Method and device for evaluating experience of automobile intelligent cockpit function

Abstract

The application relates to the technical field of automobile intelligent cabin evaluation, and provides a method and device for experience evaluation of automobile intelligent cabin functions, the method comprising the following steps: obtaining interaction behavior data generated when a user interacts with each cabin function in an automobile intelligent cabin; determining key behavior characteristics based on the interaction behavior data; wherein the key behavior characteristics comprise at least one of a behavior characteristic of a function ease-of-use index, a behavior characteristic of a function responsiveness index, a behavior characteristic of a function reliability index or a behavior characteristic of a function satisfaction index; inputting the key behavior characteristics into a quantitative evaluation model, and outputting an experience evaluation result of the user on the automobile intelligent cabin function through the quantitative evaluation model; wherein the quantitative evaluation model is obtained by training a long short-term memory (LSTM) network based on a fusion attention mechanism. The technical scheme provided by the application can improve the experience evaluation efficiency and accuracy to a certain extent.

Description

Technical Field

[0001] This application relates to the field of automotive intelligent cockpit evaluation technology, and in particular to a method and apparatus for evaluating the experience of automotive intelligent cockpit functions. Background Technology

[0002] With the development of automotive intelligence and connectivity technologies, the intelligent cockpit is gradually transforming from a simple driving space into a "third living space" integrating social interaction, entertainment, and services. Against this backdrop, how to objectively and efficiently evaluate the cockpit's functional experience has become a key issue for product iteration and user experience optimization.

[0003] Currently, the evaluation methods for intelligent cockpit functions are mainly divided into two categories: manual evaluation and automated evaluation. Manual evaluation includes subjective evaluation and objective evaluation; subjective evaluation includes questionnaires and expert reviews, while objective evaluation includes EEG and eye-tracking tests; automated evaluation relies on preset scripts or simulation scenarios, automatically executing test cases and recording evaluation results through programs.

[0004] However, the aforementioned manual evaluations are limited by individual biases, incomplete scenario coverage, and the high cost and stringent conditions of physiological measurement equipment, which can affect evaluation efficiency and the accuracy of results. While automated evaluations improve efficiency, they are mostly based on idealized scenario designs, which deviate significantly from actual user experiences and cannot accurately reflect users' real needs, thus affecting the accuracy of evaluation results. Therefore, improving the efficiency and accuracy of experience evaluations for automotive intelligent cockpit functions is a technical problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] This application provides a method and apparatus for evaluating the experience of automotive intelligent cockpit functions, which can improve the efficiency and accuracy of experience evaluation to a certain extent when evaluating the experience of automotive intelligent cockpit functions.

[0006] This application provides a method for evaluating the experience of intelligent cockpit functions in automobiles, including: Acquire interaction behavior data generated when users interact with various cockpit functions in the car's smart cockpit; Based on the interaction behavior data, key behavioral features are determined; wherein, the key behavioral features include at least one of the behavioral features of functional usability indicators, functional responsiveness indicators, functional reliability indicators, or functional satisfaction indicators. The key behavioral features are input into a quantitative evaluation model, which then outputs the user's evaluation of the car's smart cockpit functions. The quantitative evaluation model is trained based on a Long Short-Term Memory (LSTM) network with an attention-integrated mechanism.

[0007] According to the experience evaluation method for automotive intelligent cockpit functions provided in this application, the step of inputting the key behavioral features into a quantitative evaluation model and outputting the user's experience evaluation result for the automotive intelligent cockpit functions through the quantitative evaluation model includes: The key behavioral features are input into the Long Short-Term Memory (LSTM) network in the quantitative evaluation model. The temporal dependencies of the key behavioral features are extracted through the LSTM network, and the hidden feature matrix is ​​output. The hidden feature matrix is ​​input into the attention mechanism module in the quantitative evaluation model. The attention mechanism module calculates dynamic weights on the hidden feature matrix to obtain the weighted fused global features. The weighted and fused global features are input into the fully connected network in the quantitative evaluation model, and the experience evaluation result is output through the fully connected network.

