A Driving Style Recognition Method and System Based on Multi-kernel Multi-view Clustering

Abstract

This invention belongs to the field of vehicle technology and provides a driving style recognition method and system based on multi-core multi-view clustering. The driving style recognition method based on multi-core multi-view clustering includes: processing the original driving data of the target vehicle to obtain a driving condition view, a driving operation view, and a frequency domain view; calculating a first kernel matrix, a second kernel matrix, and a third kernel matrix based on the driving condition view, the driving operation view, and the frequency domain view, respectively; calculating a first alignment degree, a second alignment degree, and a third alignment degree based on the first kernel matrix, the second kernel matrix, and the third kernel matrix, respectively; calculating a fusion feature matrix based on the first alignment degree, the second alignment degree, and the third alignment degree; and performing clustering processing based on the fusion feature matrix to determine the target driving style. This improves the ability to identify complex behavioral patterns and achieves fine integration and structural alignment of multi-source features.

Description

Technical Field

[0001] This invention relates to the field of vehicle technology, and in particular to a driving style recognition method and system based on multi-core multi-view clustering. Background Technology

[0002] In recent years, as automobiles have become increasingly intelligent, users have placed higher demands on adaptive intelligent control, leading to the emergence of driving style recognition technology. Driving style refers to the stable yet differentiated operating patterns and behavioral characteristics that an individual develops over a long period under specific road, traffic, and environmental conditions. It typically manifests as habitual choices regarding vehicle acceleration, deceleration, steering, and following distance. Driving style reflects the driver's personality traits and style preferences. Furthermore, driving style is influenced by factors such as driving experience, traffic rule awareness, emotional state, and the external environment. Properly classifying driving styles is fundamental to identifying driving patterns and analyzing vehicle operating status, thereby reducing traffic accidents and optimizing vehicle energy management.

[0003] Classifying driving styles typically involves first dividing the original time series into smaller segments using a specific method, and then extracting statistical or frequency domain features from each segment to form a feature vector. Statistical features include average speed and average throttle opening, while frequency domain features include wavelet transform. The unsupervised and high-dimensional nature of driving style data places higher demands on feature extraction and clustering modeling in driving behavior analysis.

[0004] Traditional feature processing methods, such as Principal Component Analysis (PCA), typically assume linear relationships between data and remove redundant features through dimensionality reduction. However, driving style data is often complex and non-linear due to variations in driving behavior. Therefore, using PCA or similar linear dimensionality reduction methods may result in the loss of important non-linear information, thus affecting the quality of clustering results. Traditional feature fusion methods usually employ concatenation or simple weighted averaging to merge data from different views before clustering. However, this method often ignores the differences in importance between views, thus failing to fully consider the complex relationships between them.

[0005] Therefore, how to solve the problems of independent processing of various views and the lack of a joint modeling mechanism in driving style classification is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] In view of this, embodiments of the present invention provide a driving style recognition method and system based on multi-core multi-view clustering to solve the problems of independent processing of each view and lack of joint modeling mechanism in the prior art driving style classification; that is, embodiments of the present invention can improve the ability to identify complex behavior patterns.

[0007] According to one aspect of the present invention, a driving style recognition method based on multi-kernel multi-view clustering is provided. The method includes: processing raw driving data of a target vehicle to obtain a driving condition view, a driving operation view, and a frequency domain view; calculating a first kernel matrix based on the driving condition view, a second kernel matrix based on the driving operation view, and a third kernel matrix based on the frequency domain view; calculating a first alignment degree, a second alignment degree, and a third alignment degree based on the first kernel matrix, the second kernel matrix, and the third kernel matrix; calculating a fusion feature matrix based on the first alignment degree, the second alignment degree, and the third alignment degree; and performing clustering processing based on the fusion feature matrix to determine the target driving style.

[0008] In one embodiment, the process of processing the original driving data of the target vehicle to obtain a driving condition view, a driving operation view, and a frequency domain view includes: acquiring the original driving data of the target vehicle and cleaning the original driving data to obtain target driving data; dividing the target driving data into segments based on a preset driving duration to obtain at least one segment of target driving data; extracting features from the target driving data in the segment to obtain driving condition features, driving operation features, and frequency domain features; standardizing the driving condition features to obtain the driving condition view; standardizing the driving operation features to obtain the driving operation view; and standardizing the frequency domain features to obtain the frequency domain view.

