abnormal driving state sample group, respectively, where N is the number of all mixed driving state samples, and S is the variance of the similarity scores between the normal driving state sample group and the abnormal driving state sample group. b S represents the difference in similarity scores between groups. w This indicates the difference in similarity scores within a group.

[0080] A collaborative search strategy is employed to optimize the objective function (maximizing objective function B) and obtain the threshold at the boundary between normal and abnormal driving states, which serves as the similarity score for group division. The collaborative search strategy starts from a set of randomly initialized candidate solutions (i.e., a set of initialized thresholds) and iteratively updates the position of each solution. In each iteration, the direction of movement of each solution is determined by its historical best position and the current global best position of the group, thus guiding the entire population to efficiently explore and mine the solution space, ultimately converging to the vicinity of the global optimal solution. This embodiment automatically and efficiently finds the optimal segmentation point (i.e., the optimal threshold) among all possible candidate solutions, which best distinguishes the normal driving state sample group from the abnormal driving state sample group, while also ensuring the purity of each sample group. This method avoids the subjectivity and inefficiency of manual trial and error, achieving objective and automated determination of the threshold.

[0081] Group cooperative search strategies include particle swarm optimization algorithm, artificial bee colony algorithm and bacterial foraging algorithm, etc. For specific algorithm flow, please refer to the prior art, and will not be repeated in this embodiment.

[0082] S260. Classify the mixed driving state samples according to the threshold to obtain normal driving state samples and abnormal driving state samples.

[0083] Based on the above description, each mixed driving state sample has a similarity score, and S250 obtains a threshold for segmentation. Samples with a similarity score greater than or equal to the threshold are normal driving state samples, while samples with a similarity score less than the threshold are abnormal driving state samples. Finally, abnormal driving state samples are output, achieving data purification and providing a data foundation for subsequent abnormal driving state identification.

[0084] like Figure 3 As shown, this embodiment provides an electronic device, including:

[0085] At least one processor; and

[0086] A memory that is communicatively connected to at least one processor; wherein,

[0087] The memory stores instructions executable by at least one processor, which, when executed, enables the processor to perform the described method. Since at least one processor in this electronic device is capable of performing the described method, it possesses at least the same advantages as the described method.

[0088] Optionally, the electronic device also includes interfaces for connecting the various components, including high-speed interfaces and low-speed interfaces. The components are interconnected using different buses and can be mounted on a common motherboard or otherwise installed as needed. The processor can process instructions executed within the electronic device, including instructions stored in or on memory to display graphical information of a GUI (Graphical User Interface) on an external input / output device (such as a display device coupled to the interface). In other embodiments, multiple processors can be used with multiple memories, and / or multiple buses can be used with multiple memories, if desired. Similarly, multiple electronic devices (e.g., as a server array, a group of blade servers, or a multiprocessor system) can be connected, each providing some of the necessary operations. Figure 3 Take processor 301 as an example.

[0089] The memory 302, as a computer-readable storage medium, can be used to store software programs, computer-executable programs, and modules, such as the program instructions / modules corresponding to the abnormal driving state sample extraction method in this embodiment. The processor 301 executes various functional applications and data processing of the device by running the software programs, instructions, and modules stored in the memory 302, thereby implementing the above-mentioned abnormal driving state sample extraction method.

[0090] The memory 302 may primarily include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on terminal usage. Furthermore, the memory 302 may include high-speed random access memory and non-volatile memory, such as at least one disk storage device, flash memory device, or other non-volatile solid-state storage device. In some instances, the memory 302 may further include memory remotely located relative to the processor 301, which can be connected to the device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0091] The electronic device may further include an input device 303 and an output device 304. The processor 301, memory 302, input device 303, and output device 304 can be connected via a bus or other means. Figure 3Taking the example of a connection between China and Israel via a bus.

[0092] Input device 303 can receive input digital or character information, and output device 304 may include a display device, an auxiliary lighting device (e.g., an LED), and a haptic feedback device (e.g., a vibration motor). The display device may include, but is not limited to, a liquid crystal display (LCD), a light-emitting diode (LED) display, and a plasma display. In some embodiments, the display device may be a touchscreen.

[0093] This embodiment provides a computer-readable storage medium storing computer instructions for causing a computer to perform the methods described above. The computer instructions on this computer-readable storage medium, used to cause a computer to perform the methods described above, thus have at least the same advantages as the methods described above.

[0094] The medium in this application may be any combination of one or more computer-readable media. The medium may be a computer-readable signal medium or a computer-readable storage medium. The medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of the medium (a non-exhaustive list) include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this document, the medium may be any tangible medium containing or storing a program that may be used by or in connection with an instruction execution system, apparatus, or device.

[0095] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, capable of sending, propagating, or transmitting programs for use by or in connection with an instruction execution system, apparatus, or device.