[0008] According to the experience evaluation method for automotive intelligent cockpit functions provided in this application, the quantitative evaluation model is trained based on a Long Short-Term Memory (LSTM) network with an attention fusion mechanism, and includes: Obtain a training dataset, which includes key behavioral features and experience evaluation result labels corresponding to historical interaction behavior data; The key behavioral features corresponding to the historical interaction data are used as input, and the experience evaluation result label is used as the target value. A weighted mean squared error loss function is adopted, and the mean squared error loss function is minimized by an optimizer. The Long Short-Term Memory (LSTM) network based on the fusion attention mechanism is trained until the mean squared error on the validation set is lower than a preset threshold, thus obtaining the quantitative evaluation model.

[0009] According to the experience evaluation method for automotive intelligent cockpit functions provided in this application, the method further includes: Obtain the key behavioral features corresponding to the historical interaction behavior data; The key behavioral features corresponding to the historical interaction behavior data are input into the Gradient Boosting Tree (XGBoost) model, and the experience evaluation result label is generated through the Gradient Boosting Tree (XGBoost) model; wherein, the Gradient Boosting Tree (XGBoost) model is trained based on the known experience evaluation results and the corresponding key behavioral features.

[0010] According to the experience evaluation method for automotive intelligent cockpit functions provided in this application, the gradient boosting tree XGBoost model employs a differential regularization strategy during training: Based on the importance of the known experience evaluation results, the corresponding key behavioral features are divided into core features and secondary features; Set a small regularization coefficient for the core feature. Set a larger regularization coefficient for the secondary features. ,and < ; Set a small split gain threshold for the core feature. Set a larger split gain threshold for the secondary feature. ,and < ; The gradient boosting tree XGBoost model is trained based on the regularization coefficient and the split gain threshold.

[0011] According to the experience evaluation method of the intelligent cockpit function of automobile provided in this application, the function usability index includes at least one of the following: task completion rate, average task completion time, number of operation steps, or first-time use success rate. The task completion rate is used to measure the user's ability to successfully complete the cockpit functions; The average task completion time is used to measure the average efficiency of the user in completing the cockpit functions; The number of operation steps is used to measure the average number of operation steps required for the user to complete the cockpit function. The first-time use success rate is used to measure the proportion of users who successfully complete the first use of the cockpit function.

[0012] According to the experience evaluation method of the intelligent cockpit function of automobile provided in this application, the function responsiveness index includes at least one of system response time, task flow time or task switching time; The system response time is used to measure the time required for the cockpit function to respond to user commands; The task flow time is used to measure the total time required for the user to complete a complex task that includes multiple steps. The task switching time is used to measure the time required for the user to switch between different cockpit functions and restore the context.

[0013] According to the experience evaluation method of the intelligent cockpit function of automobile provided in this application, the function reliability index includes at least one of the following: function abnormal exit rate, recognition accuracy rate, or interaction failure rate; The abnormal exit rate is used to measure the percentage of times the cockpit function is abnormally terminated during use without the user's active operation. The recognition accuracy rate is used to measure the proportion of times that the cockpit function correctly recognizes user commands or gestures and executes the expected operation. The interaction failure rate is used to measure the proportion of times that the user's interaction intent is incorrectly recognized or fails to be correctly executed by the cockpit function.

[0014] According to the experience evaluation method of the intelligent cockpit function of automobile provided in this application, the function satisfaction index includes at least one of the following: function usage frequency, single usage duration, active usage rate, or user feedback sentiment analysis. The frequency of use of the function is used to measure the average number of times the user actively uses the cockpit function per unit of time. The single-use duration is used to measure the average duration for which the user starts and uses the cockpit function each time; The active usage rate is used to measure the proportion of times the user actively activates the cockpit function to the total number of times it is used. The user feedback sentiment analysis is used to measure the proportion of positive sentiment among users during the use of the cockpit functions.