[0009] In one embodiment, the step of calculating the first alignment degree, the second alignment degree, and the third alignment degree based on the first kernel matrix, the second kernel matrix, and the third kernel matrix respectively includes: calculating an average kernel based on the first kernel matrix, the second kernel matrix, and the third kernel matrix; calculating the first alignment degree based on the first kernel matrix and the average kernel; calculating the second alignment degree based on the second kernel matrix and the average kernel; and calculating the third alignment degree based on the third kernel matrix and the average kernel.

[0010] In one embodiment, the step of clustering based on the fused feature matrix to determine the target driving style includes: calculating a Laplace matrix based on the fused feature matrix; and performing clustering based on the Laplace matrix to determine the target driving style.

[0011] In one embodiment, the first alignment degree is calculated based on the first kernel matrix and the average kernel, and the first alignment degree is:

[0012]

[0013] in, For Frobenius inner product, and They are respectively and The Frobenius norm, This is the first kernel matrix. For average kernel.

[0014] In one embodiment, the calculation based on the first alignment, the second alignment, and the third alignment yields a fused feature matrix, which is:

[0015]

[0016] in, As weight, It is the kernel matrix.

[0017] According to another aspect of the present invention, a driving style recognition system based on multi-core multi-view clustering is provided. The driving style recognition system based on multi-core multi-view clustering includes: a data processing module, a first calculation module, a second calculation module, a third calculation module, and a determination module. The data processing module processes raw driving data of a target vehicle to obtain a driving condition view, a driving operation view, and a frequency domain view. The first calculation module calculates a first kernel matrix based on the driving condition view, a second kernel matrix based on the driving operation view, and a third kernel matrix based on the frequency domain view. The second calculation module calculates a first alignment degree, a second alignment degree, and a third alignment degree based on the first kernel matrix, the second kernel matrix, and the third kernel matrix, respectively. The third calculation module calculates a fusion feature matrix based on the first alignment degree, the second alignment degree, and the third alignment degree. The determination module performs clustering processing based on the fusion feature matrix to determine the target driving style.

[0018] In one embodiment, the data processing module includes: a data cleaning unit, a data segmentation unit, a feature extraction unit, and a standardization unit. The data cleaning unit acquires the original driving data of the target vehicle and performs cleaning processing on the original driving data to obtain target driving data. The data segmentation unit divides the target driving data into segments based on a preset driving duration to obtain at least one segment of target driving data. The feature extraction unit extracts features from the target driving data in the segment to obtain driving condition features, driving operation features, and frequency domain features. The standardization unit performs standardization processing based on the driving condition features to obtain the driving condition view, performs standardization processing based on the driving operation features to obtain the driving operation view, and performs standardization processing based on the frequency domain features to obtain the frequency domain view.

[0019] In one embodiment, the second calculation module includes an average kernel calculation unit and an alignment calculation unit, wherein the average kernel calculation unit is used to calculate an average kernel based on the first kernel matrix, the second kernel matrix, and the third kernel matrix; the alignment calculation unit is used to calculate a first alignment degree based on the first kernel matrix and the average kernel, calculate a second alignment degree based on the second kernel matrix and the average kernel, and calculate a third alignment degree based on the third kernel matrix and the average kernel.

[0020] In one embodiment, the determining module includes a calculation unit and a determining unit, wherein the calculation unit is used to calculate based on the fused feature matrix to obtain a Laplacian matrix; and the determining unit is used to perform clustering based on the Laplacian matrix to determine the target driving style.

[0021] In summary, in this embodiment of the invention, the original driving data of the target vehicle is processed to obtain a driving condition view, a driving operation view, and a frequency domain view. A first kernel matrix is ​​calculated based on the driving condition view, a second kernel matrix is ​​calculated based on the driving operation view, and a third kernel matrix is ​​calculated based on the frequency domain view. A first alignment degree, a second alignment degree, and a third alignment degree are calculated based on the first kernel matrix, the second kernel matrix, and the third kernel matrix, respectively. A fusion feature matrix is ​​calculated based on the first alignment degree, the second alignment degree, and the third alignment degree. Clustering processing is performed based on the fusion feature matrix to determine the target driving style, which improves the ability to identify complex behavior patterns. This ensures that important views play a greater role in clustering, reduces interference from weak views, and achieves fine integration and structural alignment of multi-source features, significantly improving clustering accuracy and robustness. Attached Figure Description

[0022] Further details, features, and advantages of the invention are disclosed in the following description of exemplary embodiments in conjunction with the accompanying drawings, in which:

[0023] Figure 1 A flowchart illustrating a driving style recognition method based on multi-core multi-view clustering disclosed in an embodiment of this application is shown.

