Automobile function personalized pushing method and system based on end-to-end learning

Abstract

The invention discloses an automobile function personalized pushing method and system based on end-to-end learning, and belongs to the technical field of vehicle engineering. Comprises: acquiring voice data of a driver in a vehicle; inputting the voice data into an optimal character feature prediction model for processing, and obtaining a character feature prediction result corresponding to the driver; combining the character feature prediction result with a user tag, and controlling a function pushing mode of the vehicle; wherein training is carried out through an end-to-end learning strategy by utilizing a pre-trained representation model and a predictor, and an optimal character feature prediction model is obtained. According to the method, personality characteristic identification of the driver is realized through an end-to-end neural network deep learning technology, the development cost is low, iterative upgrading of an algorithm is easier to complete, and the real-time performance is higher; the problems that in the prior art, feature recognition based on voice is low in processing efficiency and high in cost are solved.

Description

Technical Field

[0001] The present invention relates to the field of vehicle engineering technology, and in particular to a method and system for personalized push of automobile functions based on end-to-end learning. Background Art

[0002] The statements in this section merely mention background art related to the present invention and do not necessarily constitute prior art.

[0003] With the development of automobile technology, the functions of automobiles are gradually becoming more humanized, and they can optimize services, upgrade functions and systems, and push corresponding services according to the personalities of drivers and passengers.

[0004] Statistical surveys and studies have shown that there is a close connection between voice and personality traits, and each factor of personality traits (such as extroversion, conscientiousness, openness, emotional stability, etc.) has different performance values ​​in different acoustic and prosodic features. For example, people with a high extroversion index have better voice fluency, fewer pauses, and less silence.

[0005] Based on this, existing personality recognition is mostly based on extracted speech features. This method requires pre-extraction of speech prosodic features, and then establishes effective mapping connections between prosodic features and various personality factors through machine learning training. However, in actual applications, it has the following disadvantages:

[0006] (1) To extract the prosodic features of speech, it is necessary to develop separate speech feature recognition software, which will increase additional development costs and processing procedures and affect the timeliness of data processing.

[0007] (2) After the model is deployed on the cloud, iterative upgrades require reprocessing of sound feature extraction, which makes the iterative upgrade work more cumbersome and increases the workload. Summary of the invention

[0008] In order to address the deficiencies of the prior art, the present invention provides a method, system, electronic device, computer-readable storage medium and computer program product for personalized push of automobile functions based on end-to-end learning. By collecting the driver's smart cockpit command voice, the driver's personality characteristics are identified in the cloud, and the automobile cockpit functions are upgraded and pushed according to the personality characteristics.

[0009] In a first aspect, the present invention provides a method for personalized push of automobile functions based on end-to-end learning;

[0010] A personalized push method for automobile functions based on end-to-end learning, comprising:

[0011] Acquire voice data of the driver in the vehicle;

[0012] Inputting the voice data into the optimal personality characteristic prediction model for processing to obtain the personality characteristic prediction result corresponding to the driver;

[0013] Combining the personality trait prediction result with the user tag to control the function push mode of the vehicle;

[0014] Among them, the pre-trained representation model and predictor are used to train through an end-to-end learning strategy to obtain the optimal personality trait prediction model.

[0015] In some embodiments, the method of using the pre-trained representation model and predictor to train through an end-to-end learning strategy to obtain the optimal personality trait prediction model specifically includes:

[0016] Based on the model parameters of the representation model, a grid search method is used to generate multiple hyperparameter combinations; the initial parameters of the predictor are randomly generated and combined with the representation model corresponding to the hyperparameter combination as the preset model for training;

[0017] The preset model is trained in batches using the training set, and the parameters of the predictor are updated using the back-propagation algorithm. The optimal hyperparameter combination is selected based on the training results.

[0018] The preset model corresponding to the optimal hyperparameter combination is trained through the training set to generate the optimal personality trait prediction model.