[0015] According to the experience evaluation method for automotive intelligent cockpit functions provided in this application, the method further includes: The experience scores corresponding to the functional usability index, functional responsiveness index, functional reliability index, and functional satisfaction index are calculated using a hierarchical weighted method to obtain a comprehensive experience score. Output an automated evaluation report, which uses a bar chart or radar chart to display the experience score corresponding to each indicator and the overall experience score.

[0016] According to the experience evaluation method for automotive intelligent cockpit functions provided in this application, the step of acquiring interaction behavior data generated when a user interacts with various cockpit functions in the automotive intelligent cockpit includes: Acquire initial interaction behavior data generated when users interact with various cockpit functions in the car's smart cockpit; The initial interaction behavior data is sequentially cleaned and standardized to obtain the interaction behavior data. The data cleaning includes outlier detection, duplicate data detection, and missing value imputation; the standardization process includes structured feature transformation and normalization.

[0017] This application also provides an experience evaluation device for automotive intelligent cockpit functions, including: The acquisition unit is used to acquire interaction behavior data generated when the user interacts with various cockpit functions in the car's smart cockpit; The processing unit is configured to determine key behavioral features based on the interactive behavior data; wherein the key behavioral features include at least one of the behavioral features of functional usability indicators, functional responsiveness indicators, functional reliability indicators, or functional satisfaction indicators. The processing unit is further configured to input the key behavioral features into a quantitative evaluation model, and output the user's experience evaluation result of the intelligent cockpit function of the car through the quantitative evaluation model; wherein, the quantitative evaluation model is trained based on a long short-term memory LSTM network with an attention fusion mechanism.

[0018] This application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement an experience evaluation method for automotive smart cockpit functions as described above.

[0019] This application also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements an experience evaluation method for automotive smart cockpit functions as described above.

[0020] This application also provides a computer program product, including a computer program that, when executed by a processor, implements an experience evaluation method for automotive smart cockpit functions as described above.

[0021] The method and apparatus for evaluating the experience of automotive intelligent cockpit functions provided in this application, when evaluating the experience of automotive intelligent cockpit functions, acquires interaction behavior data generated when the user interacts with each cockpit function in the automotive intelligent cockpit; based on the interaction behavior data, key behavioral features are determined; wherein, the key behavioral features include at least one of the behavioral features of functional usability indicators, functional responsiveness indicators, functional reliability indicators, or functional satisfaction indicators; the key behavioral features are input into a quantitative evaluation model, and the quantitative evaluation model outputs the user's experience evaluation result of the automotive intelligent cockpit functions; wherein, the quantitative evaluation model is trained based on a Long Short-Term Memory (LSTM) network with an attention fusion mechanism. Since the interactive behavior data is collected in real time during actual use, and the key behavioral features determined based on the interactive behavior data cover multiple behavioral feature dimensions, it can comprehensively reflect the user's operation performance and avoid the individual bias of traditional manual evaluation. Furthermore, the quantitative evaluation model automatically completes the end-to-end scoring mapping without human intervention, which can effectively improve the evaluation efficiency. In addition, the Long Short-Term Memory (LSTM) network based on the fusion attention mechanism can capture the temporal dependence of behavior, and the attention mechanism focuses on key behavioral features, which improves the consistency between the evaluation results and the actual experience. Therefore, it can improve the efficiency and accuracy of the evaluation of the intelligent cockpit function experience to a certain extent. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the 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 based on these drawings without creative effort.

[0023] Figure 1 This is a flowchart illustrating an experience evaluation method for automotive intelligent cockpit functions provided in an embodiment of this application.

[0024] Figure 2 This is a schematic diagram of the structure of an experience evaluation device for an intelligent cockpit function in an automobile, provided in an embodiment of this application.

[0025] Figure 3 This is a schematic diagram of the physical structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0027] In the embodiments of this application, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone, where A and B can be singular or plural. In the textual description of this application, the character " / " generally indicates that the preceding and following related objects have an "or" relationship.