[0024] Figure 2 It shows Figure 1 The flowchart of step S110 shown is a schematic diagram.

[0025] Figure 3 This shows a portion of the target driving data;

[0026] Figure 4 It shows Figure 1 The flowchart of step S130 shown is shown below;

[0027] Figure 5 It shows Figure 1 The flowchart of step S150 shown is a schematic diagram.

[0028] Figure 6 This illustration shows a schematic diagram of the structure of a driving style recognition system based on multi-core multi-view clustering disclosed in an embodiment of this application. Detailed Implementation

[0029] Embodiments of the present invention will now be described in more detail with reference to the accompanying drawings. While some embodiments of the invention are shown in the drawings, it should be understood that the invention can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of the invention. It should be understood that the accompanying drawings and embodiments are for illustrative purposes only and are not intended to limit the scope of protection of the invention.

[0030] It should be understood that the various steps described in the method embodiments of the present invention may be performed in different orders and / or in parallel. Furthermore, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present invention is not limited in this respect.

[0031] The term "comprising" and its variations as used herein are open-ended, meaning "including but not limited to". The term "based on" means "at least partially based on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". Definitions of other terms will be given in the following description. It should be noted that the concepts of "first", "second", etc., mentioned in this invention are used only to distinguish different devices, modules, or units, and are not intended to limit the order of functions performed by these devices, modules, or units or their interdependencies.

[0032] It should be noted that the terms "a" and "a plurality of" used in this invention are illustrative rather than restrictive. Those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0033] The names of the messages or information exchanged between the multiple devices in the embodiments of the present invention are for illustrative purposes only and are not intended to limit the scope of these messages or information.

[0034] It should be noted that the execution subject of the driving style recognition method based on multi-core multi-view clustering provided in this embodiment of the invention can be one or more electronic devices, and this invention does not limit this. The electronic device can be a terminal (i.e., a client) or a server. Therefore, when the execution subject includes multiple electronic devices, and these multiple electronic devices include at least one terminal and at least one server, the driving style recognition method based on multi-core multi-view clustering provided in this embodiment of the invention can be jointly executed by the terminal and the server. Accordingly, the terminal mentioned herein can include, but is not limited to: smartphones, tablets, laptops, desktop computers, smartwatches, smart voice interaction devices, smart home appliances, vehicle terminals, aircraft, etc. The server mentioned herein can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN (Content Delivery Network), and big data and artificial intelligence platforms, etc.

[0035] Based on the above description, this embodiment of the invention proposes a driving style recognition method based on multi-core multi-view clustering. This driving style recognition method can be executed by the aforementioned electronic device (terminal or server); or, it can be executed jointly by the terminal and the server. For ease of explanation, the following description will use the execution of this driving style recognition method based on multi-core multi-view clustering by an electronic device as an example.

[0036] Please see Figure 1 This is a flowchart illustrating a driving style recognition method based on multi-core multi-view clustering disclosed in an embodiment of this application. The driving style recognition method based on multi-core multi-view clustering addresses the problem of improving the objectivity and consistency of scoring results in driving style recognition technology. It should be noted that the driving style recognition method based on multi-core multi-view clustering in this application is not limited to... Figure 1 The flowchart shown illustrates the steps and their order. Depending on different needs, the steps in the flowchart can be added, removed, or their order changed. In this embodiment, as... Figure 1 As shown, the process of a driving style recognition method based on multi-core multi-view clustering includes at least the following steps.

[0037] S110. Based on the original driving data of the target vehicle, process to obtain a driving condition view, a driving operation view, and a frequency domain view.

[0038] like Figure 2 As shown, in an embodiment of the present invention, Figure 2 Step S110 includes at least the following steps:

[0039] S111. Obtain the original driving data of the target vehicle, and perform cleaning processing on the original driving data to obtain the target driving data.

[0040] In this embodiment of the invention, raw driving data of a target vehicle with a sampling rate of a preset frequency is acquired, and the raw driving data is cleaned to obtain target driving data. The target driving data may include at least vehicle acceleration, speed, and pedal opening. The preset frequency can be 10Hz; however, it is understood that the preset frequency can be set based on actual needs, and this embodiment does not limit it. Partial data of the target driving data is shown below. Figure 3 As shown.