[0019] In some embodiments, the optimal personality trait prediction model includes an encoder, a multi-layer perceptron, and a sigmoid activation layer connected in sequence.

[0020] In some implementations, before combining the personality trait prediction result with the user tag, the method further includes: obtaining basic information of the driver, and correcting the personality trait prediction result in combination with the basic information of the driver.

[0021] In some implementations, the correction of the personality trait prediction result in combination with the basic information of the driver is specifically: quantitative correction of the personality trait prediction result in combination with the age and personality trait type of the driver.

[0022] In a second aspect, the present invention provides a personalized push system for automobile functions based on end-to-end learning;

[0023] A personalized push system for automobile functions based on end-to-end learning, including:

[0024] The acquisition module is configured to: acquire voice data of a driver in a vehicle;

[0025] The personality characteristic prediction module is configured to: input the voice data into the optimal personality characteristic prediction model for processing, and obtain the personality characteristic prediction result corresponding to the driver;

[0026] A function personalized push module is configured to: combine the personality feature prediction result with the user tag to control the function push mode of the vehicle;

[0027] Among them, the pre-trained representation model and predictor are used to train through an end-to-end learning strategy to obtain the optimal personality trait prediction model.

[0028] In a third aspect, the present invention provides an electronic device;

[0029] An electronic device includes a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the above-mentioned method for personalized push of automobile functions based on end-to-end learning.

[0030] In a fourth aspect, the present invention provides a computer-readable storage medium;

[0031] A computer-readable storage medium stores a computer program / instruction, which, when executed by a processor, implements the steps of the above-mentioned method for personalized push of automobile functions based on end-to-end learning.

[0032] In a fifth aspect, the present invention provides a computer program product;

[0033] A computer program product includes a computer program / instruction, which, when executed by a processor, implements the steps of the above-mentioned method for personalized push of automobile functions based on end-to-end learning.

[0034] Compared with the prior art, the present invention has the following beneficial effects:

[0035] 1. The technical solution provided by the present invention designs an optimal personality trait prediction model developed through end-to-end neural network deep learning technology. During the learning process, the input data is directly passed to the neural network model for black box learning. Through multiple neural layer processing, the final output result is directly generated. There is no need to extract speech features, which simplifies the process and helps to improve the real-time performance of personality trait prediction.

[0036] 2. The technical solution provided by the present invention makes it easier to complete the iterative upgrade of the algorithm when it is aimed at the entire driving user group.

[0037] 3. The technical solution provided by the present invention fully considers the basic information of the driver to predict the personality characteristics, improves the accuracy and authenticity of the personality characteristics prediction, and at the same time, considers the driver's habits to push personalized functions, which is well matched with the actual driving scene. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0039] Figure 1 A schematic diagram of a process flow provided by an embodiment of the present invention;

[0040] Figure 2 A schematic diagram of a connection system between a driver, a vehicle computer and the cloud provided in an embodiment of the present invention;

[0041] Figure 3 A schematic diagram of a training model flow chart provided by an embodiment of the present invention;

[0042] Figure 4 This is a diagram showing an example of the architecture of the optimal personality trait prediction model provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0043] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used in the present invention have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0044] It should be noted that the terms used herein are only for describing specific embodiments, and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that the terms "include" and "have" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0045] In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other.

[0046] Embodiment 1

[0047] Currently, the personality characteristics of drivers are mostly identified based on extracted voice features, which is cumbersome and not very real-time, and is not conducive to practical applications. In particular, in terms of feature extraction, the traditional method of identifying personality characteristics from voice relies on the method of manually extracting features, and then using machine learning algorithms for identification. However, the method of manually extracting features may result in the omission of some important feature extractions. The end-to-end method can more comprehensively utilize the objective information of the voice itself, thereby having better algorithm performance; therefore, the present invention provides a method for personalized push of automobile functions based on end-to-end learning, which collects the driver's intelligent cockpit command voice, identifies the driver's personality characteristics in the cloud, and then upgrades and pushes the car cockpit functions according to the personality characteristics.