[0028] The technical solution provided in this application can be applied to the in-vehicle cockpit system of intelligent connected vehicles for automated evaluation of the user experience of intelligent cockpit functions. Currently, intelligent cockpit function experience evaluation methods are mainly divided into two categories: manual evaluation and automated evaluation. Manual evaluation is limited by individual bias, incomplete scenario coverage, and the high cost and stringent conditions of physiological measurement equipment, which affects evaluation efficiency and the accuracy of evaluation results. While automated evaluation improves efficiency, it is mostly based on idealized scenario designs, which deviate significantly from actual user experience and cannot accurately reflect users' real needs, thus affecting the accuracy of evaluation results.

[0029] To improve the efficiency and accuracy of experience evaluation, this application provides a method for evaluating the experience of automotive intelligent cockpit functions. The following specific embodiments will provide a detailed description of this method. It is understood that these specific embodiments can be combined with each other, and similar concepts or processes may not be repeated in some embodiments.

[0030] Figure 1 A flowchart illustrating an experience evaluation method for automotive intelligent cockpit functions provided in this application embodiment is shown below. For example, please refer to... Figure 1 As shown, the experience evaluation method for the intelligent cockpit function of this car may include: S101. Obtain interaction behavior data generated when the user interacts with various cockpit functions in the car's intelligent cockpit.

[0031] For example, interaction behavior data may include at least one of the following: interaction log data, sensor data, vehicle and environment data, or user basic attribute data, which can be set according to actual needs. Interaction log data may include operation sequences, click coordinates, press duration, and user feedback logs; sensor data may include voice commands and facial expression features; vehicle and environment data may include vehicle speed, acceleration, light intensity, and weather data; and user basic attribute data may include age, gender, and driving habits.

[0032] For example, interactive behavior data can be acquired through multi-source data acquisition devices built into the smart cockpit. These devices may include an interaction log module, multiple sensors, and a vehicle-to-everything (V2X) data interface; the interaction log module records system-level interaction events and response information to obtain interaction log data; the multiple sensors include microphones, cameras, touch sensors, and biosensors to acquire sensor data; and the V2X data interface acquires vehicle and environmental data.

[0033] It should be noted that if the interaction behavior data contains user privacy information, such as sensitive personal content or user identification information in user voice commands, the user identification information in the interaction behavior data can be anonymized or desensitized to ensure compliance with relevant data security and privacy protection requirements.

[0034] For example, acquiring interaction behavior data generated when a user interacts with various cockpit functions in a car's smart cockpit may include: The system acquires initial interaction behavior data generated when users interact with various cockpit functions in the car's intelligent cockpit; the initial interaction behavior data is then subjected to data cleaning and standardization processing to obtain the interaction behavior data; data cleaning includes outlier detection, duplicate data detection, and missing value imputation; standardization processing includes structured feature transformation and normalization processing.

[0035] For example, outlier detection can be based on... The outlier detection algorithm removes outlier data from the initial interaction behavior data whose response time exceeds a reasonable range, resulting in outlier-detected interaction behavior data. Duplicate data detection can employ a duplicate data detection algorithm to identify and remove duplicate interaction records generated by sensor malfunctions or system failures from the outlier-detected interaction behavior data, resulting in cleaned interaction behavior data. Missing value imputation can be performed based on the data type of the imputed interaction behavior data, using mean imputation, median imputation, or interpolation based on neighboring data to obtain cleaned interaction behavior data.

[0036] For example, structured feature transformation refers to converting cleaned unstructured interactive behavior data into structured features using natural language processing techniques. For instance, it involves classifying intent, extracting keywords, and analyzing sentiment tendencies for voice commands to generate instruction intent labels and sentiment scores, and converting time data in different formats into a unified timestamp format to obtain structured interactive behavior data. Normalization processing refers to normalizing the numerical data in the structured interactive behavior data, mapping it to the [0, 1] interval to unify the dimensions of the interactive behavior data parameters.

[0037] S102. Based on interactive behavior data, determine key behavioral characteristics; wherein, key behavioral characteristics include at least one of the behavioral characteristics of functional usability indicators, functional responsiveness indicators, functional reliability indicators, or functional satisfaction indicators.

[0038] Among them, key behavioral features are used to reflect the user's experience when using various cockpit functions in the smart cockpit, and serve as the basic input for subsequent quantitative evaluation models to score the experience.