[0096] Program code contained on a computer-readable medium may be transmitted using any suitable medium, including but not limited to wireless, wire, optical fiber, RF (Radio Frequency), or any suitable combination thereof.

[0097] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof. Programming languages ​​include object-oriented programming languages—such as Java, Smalltalk, and C++—as well as conventional procedural programming languages—such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0098] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions. When the computer instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired means, such as coaxial cable, optical fiber, digital subscriber line (DSL), or wireless means, such as infrared, wireless, microwave, etc. The computer-readable storage medium can be any available medium that a computer can access, or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium, or a semiconductor medium, etc. It is worth noting that the computer-readable storage medium mentioned in the embodiments of this application can be a non-volatile storage medium, in other words, it can be a non-transient storage medium.

[0099] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this application can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this application can be achieved, and this is not limited herein.

[0100] The specific embodiments described above do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for extracting an abnormal driving state sample, characterized by, The method comprises the following steps: obtaining normal driving state samples and mixed driving state samples; extracting features from the normal driving state samples and the mixed driving state samples respectively to obtain normal feature vectors and mixed feature vectors; determining the similarity score of each mixed driving state sample according to the distance between the mixed feature vector and the normal feature vector; determining the threshold value at the boundary between the normal driving state and the abnormal driving state according to the similarity score; classifying the mixed driving state samples according to the threshold value to obtain normal driving state samples and abnormal driving state samples; wherein, determining the threshold value at the boundary between the normal driving state and the abnormal driving state according to the similarity score comprises: defining the normal driving state sample group and the abnormal driving state sample group to which the mixed driving state sample is divided; constructing an objective function by maximizing the difference in similarity scores between groups and minimizing the difference in similarity scores within groups; optimizing the objective function using a group collaboration search strategy to obtain the threshold value at the boundary between the normal driving state and the abnormal driving state; the objective function B is: ； ； ； wherein, , are the average of the similarity scores of the normal driving state sample group and the abnormal driving state sample group, respectively, , are the average of the similarity scores of the normal driving state sample group and the abnormal driving state sample group, respectively, is the average of the similarity scores of all mixed driving state samples, , are the variance of the similarity scores of the normal driving state sample group and the abnormal driving state sample group, respectively, N is the number of all mixed driving state samples, S b represents the difference of the similarity scores between groups, S w represents the difference of the similarity scores within groups.

2. The method of claim 1, wherein the non-normal driving state sample is extracted by: the normal driving state samples and the mixed driving state samples comprise multi-dimensional physiological data and multi-dimensional driving behavior data collected under different states of the driver.

3. The method of claim 1, wherein extracting features from the normal driving state samples and the mixed driving state samples respectively to obtain normal feature vectors and mixed feature vectors comprises: preliminarily selecting multi-dimensional feature indicators in combination with paired t-test, effect size and Spearman correlation coefficient; dimensionless processing the vectors of the multi-dimensional feature indicators of the normal driving state samples and the mixed driving state samples respectively; dimension reduction processing the vectors of the multi-dimensional feature indicators using a neural network model to obtain normal feature vectors and mixed feature vectors.

4. The method of extracting a non-normal driving state sample according to claim 3, wherein, Before dimension reduction processing the vectors of the multi-dimensional feature indicators using a neural network model to obtain normal feature vectors and mixed feature vectors, it further comprises: performing principal component analysis on the multi-dimensional feature indicators of the normal driving state samples to obtain the first three principal components; training the neural network model with the first three principal components as the supervision target.

5. The method of claim 1, wherein, determining the similarity score of each mixed driving state sample according to the distance between the mixed feature vector and the normal feature vector comprises: calculating the center vector of the normal driving state samples according to the normal feature vectors of all normal driving state samples; obtaining the covariance matrix representing the cooperative change of the normal driving state samples according to the normal feature vector of each normal driving state sample and the center vector; calculating the initial distance between each mixed feature vector and the normal feature vector according to the center vector and the covariance matrix; normalizing the initial distance to obtain the similarity score.

6. The method of claim 3, wherein the non-normal driving state sample is extracted by: the multi-dimensional feature indicators comprise: root mean square of respiratory rate change value, low frequency power of heart rate, logarithmic transformed difference of perinasal skin conductance feature, mean value of perinasal skin conductance feature, standard deviation of heart rate, lane keeping score, maximum value of brake force, maximum value of speed, 5% quantile of steering angle, standard deviation of throttle position.

7. The method according to any one of claims 1 to 6, wherein the method is characterized by, Before feature extraction is performed on the normal driving state sample and the mixed driving state sample respectively, the method further comprises: Data augmentation is performed on the normal driving state sample and the mixed driving state sample by using a sliding window method.

8. An electronic device, comprising: The method comprises: at least one processor, and a memory connected in communication with the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method for extracting the non-normal driving state sample according to any one of claims 1-7.

Images