[0048] Next, combine Figure 1-Figure 4 , a method for personalized push of automobile functions based on end-to-end learning disclosed in this embodiment is described in detail. The method for personalized push of automobile functions based on end-to-end learning is deployed in the cloud and includes the following steps:

[0049] S1. Acquire voice data of the driver in the vehicle.

[0050] In this embodiment, when the driver starts to use voice control cockpit, the driver's voice is recorded through the vehicle microphone installed in the cabin; determine whether the sampling frequency of the recording file is less than 16kHz. If it is less than 16kHz, the sound file is directly uploaded to the cloud, and the recording file is converted in the cloud, and the sampling frequency is interpolated and filled to 16kHz. If it is greater than 16kHz, the sampling frequency of the recording file is compressed to 16kHz locally in the vehicle node and then uploaded to the cloud to save bandwidth.

[0051] S2. Input the voice data into the optimal personality trait prediction model deployed in the cloud to obtain the personality trait prediction result corresponding to the driver.

[0052] As an implementation method, before inputting the speech data into the optimal personality trait prediction model, it also includes: using the pre-trained representation model and predictor to train through an end-to-end learning strategy to obtain the optimal personality trait prediction model. The specific process is as follows:

[0053] (1) Data collection.

[0054] Considering that the language characteristics (word usage habits) and phonetic characteristics (prosodic characteristics of speech) of different regions may be different, a certain number of driver volunteers (male-female ratio close to 1:1) were recruited in eight Mandarin dialect areas. The driver volunteers filled out the Big Five Personality Questionnaire to obtain the scores of the five personality traits of the drivers.

[0055] Calculate the overall mean and overall standard deviation of each personality trait. For a single personality trait, if the driver's score is higher than the mean plus the standard deviation, then the classification under this personality trait is recorded as "high". If the driver's score is lower than the mean minus the standard deviation, then the classification under this personality trait is recorded as "low". For the classification model, the "high" and "low" scores of each personality trait are used as training labels. For the regression model, the specific values ​​of each personality trait score are used as training labels. In the driving environment, record the voice operation of the cockpit function, such as "navigate to XX place", "open XX software", "play XX song", "lower the air conditioning temperature", "open / close the window", and other voice commands. Try to cover all voice commands that may occur in the cockpit. The sampling frequency is 16KHz, mono.

[0056] (2) Construct training set and validation set.

[0057] The collected voice data of the driver volunteers are preprocessed, including data cleaning and normalization. In this embodiment, the input of the optimal personality trait prediction model is the voice data, and the output is the specific value of the personality trait score. Based on this, during the training process, the voice data of the driver volunteers and the training labels are associated to construct a data set, and the training labels are the specific values ​​of the corresponding personality trait scores.

[0058] 20% of the sample data in the dataset is divided as a validation set, and the rest is divided as a training set.

[0059] (3) Use the speech neural network model as the representation model, and use the model parameters pre-trained on the existing speech data set as the basic parameters to reduce the amount of training calculations and improve model performance. Freeze the parameters of the above representation model, and connect the output representation of the model to the neural network-based predictor. Use the personality trait label as the prediction target and use the above training set for training.

[0060] Specifically, a grid search method is used to generate multiple sets of hyperparameter combinations. Under each set of hyperparameter combinations, the initial parameters of the predictor are randomly generated multiple times and recorded. The combination of hyperparameters and the randomly generated initial parameter combination is used as the preset model for each training. Based on the preset model, the training set is divided into batches of 32 per time and input into the neural network for batch learning to speed up the training speed and stabilize the training results. The back propagation algorithm is used to update the parameters in the predictor. The training is stopped when the training loss of the validation set tends to be stable.

[0061] (4) Observe and select the best hyperparameter combination in multiple experiments and record it as the most optimal parameter combination. Under the above optimal hyperparameter combination, use the initial parameters in the corresponding training and retrain on the entire data set to generate the optimal personality trait prediction model, and the training is completed.