[0039] For example, when determining key behavioral features based on interactive behavior data, rule-based calculation or statistical summarization methods can be used to extract these features. Specifically, rule-based calculation or statistical summarization methods refer to: performing deterministic calculations based on predefined business rules and mathematical formulas for different types of interactive behavior data. This includes subtracting timestamp fields in operation logs to obtain response durations, counting and summing continuous time series data to statistically analyze usage frequency, classifying and counting success and failure markers in task execution records and calculating ratios to obtain completion rates, accumulating the number of events in operation sequences to obtain the number of steps, and calculating sentiment tendency scores for speech and text data through sentiment lexicon matching and polarity scoring. This transforms interactive behavior data into key behavioral features with clear evaluative semantics.

[0040] The behavioral characteristics of the above-mentioned usability indicators are used to measure the efficiency and learning cost of users operating cockpit functions. The core is whether users can quickly master and complete the functions and reduce driving distraction. The function responsiveness indicators are used to measure the response speed and operation smoothness in the interaction between users and cockpit functions. The core goal is to maximize the smoothness of task execution. The function reliability indicators are used to measure the stability and adaptability of cockpit functions in different scenarios and reduce function failures or task interruptions. The function satisfaction indicators are used to measure the emotional experience of users in the process of operating cockpit functions.

[0041] S103. Input key behavioral features into the quantitative evaluation model, and output the user's experience evaluation results of the car's intelligent cockpit function through the quantitative evaluation model; wherein, the quantitative evaluation model is trained based on a long short-term memory LSTM network with an attention fusion mechanism.

[0042] For example, in some embodiments, the experience evaluation results may include at least one of the following: functional usability score, functional responsiveness score, functional reliability score, functional satisfaction score, or comprehensive experience score calculated by the analytic hierarchy process based on the above scores. The specific settings can be configured according to actual needs.

[0043] It is understandable that the functional usability score corresponds to the experience evaluation results of the functional usability index; the functional responsiveness score corresponds to the experience evaluation results of the functional responsiveness index; the functional reliability score corresponds to the experience evaluation results of the functional reliability index; and the functional satisfaction score corresponds to the experience evaluation results of the functional satisfaction index.

[0044] For example, in some embodiments, key behavioral features are input into a quantitative evaluation model, and the model outputs user experience evaluation results for the intelligent cockpit functions of the vehicle, which may include: Key behavioral features are input into the Long Short-Term Memory (LSTM) network in the quantitative evaluation model. The LSTM network extracts the temporal dependencies of key behavioral features and outputs a hidden feature matrix. The hidden feature matrix is ​​then input into the attention mechanism module of the quantitative evaluation model. The attention mechanism calculates dynamic weights on the hidden feature matrix to obtain a weighted and fused global feature. The weighted and fused global feature is then input into the fully connected network of the quantitative evaluation model, which outputs the experience evaluation result. This approach, using the LSTM network to extract the temporal dependencies of key behavioral features, fully captures the sequential relationships in user interactions, avoiding evaluation distortion caused by ignoring the order of behavioral sequences. The attention mechanism's calculation of dynamic weights on the hidden feature matrix adaptively distinguishes the importance of different behavioral features, improving the accuracy of the evaluation results. End-to-end automated evaluation avoids the subjective bias of manual evaluation and the high cost of physiological measurement equipment, improving evaluation efficiency while ensuring accuracy.

[0045] For example, a Long Short-Term Memory (LSTM) network may include an input layer, a hidden layer, a dropout layer, and an output layer. The input layer receives key behavioral features, and its output is connected to the input of the hidden layer to pass these features to the hidden layer. The output of the hidden layer is connected to the input of the dropout layer. The hidden layer uses the ReLU activation function and employs a gating mechanism to extract temporal dependencies from the key behavioral features. This gating mechanism includes a forget gate, an input gate, and an output gate. The forget gate selectively discards irrelevant historical information, the input gate selectively retains key features from the current input, and the output gate controls the output of the hidden state at the current time step. Through the synergistic effect of these gating mechanisms, the hidden layer can dynamically update the network state, fully capturing the long-term and short-term dependencies in user interactions to maintain a stable understanding of user behavior patterns. The hidden layer then passes the first intermediate feature obtained after extracting temporal dependencies to the dropout layer. The output of the dropout layer is connected to the input of the output layer to randomly deactivate some neuron connections in the first intermediate feature to suppress overfitting, obtaining a second intermediate feature, which is then passed to the output layer. The output layer performs a linear transformation on the second intermediate feature, outputting the hidden feature matrix.