[0041] The raw driving data undergoes cleaning processing to obtain the target driving data. Specifically, firstly, due to sensor errors or environmental interference, the raw driving data may contain noise, which can affect the accuracy of data analysis; secondly, during the data sampling process, there may be discontinuities in time or data loss, requiring interpolation to supplement the data; thirdly, extreme values ​​(i.e., abnormal accelerations) may appear during the data sampling process, distorting the statistical results and requiring removal or correction; fourthly, data fluctuations (i.e., sudden changes in vehicle speed) require smoothing through filtering to reflect the actual driving situation; and fifthly, the target driving data obtained through cleaning processing can more accurately represent the driving cycle and driving style.

[0042] S112. Divide the target driving data into segments based on a preset driving time to obtain at least one segment of target driving data.

[0043] In this embodiment of the invention, the target driving data is segmented based on a preset driving time to obtain at least one segment of target driving data. The preset driving time can be 5 seconds; however, it is understood that the preset driving time can be set based on actual needs, and this embodiment does not limit it in this regard.

[0044] S113. Based on the target driving data in the segment, feature extraction is performed to obtain driving condition features, driving operation features, and frequency domain features.

[0045] In this embodiment of the invention, feature extraction is performed based on the target driving data in each segment to obtain driving condition features, driving operation features, and frequency domain features. The calculation of frequency domain features will be explained as an example.

[0046] The Fast Fourier Transform (FFT) is used to process time-series signals of velocity or acceleration. It transforms the time-series signal from the time domain to the frequency domain, thereby extracting frequency domain features. The FFT formula is as follows:

[0047] Formula (1)

[0048] in, This is the frequency domain representation of the time-domain signal. A time-series signal of velocity or acceleration. For frequency, It is an exponential function (also known as a rotated complex number).

[0049] The energy spectrum of a time-series signal of velocity or acceleration is given by the following formula:

[0050] Formula (2)

[0051] The main frequency is as follows:

[0052] Formula (3)

[0053] The root mean square of the spectrum is given by the following formula:

[0054] Formula (4)

[0055] in, This represents the length of the time series sample.

[0056] The power spectral density is given by the following formula:

[0057] Formula (5)

[0058] in, This refers to the frequency resolution.

[0059] The characteristics of the driving conditions are shown in Table 1 below:

[0060] Table 1 Characteristics of Driving Conditions

[0061]

[0062] The driving operation characteristics are shown in Table 2 below:

[0063]

[0064] The frequency domain characteristics are shown in Table 3 below:

[0065]

[0066] S114. Based on the driving condition features, perform standardization processing to obtain the driving condition view; based on the driving operation features, perform standardization processing to obtain the driving operation view; and based on the frequency domain features, perform standardization processing to obtain the frequency domain view.

[0067] In this embodiment of the invention, standardization processing is performed based on driving condition features, driving operation features, and frequency domain features to obtain a driving condition view, a driving operation view, and a frequency domain view. Each view describes a different aspect of the target driving data segment.

[0068] S120. A first core matrix is ​​calculated based on the driving condition view, a second core matrix is ​​calculated based on the driving operation view, and a third core matrix is ​​calculated based on the frequency domain view.

[0069] In this embodiment of the invention, for the radial basis function (RBF) kernel, the kernel matrix... elements As shown in the following formula:

[0070] Formula (6)

[0071] in, Let i be the feature vector of sample i. Let j be the feature vector of sample j. Let be the squared Euclidean distance between the feature vectors of sample i and the feature vectors of sample j. To control the degree of influence of distance on similarity, The larger the value, the smaller the similarity between samples that are far apart.

[0072] S130. Based on the first kernel matrix, the second kernel matrix, and the third kernel matrix, calculations are performed to obtain the first alignment degree, the second alignment degree, and the third alignment degree, respectively.

[0073] like Figure 4 As shown, in an embodiment of the present invention, Figure 4 Step S130 includes at least the following steps:

[0074] S131. Calculate the average kernel based on the first kernel matrix, the second kernel matrix, and the third kernel matrix.

[0075] In this embodiment of the invention, based on the first kernel matrix Second kernel matrix and the third core matrix Calculations were performed to obtain the average kernel. Average kernel As shown in the following formula:

[0076] Formula (7)

[0077] in, The number of kernel matrices.

[0078] S132. Calculate the first alignment degree based on the first kernel matrix and the average kernel, calculate the second alignment degree based on the second kernel matrix and the average kernel, and calculate the third alignment degree based on the third kernel matrix and the average kernel.

[0079] In this embodiment of the invention, the first alignment degree is taken as an example for explanation. The first alignment degree is calculated based on the first kernel matrix and the average kernel. First alignment As shown in the following formula:

[0080] Formula (8)

[0081] in, For Frobenius inner product, and They are respectively and The Frobenius norm.