[0062] In this embodiment, the model architecture of the optimal personality prediction model includes an encoder (representation model), a multilayer perceptron (predictor) and a Sigmoid activation layer connected in sequence. Next, the representation model is an encoder (Encoder) and the predictor is a multilayer perceptron (Multilayer Perceptron) as an example for further explanation.

[0063] Exemplarily, the encoder part of the speechtokenizer_hubert_avg model pre-trained on the LibriSpeech public dataset is used as the representation model; a multilayer perceptron (i.e., multiple layers of fully connected layers) is connected after the representation output, and each fully connected layer uses BatchNorm, ReLU activation layer, Dropout and other anti-overfitting technologies; the output of the multilayer perceptron is passed through a Sigmoid activation layer to output the specific numerical value of each personality trait score.

[0064] Input: Take 16KHz speech clips as input data.

[0065] Output: Output the Big Five personality trait scores of the speaker of the speech clip.

[0066] Training process: set the step size and range of the hyperparameters, train in batches, and find the optimal model hyperparameters through verification. Under the hyperparameters, select the optimal end-to-end model, which is the optimal personality trait prediction model.

[0067] During the training process, based on the black-box learning principle, the encoder and multi-layer perceptron fully learn the features contained in the speech data.

[0068] S3. Setting user tags according to the personality feature prediction results, selecting a target function push mode that matches the user tag from a variety of function push modes, and executing the target function push mode in combination with preset driving habits.

[0069] In this embodiment, the personality trait prediction results refer to the five personality traits of the Big Five, and the preset driving habits refer to the prompt options manually set by the driver, including "temporary exit", "permanent exit" and "custom exit time". The "temporary exit" option means that reminders will no longer be pushed during this driving process, and the "permanent exit" option means that reminders will no longer be pushed during future driving processes; the "custom exit time" option allows the driver to freely choose the time to push reminders again.

[0070] Specifically, the user label is the predicted result of the driver's corresponding personality characteristics, that is, personality traits; according to the user label and the user's current setting conditions, the judgment is made, and different functions are judged and pushed according to different personality traits within the personalized push time set by the driver. The corresponding relationship between the five personality traits of the Big Five and the function push mode is expressed as follows:

[0071] High extroversion: When the social function is turned off, prompt whether it needs to be turned on.

[0072] High Pleasantness: When the traffic sharing function ahead is turned off, it prompts whether to turn it on.

[0073] High Neuroticism: When the road rage emotional counseling function is turned off, it will prompt whether to turn it on.

[0074] High conscientiousness: When the nearby gas station reminder function is turned off, it prompts whether it needs to be turned on.

[0075] High openness: When pushing advertisements, game advertisements are given priority.

[0076] Embodiment 2

[0077] The difference between this embodiment and the first embodiment is that, considering the influence of age and gender on the prediction of the driver's personality, the personality prediction result output by the optimal personality prediction model is corrected, so that the personality prediction result is closer to the driver's actual state. At the same time, the user tag is formed in combination with the driving habit information, and the personalized selection of the function push mode is performed according to the user tag to improve the driver's driving comfort.

[0078] In summary, the method for personalized push of automobile functions based on end-to-end learning described in this embodiment further includes:

[0079] (1) Obtain the basic information of the driver and modify the personality characteristics prediction results based on the basic information of the driver.

[0080] In this embodiment, the basic information of the driver is the driver's age. The existing intelligent driving cockpit system usually obtains the above information of the driver during registration. Here, it can be directly called with the driver's consent.

[0081] Exemplarily, the personality trait prediction result is a personality trait score, and the modified personality trait prediction result is expressed as:

[0082]

[0083] In the formula, G represents the personality trait score, a is a positive or negative number, corresponding to the personality trait. If the personality trait is agreeableness or conscientiousness, a is a positive number, indicating that the corresponding personality trait score will increase with age; if the personality trait is extraversion, neuroticism or openness, a is a negative number, indicating that the corresponding personality trait score will decrease with age; age represents age.