[0046] For example, the attention mechanism module calculates dynamic weights for the hidden feature matrix, which may include: connecting the input of the attention mechanism module to the output of the Long Short-Term Memory (LSTM) network to receive the hidden feature matrix output by the LSTM network; performing global max pooling and global average pooling operations on the hidden feature matrix through the pooling layer in the attention mechanism module to obtain max pooling vectors and average pooling vectors; concatenating the max pooling vectors and average pooling vectors along the feature dimension through the concatenation layer in the attention mechanism module to obtain a concatenated vector; inputting the concatenated vector to a single-layer perceptron in the attention mechanism module, performing a nonlinear transformation through the ReLU activation function, and outputting a preliminary score; normalizing the preliminary score through the softmax layer in the attention mechanism module to obtain dynamic weights; and weighting and fusing the dynamic weights with the hidden feature matrix through the weighted fusion layer in the attention mechanism module to obtain weighted and fused global features, which are then output to the fully connected network through the output of the attention mechanism module.

[0047] As can be seen, in this embodiment of the application, when evaluating the experience of the intelligent cockpit function of a car, interactive behavior data generated when the user interacts with each cockpit function in the intelligent cockpit is obtained; based on the interactive behavior data, key behavioral features are determined; wherein, the key behavioral features include at least one of the behavioral features of functional usability indicators, functional responsiveness indicators, functional reliability indicators, or functional satisfaction indicators; the key behavioral features are input into a quantitative evaluation model, and the quantitative evaluation model outputs the user's experience evaluation result of the intelligent cockpit function of the car; wherein, the quantitative evaluation model is trained based on a Long Short-Term Memory (LSTM) network with an attention fusion mechanism. Since the interactive behavior data is collected in real time during actual use, and the key behavioral features determined based on the interactive behavior data cover multiple behavioral feature dimensions, it can comprehensively reflect the user's operation performance and avoid the individual bias of traditional manual evaluation. Furthermore, the quantitative evaluation model automatically completes the end-to-end scoring mapping without human intervention, which can effectively improve the evaluation efficiency. In addition, the Long Short-Term Memory (LSTM) network based on the fusion attention mechanism can capture the temporal dependence of behavior, and the attention mechanism focuses on key behavioral features, which improves the consistency between the evaluation results and the actual experience. Therefore, it can improve the efficiency and accuracy of the evaluation of the intelligent cockpit function experience to a certain extent.

[0048] Based on the above Figure 1 In the illustrated embodiments, for example, in some embodiments, the usability metrics include at least one of the following: task completion rate, average task completion time, number of operation steps, or first-time use success rate, which can be set according to actual needs.

[0049] The task completion rate measures a user's ability to successfully complete cockpit functions. For example, the task completion rate can be calculated using Formula 1 below: Formula 1 in, Indicates the task completion rate. This indicates the number of times a user has successfully completed the core tasks corresponding to the cockpit functions. Completing a task means detecting a clear signal that the task objective has been achieved. If the user actively cancels the task midway, fails to complete it within the time limit, or the task is interrupted due to a system error, it is determined that the task has not been completed. This indicates the total number of times a user attempted to launch the core task of this feature. Attempting to launch the core task of this feature means detecting a clear signal that the user has launched the task.

[0050] Average task completion time measures the average efficiency with which a user completes cockpit functions. For example, the calculation of average task completion time can be seen in Formula 2 below: Formula 2 in, This indicates the average task completion time. A timestamp indicating the time the task was completed. The timestamp indicating when the task started. This indicates the number of times a user has successfully completed the core tasks corresponding to the cockpit functions.