[0082] S140. Calculate the fused feature matrix based on the first alignment, the second alignment, and the third alignment.

[0083] In this embodiment of the invention, the first alignment degree is first set. Second alignment and third alignment Normalization is performed to obtain the kernel weights. Core weight As shown in the following formula:

[0084] Formula (9)

[0085] Then, the kernel weights The fusion feature matrix is ​​obtained by weighted summation of the kernel matrix and the fusion feature matrix. As shown in the following formula:

[0086] Formula (10)

[0087] Among them, the fusion feature matrix It reflects the driving style characteristics of each corresponding driving segment, which integrates driving condition characteristics, driving operation characteristics, and frequency domain characteristics.

[0088] S150. Based on the fused feature matrix, perform clustering processing to determine the target driving style.

[0089] like Figure 5 As shown, in an embodiment of the present invention, Figure 5 Step S150 includes at least the following steps:

[0090] S151. Based on the fused feature matrix, the Laplacian matrix is ​​obtained.

[0091] In this embodiment of the invention, the Laplacian matrix is ​​obtained by calculation based on the fused feature matrix. Laplace matrix As shown in the following formula:

[0092] Formula (11)

[0093] Formula (12)

[0094] in, For the angle matrix, This is the sum of the weights of each row in the fused feature matrix.

[0095] S152. Clustering is performed based on the Laplace matrix to determine the target driving style.

[0096] In this embodiment of the invention, the Laplacian matrix is ​​processed using a spectral clustering algorithm. The eigenvalue decomposition is performed as follows:

[0097] Formula (12)

[0098] in, ≤ ≤⋯≤ The eigenvalues ​​are arranged in ascending order. This is the corresponding feature vector.

[0099] Select the eigenvectors corresponding to the first n smallest eigenvalues. When n = 4, the first four eigenvectors are arranged into a matrix as follows:

[0100] Formula (13)

[0101] In each row This provides a low-dimensional representation of the original sample i, thus mapping the original sample from a high-dimensional space to a k-dimensional space that preserves the clustering structure. These feature vectors constitute a low-dimensional embedding space, retaining the clustering structure of the original data.

[0102] Then, the embedding results are clustered using the K-means algorithm, with the number of clusters set to 4 to reflect different driving style types. Finally, each segment is assigned to a cluster, and each cluster represents a driving style.

[0103] In summary, the driving style recognition method based on multi-core multi-view clustering proposed in this application processes the original driving data of the target vehicle to obtain a driving condition view, a driving operation view, and a frequency domain view. A first kernel matrix is ​​calculated based on the driving condition view, a second kernel matrix is ​​calculated based on the driving operation view, and a third kernel matrix is ​​calculated based on the frequency domain view. A first alignment degree, a second alignment degree, and a third alignment degree are calculated based on the first, second, and third kernel matrices, respectively. A fusion feature matrix is ​​then calculated based on the first, second, and third alignment degrees. Clustering processing is performed based on the fusion feature matrix to determine the target driving style. This method improves the ability to identify complex behavioral patterns, ensuring that important views play a greater role in clustering, reducing interference from weak views, and achieving fine integration and structural alignment of multi-source features, significantly improving clustering accuracy and robustness.

[0104] Please see Figure 6 This is a schematic diagram of the structure of a driving style recognition system based on multi-core multi-view clustering disclosed in an embodiment of this application. In one embodiment, such as Figure 6 As shown, this application provides a driving style recognition system 100 based on multi-core multi-view clustering. The driving style recognition system 100 based on multi-core multi-view clustering may include at least: a data processing module 110, a first calculation module 130, a second calculation module 150, a third calculation module 170, and a determination module 180. Information interaction exists between the data processing module 110 and the first calculation module 130, between the first calculation module 130 and the second calculation module 150, between the second calculation module 150 and the third calculation module 170, and between the third calculation module 170 and the determination module 180.

[0105] The data processing module 110 is used to process the raw driving data of the target vehicle to obtain a driving condition view, a driving operation view, and a frequency domain view. The data processing module 110 may include at least: a data cleaning unit 111, a data partitioning unit 113, a feature extraction unit 115, and a standardization processing unit 117.

[0106] The data cleaning unit 111 is used to acquire the raw driving data of the target vehicle and perform cleaning processing on the raw driving data to obtain the target driving data. In this embodiment of the invention, the raw driving data of the target vehicle with a sampling rate of a preset frequency is acquired, and the raw driving data is cleaned to obtain the target driving data. The target driving data may include at least the vehicle's acceleration, speed, pedal opening, etc. The preset frequency can be 10Hz; it is understood that the preset frequency can be set based on actual needs, and this embodiment does not limit it. Partial data of the target driving data is shown below. Figure 3 As shown.