[0084] At the same time, considering the impact of age development on the stability of personal personality, the younger the age, the more likely the personality is to fluctuate; if 30>age>0, then 1+log(age)>a>1, otherwise, then a=1.

[0085] (2) Obtain driving habit information and construct user labels by combining the driving habit information and the corrected personality trait prediction results.

[0086] In this embodiment, the driving habit information includes driving parameters (such as speed, braking force and steering angle, etc.), driving time and other data collected by the driver when driving the vehicle. The driving parameters are collected by the vehicle's built-in sensors. The categories of user tags include safe drivers, adventurous drivers, urban commuters, long-distance travelers and night travel enthusiasts.

[0087] Specifically, based on the driving habit information, the driving habit characteristics of the driver are extracted, including the average speed, number of emergency brakes, steering frequency and driving time distribution. According to the driving habit characteristics and the corrected personality characteristics prediction results, the user label is determined.

[0088] For example, if the conscientiousness score is >70, the number of emergency brakes is <3 times / h and the average speed is ≤ the local speed limit + 10%, the person is marked as a safe driver; if the conscientiousness score is <70, the number of emergency brakes is >5 times / h, the average speed is > the local speed limit + 20% and the turning frequency is greater than 4, the person is marked as an adventurous driver; if the main driving time is concentrated in the morning and evening rush hours, the travel distance is less than 20 kilometers, the average speed is <30kn / h and the extroversion score is ≥60, the person is marked as an urban commuter; if the average single trip distance is >100 kilometers, the average speed is >80km / h and the conscientiousness score is ≥70, the person is marked as a long-distance traveler; if the extroversion score is >60 and the night driving time is >20%, the person is marked as a night travel enthusiast.

[0089] (3) Select a corresponding function push mode according to the user tag, and push the vehicle functions according to the function push mode.

[0090] Exemplarily, the correspondence between user tags and function push modes is as follows:

[0091] Safety drivers: When the function of sharing road conditions ahead is turned off, you will be prompted whether to turn it on; when the function of reminding nearby gas stations is turned off, you will be prompted whether to turn it on.

[0092] Adventurous drivers: When the road rage emotional counseling function is turned off, you will be prompted whether to turn it on.

[0093] Urban commuters: When the traffic sharing feature is turned off, you will be prompted whether to turn it on.

[0094] Long-distance travelers: When the function of sharing road conditions ahead is turned off, you will be prompted whether to turn it on. When the function of reminding nearby gas stations is turned off, you will be prompted whether to turn it on. When the social function is turned off, you will be prompted whether to turn it on.

[0095] Night enthusiasts: When the music function is turned off, you will be prompted whether to turn it on.

[0096] Embodiment 3

[0097] This embodiment discloses a personalized push system for automobile functions based on end-to-end learning, including:

[0098] The acquisition module is configured to: acquire voice data of a driver in a vehicle;

[0099] The personality characteristic prediction module is configured to: input the voice data into the optimal personality characteristic prediction model to obtain the personality characteristic prediction result corresponding to the driver;

[0100] The function personalized push module is configured to: set a user tag according to the personality feature prediction result, and control the function push mode of the vehicle based on the user tag;

[0101] Among them, the pre-trained representation model and predictor are used to train through an end-to-end learning strategy to obtain the optimal personality trait prediction model.

[0102] It should be noted that the acquisition module, personality trait prediction module and personalized function push module correspond to the steps in Embodiment 1, and the examples and application scenarios implemented by the modules are the same as those of the corresponding steps, but are not limited to the contents disclosed in Embodiment 1. It should be noted that the modules as part of the system can be executed in a computer system such as a set of computer executable instructions.

[0103] Embodiment 4

[0104] Embodiment 4 of the present invention provides an electronic device, including a memory and a processor, and computer instructions stored in the memory and executed on the processor. When the computer instructions are executed by the processor, the steps of the above-mentioned method for personalized push of automobile functions based on end-to-end learning are completed.