[0107] The raw driving data undergoes cleaning processing to obtain the target driving data. Specifically, firstly, due to sensor errors or environmental interference, the raw driving data may contain noise, which can affect the accuracy of data analysis; secondly, during the data sampling process, there may be discontinuities in time or data loss, requiring interpolation to supplement the data; thirdly, extreme values ​​(i.e., abnormal accelerations) may appear during the data sampling process, distorting the statistical results and requiring removal or correction; fourthly, data fluctuations (i.e., sudden changes in vehicle speed) require smoothing through filtering to reflect the actual driving situation; and fifthly, the target driving data obtained through cleaning processing can more accurately represent the driving cycle and driving style.

[0108] The data segmentation unit 113 is used to segment the target driving data based on a preset driving time to obtain at least one segment of target driving data. In this embodiment of the invention, the target driving data is segmented based on a preset driving time to obtain at least one segment of target driving data. The preset driving time can be 5 seconds. It is understood that the preset driving time can be set based on actual needs, and this embodiment does not limit it in this way.

[0109] The feature extraction unit 115 is used to extract features based on the target driving data in each segment, obtaining driving condition features, driving operation features, and frequency domain features. In this embodiment of the invention, feature extraction is performed based on the target driving data in each segment to obtain driving condition features, driving operation features, and frequency domain features. The calculation of frequency domain features will be explained as an example.

[0110] The Fast Fourier Transform (FFT) is used to process time-series signals of velocity or acceleration. It transforms the time-series signal from the time domain to the frequency domain, thereby extracting frequency domain features. The FFT formula is as follows:

[0111] Formula (1)

[0112] in, This is the frequency domain representation of the time-domain signal. A time-series signal of velocity or acceleration. For frequency, It is an exponential function (also known as a rotated complex number).

[0113] The energy spectrum of a time-series signal of velocity or acceleration is given by the following formula:

[0114] Formula (2)

[0115] The main frequency is as follows:

[0116] Formula (3)

[0117] The root mean square of the spectrum is given by the following formula:

[0118] Formula (4)

[0119] in, This represents the length of the time series sample.

[0120] The power spectral density is given by the following formula:

[0121] Formula (5)

[0122] in, This refers to the frequency resolution.

[0123] The characteristics of the driving conditions are shown in Table 1 below:

[0124] Table 1 Characteristics of Driving Conditions

[0125]

[0126] The driving operation characteristics are shown in Table 2 below:

[0127]

[0128] The frequency domain characteristics are shown in Table 3 below:

[0129]

[0130] The standardization processing unit 117 is used to perform standardization processing based on the driving condition features to obtain the driving condition view, and to perform standardization processing based on the driving operation features to obtain the driving operation view, and to perform standardization processing based on the frequency domain features to obtain the frequency domain view. In this embodiment of the invention, standardization processing is performed based on the driving condition features, driving operation features, and frequency domain features respectively to obtain the driving condition view, driving operation view, and frequency domain view. Different views describe different aspects of the target driving data segment.

[0131] The first calculation module 130 is used to calculate a first kernel matrix based on the driving condition view, a second kernel matrix based on the driving operation view, and a third kernel matrix based on the frequency domain view. In this embodiment of the invention, for a radial basis function (RBF) kernel, the kernel matrix... elements As shown in the following formula:

[0132] Formula (6)

[0133] in, Let i be the feature vector of sample i. Let j be the feature vector of sample j. Let be the squared Euclidean distance between the feature vectors of sample i and the feature vectors of sample j. To control the degree of influence of distance on similarity, The larger the value, the smaller the similarity between samples that are far apart.

[0134] The second calculation module 150 is used to perform calculations based on the first kernel matrix, the second kernel matrix, and the third kernel matrix to obtain a first alignment degree, a second alignment degree, and a third alignment degree, respectively. The second calculation module 150 may include at least an average kernel calculation unit 151 and an alignment degree calculation unit 153.

[0135] The average kernel calculation unit 151 is used to calculate the average kernel based on the first kernel matrix, the second kernel matrix, and the third kernel matrix. In this embodiment of the invention, the average kernel is calculated based on the first kernel matrix. Second kernel matrix and the third core matrix Calculations were performed to obtain the average kernel. Average kernel As shown in the following formula:

[0136] Formula (7)

[0137] in, The number of kernel matrices.