[0105] Embodiment 5

[0106] Embodiment 5 of the present invention provides a computer-readable storage medium for storing computer instructions. When the computer instructions are executed by a processor, the steps of the above-mentioned method for personalized push of automobile functions based on end-to-end learning are completed.

[0107] Embodiment 6

[0108] Embodiment 6 of the present invention provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the steps of the above-mentioned method for personalized push of automobile functions based on end-to-end learning.

[0109] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0110] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0111] These computer program instructions can also be loaded onto a computer or other programmable data processing device, and a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0112] The description of each embodiment in the above embodiments has different emphases. For parts not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0113] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A personalized push method for automobile functions based on end-to-end learning, characterized in that: include: Acquire voice data of the driver in the vehicle; Inputting the voice data into the optimal personality characteristic prediction model for processing to obtain the personality characteristic prediction result corresponding to the driver; Combining the personality trait prediction result with the user tag to control the function push mode of the vehicle; Among them, the pre-trained representation model and predictor are used to train through an end-to-end learning strategy to obtain the optimal personality trait prediction model.

2. The method for personalized push of automobile functions based on end-to-end learning as claimed in claim 1, characterized in that: The method of using the pre-trained representation model and predictor to train through an end-to-end learning strategy to obtain the optimal personality trait prediction model specifically includes: Based on the model parameters of the representation model, a grid search method is used to generate multiple hyperparameter combinations; the initial parameters of the predictor are randomly generated and combined with the representation model corresponding to the hyperparameter combination as the preset model for training; The preset model is trained in batches using the training set, and the parameters of the predictor are updated using the back-propagation algorithm. The optimal hyperparameter combination is selected based on the training results. The preset model corresponding to the optimal hyperparameter combination is trained through the training set to generate the optimal personality trait prediction model.

3. The method for personalized push of automobile functions based on end-to-end learning as claimed in claim 1, characterized in that: The optimal personality trait prediction model includes an encoder, a multi-layer perceptron and a Sigmoid activation layer connected in sequence.

4. The method for personalized push of automobile functions based on end-to-end learning as claimed in claim 1, characterized in that: Before combining the personality feature prediction result with the user tag, the method further includes: obtaining basic information of the driver, and correcting the personality feature prediction result in combination with the basic information of the driver.

5. The method for personalized push of automobile functions based on end-to-end learning as claimed in claim 4, characterized in that: The correcting of the personality characteristic prediction result in combination with the basic information of the driver specifically includes: quantitatively correcting the personality characteristic prediction result in combination with the age and personality characteristic type of the driver.

6. The method for personalized push of automobile functions based on end-to-end learning as claimed in claim 1, characterized in that: The user tag is constructed by using driving habit information and personality feature prediction results.

7. The personalized push system for automobile functions based on end-to-end learning is characterized by: include: The acquisition module is configured to: acquire voice data of a driver in a vehicle; The personality characteristic prediction module is configured to: input the voice data into the optimal personality characteristic prediction model for processing, and obtain the personality characteristic prediction result corresponding to the driver; A function personalized push module is configured to: combine the personality feature prediction result with the user tag to control the function push mode of the vehicle; Among them, the pre-trained representation model and predictor are used to train through an end-to-end learning strategy to obtain the optimal personality trait prediction model.

8. An electronic device comprising a memory, a processor and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the steps of the method for personalized push of automobile functions based on end-to-end learning as described in any one of claims 1-6.

9. A computer-readable storage medium having a computer program / instruction stored thereon, characterized in that: When the computer program / instruction is executed by a processor, the steps of the method for personalized push of automobile functions based on end-to-end learning as described in any one of claims 1-6 are implemented.

10. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instruction is executed by a processor, the steps of the method for personalized push of automobile functions based on end-to-end learning as described in any one of claims 1-6 are implemented.

Images