[0138] The alignment calculation unit 153 is used to calculate the first alignment degree based on the first kernel matrix and the average kernel, calculate the second alignment degree based on the second kernel matrix and the average kernel, and calculate the third alignment degree based on the third kernel matrix and the average kernel. In this embodiment of the invention, the first alignment degree is used as an example for explanation; the first alignment degree is calculated based on the first kernel matrix and the average kernel. First alignment As shown in the following formula:

[0139] Formula (8)

[0140] in, For Frobenius inner product, and They are respectively and The Frobenius norm.

[0141] The third calculation module 170 is used to calculate based on the first alignment degree, the second alignment degree, and the third alignment degree to obtain a fused feature matrix. In this embodiment of the invention, the first alignment degree is first calculated... Second alignment and third alignment Normalization is performed to obtain the kernel weights. Core weight As shown in the following formula:

[0142] Formula (9)

[0143] Then, the kernel weights The fusion feature matrix is ​​obtained by weighted summation of the kernel matrix and the fusion feature matrix. As shown in the following formula:

[0144] Formula (10)

[0145] Among them, the fusion feature matrix It reflects the driving style characteristics of each corresponding driving segment, which integrates driving condition characteristics, driving operation characteristics, and frequency domain characteristics.

[0146] The determination module 180 is used to perform clustering processing based on the fused feature matrix to determine the target driving style. The determination module 180 may include at least a calculation unit 181 and a determination unit 183.

[0147] The calculation unit 181 is used to perform calculations based on the fused feature matrix to obtain the Laplacian matrix. In this embodiment of the invention, the Laplacian matrix is ​​obtained by calculation based on the fused feature matrix. Laplace matrix As shown in the following formula:

[0148] Formula (11)

[0149] Formula (12)

[0150] in, For the angle matrix, This is the sum of the weights of each row in the fused feature matrix.

[0151] The determining unit 183 is used to perform clustering based on the Laplacian matrix to determine the target driving style. In this embodiment of the invention, the Laplacian matrix is ​​processed using a spectral clustering algorithm. The eigenvalue decomposition is performed as follows:

[0152] Formula (12)

[0153] in, ≤ ≤⋯≤ The eigenvalues ​​are arranged in ascending order. This is the corresponding feature vector.

[0154] Select the eigenvectors corresponding to the first n smallest eigenvalues. When n = 4, the first four eigenvectors are arranged into a matrix as follows:

[0155] Formula (13)

[0156] In each row This provides a low-dimensional representation of the original sample i, thus mapping the original sample from a high-dimensional space to a k-dimensional space that preserves the clustering structure. These feature vectors constitute a low-dimensional embedding space, retaining the clustering structure of the original data.

[0157] Then, the embedding results are clustered using the K-means algorithm, with the number of clusters set to 4 to reflect different driving style types. Finally, each segment is assigned to a cluster, and each cluster represents a driving style.

[0158] In summary, in the driving style recognition system based on multi-core multi-view clustering of this application, the data processing module 110 processes the original driving data of the target vehicle to obtain a driving condition view, a driving operation view, and a frequency domain view. The first calculation module 130 calculates a first kernel matrix based on the driving condition view, a second kernel matrix based on the driving operation view, and a third kernel matrix based on the frequency domain view. The second calculation module 150 calculates a first alignment, a second alignment, and a third alignment based on the first, second, and third kernel matrices. The third calculation module 170 calculates a fusion feature matrix based on the first, second, and third alignment. The determination module 180 performs clustering processing based on the fusion feature matrix to determine the target driving style. This improves the ability to identify complex behavior patterns, ensuring that important views play a greater role in clustering, reducing interference from weak views, achieving fine integration and structural alignment of multi-source features, and significantly improving clustering accuracy and robustness.

[0159] In the description of this specification, the references to terms such as "an embodiment," "some embodiments," "example," "specific example," "a implementation," "a preferred implementation," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0160] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

Claims

1. A driving style recognition method based on multi-core multi-view clustering, characterized in that, The driving style recognition method based on multi-core multi-view clustering comprises the following steps: Processing the original driving data of the target vehicle to obtain a driving condition view, a driving operation view and a frequency domain view; Calculating a first core matrix based on the driving condition view, a second core matrix based on the driving operation view, and a third core matrix based on the frequency domain view; Calculating a first alignment degree, a second alignment degree and a third alignment degree based on the first core matrix, the second core matrix and the third core matrix; Calculating a fusion feature matrix based on the first alignment degree, the second alignment degree and the third alignment degree; Determining the target driving style based on clustering processing of the fusion feature matrix. 2.The driving style recognition method based on multi-core multi-view clustering according to claim 1, characterized in that, The processing of the original driving data of the target vehicle to obtain a driving condition view, a driving operation view and a frequency domain view comprises the following steps: Obtaining the original driving data of the target vehicle and performing cleaning processing on the original driving data to obtain target driving data; Dividing the target driving data into at least one segment based on a preset driving time length; Extracting features from the target driving data in each segment to obtain driving condition features, driving operation features and frequency domain features; Performing standardization processing on the driving condition features to obtain the driving condition view, performing standardization processing on the driving operation features to obtain the driving operation view, and performing standardization processing on the frequency domain features to obtain the frequency domain view. 3.The driving style recognition method based on multi-core multi-view clustering according to claim 2, characterized in that, The calculation based on the first core matrix, the second core matrix and the third core matrix to obtain a first alignment degree, a second alignment degree and a third alignment degree comprises the following steps: Calculating an average core based on the first core matrix, the second core matrix and the third core matrix; Calculating the first alignment degree based on the first core matrix and the average core, calculating the second alignment degree based on the second core matrix and the average core, and calculating the third alignment degree based on the third core matrix and the average core.

4. The driving style recognition method based on multi-core multi-view clustering according to claim 3, characterized in that, The clustering processing based on the fusion feature matrix to determine the target driving style comprises the following steps: Calculating a Laplacian matrix based on the fusion feature matrix; Determining the target driving style based on clustering of the Laplacian matrix.

5. The driving style recognition method based on multi-core multi-view clustering according to claim 4, characterized in that, The first alignment degree calculated based on the first core matrix and the average core is: where is the Frobenius inner product, and are the Frobenius norms of and respectively, is the first kernel matrix, is the average kernel.

6. The driving style recognition method based on multi-core multi-view clustering according to claim 5, characterized in that, The fusion feature matrix calculated based on the first alignment degree, the second alignment degree and the third alignment degree is: wherein is a weight, is a kernel matrix.

7. A driving style recognition system based on multi-core multi-view clustering, characterized by, The driving style recognition system based on multi-core multi-view clustering comprises a data processing module, a first calculation module, a second calculation module, a third calculation module and a determination module, wherein The data processing module is configured to process the original driving data of the target vehicle to obtain a driving condition view, a driving operation view and a frequency domain view; The first calculation module is configured to calculate a first kernel matrix based on the driving condition view, calculate a second kernel matrix based on the driving operation view, and calculate a third kernel matrix based on the frequency domain view; The second calculation module is configured to calculate a first alignment degree, a second alignment degree, and a third alignment degree based on the first kernel matrix, the second kernel matrix, and the third kernel matrix, respectively; The third calculation module is configured to calculate a fusion feature matrix based on the first alignment degree, the second alignment degree, and the third alignment degree; The determination module is configured to determine a target driving style based on clustering processing of the fusion feature matrix.

8. The driving style recognition system based on multi-core multi-view clustering according to claim 7, characterized in that, The data processing module includes a data cleaning unit, a data division unit, a feature extraction unit, and a standardization processing unit, wherein The data cleaning unit is configured to obtain original driving data of the target vehicle, and clean the original driving data to obtain target driving data; The data division unit is configured to divide the target driving data into at least one segment based on a preset driving time length; The feature extraction unit is configured to extract features based on the target driving data in the one segment to obtain driving condition features, driving operation features, and frequency domain features; The standardization processing unit is configured to standardize the driving condition features to obtain the driving condition view, standardize the driving operation features to obtain the driving operation view, and standardize the frequency domain features to obtain the frequency domain view.

9. The driving style recognition system based on multi-core multi-view clustering according to claim 8, characterized in that, The second calculation module includes an average kernel calculation unit and an alignment degree calculation unit, wherein The average kernel calculation unit is configured to calculate an average kernel based on the first kernel matrix, the second kernel matrix, and the third kernel matrix; The alignment degree calculation unit is configured to calculate the first alignment degree based on the first kernel matrix and the average kernel, calculate the second alignment degree based on the second kernel matrix and the average kernel, and calculate the third alignment degree based on the third kernel matrix and the average kernel.

10. The driving style recognition system based on multi-core multi-view clustering according to claim 9, characterized in that, The determination module includes a calculation unit and a determination unit, wherein The calculation unit is configured to calculate a Laplacian matrix based on the fusion feature matrix; The determination unit is configured to determine the target driving style based on clustering of the Laplacian matrix.

Images