[0051] Operation steps count measures the average number of steps a user needs to complete cockpit functions. Specifically, operation steps count refers to the number of all valid and directly related events recorded between the start and end of a task, filtering out erroneous and invalid operations during the calculation.

[0052] First-time use success rate measures the percentage of users who successfully complete the first use of the cockpit features.

[0053] For example, in some embodiments, the functional responsiveness metrics include at least one of system response time, task flow time, or task switching time, which can be set according to actual needs.

[0054] System response time measures the time required for cockpit functions to respond to user commands. For example, system response time can be calculated using Formula 3 below: Formula 3 in, Indicates system response time. The timestamp indicating the start of system feedback, which includes function execution; The timestamp indicates the end of user command input, which includes voice interaction, physical button input, or gesture interaction.

[0055] Task flow time measures the total time required for a user to complete a complex task involving multiple steps. It focuses on the efficiency of the task operation flow, rather than the speed of a single response. The calculation of task flow time can be seen in Formula 4 below: Formula 4 in, Indicates the time required for the task process. A timestamp indicating the end time of the task. The timestamp indicating when the task started.

[0056] Task switching time measures the time required for a user to switch between different cockpit functions and restore context, thus measuring the smoothness and agility of multitasking.

[0057] For example, in some embodiments, the functional reliability metric may include at least one of the following: functional failure rate, recognition accuracy rate, or interaction failure rate, which can be set according to actual needs.

[0058] The abnormal exit rate measures the percentage of times a cockpit function terminates abnormally without the user's active intervention. For example, the calculation of the abnormal exit rate can be found in Formula 5 below: Formula 5 in, This indicates the rate of abnormal exits due to functional errors. This indicates the number of times the system log records function crash error codes, prolonged periods of unresponsiveness, or the number of times a user is forced to switch to other functions due to lag. This indicates the total number of times the user attempted to launch the core task of this feature.

[0059] Recognition accuracy measures the proportion of times a cockpit function correctly recognizes a user's command or gesture and executes the intended action. For example, recognition accuracy can be calculated using Formula 6 below: Formula 6 in, Indicates recognition accuracy. Indicates the number of times the identification was correct. This indicates the number of times the command or gesture was executed. The number of correct recognitions can be obtained by comparing the voice recognition results or gesture recognition results with the user's true intent, which can be inferred from the user's confirmation or correction of the operation.

[0060] The interaction failure rate measures the proportion of user interaction intentions that are incorrectly recognized or failed to be executed correctly by the cockpit functions. For example, the interaction failure rate can be calculated using Formula 7 below: Formula 7 in, Indicates the interaction failure rate. This indicates the number of times the system failed to produce the expected response. This indicates the total number of interactions. The number of times the system failed to produce the expected response can be inferred from subsequent user correction actions or system error logs. User correction actions include clicking again or repeating the command.

[0061] For example, in some embodiments, the function satisfaction index includes at least one of the following: function usage frequency, single usage duration, active usage rate, or user feedback sentiment analysis, which can be set according to actual needs.

[0062] Function usage frequency measures the average number of times a user actively uses a cockpit function per unit of time. Specifically, function usage frequency can be obtained from function activation logs. When calculating, it is necessary to distinguish between user-initiated activations and system-automatic activations, and only the number of user-initiated activations is counted.

[0063] Duration per use measures the average duration for which a user starts up and uses the cockpit features each time.

[0064] Active usage rate measures the percentage of times a user actively activates a cockpit function out of the total number of uses. Specifically, active usage rate refers to the percentage of times a cockpit function is activated by the user, rather than being pre-set or automatically triggered by the system.

[0065] User feedback sentiment analysis measures the proportion of positive sentiment expressed by users during cockpit use. Specifically, it involves natural language processing of user-initiated voice recordings, system-recorded text feedback, or feedback forms to analyze the percentage of positive sentiment expressed.

[0066] Understandably, the frequency of function usage reflects users' dependence on and habitual preferences for cockpit functions; the duration of a single use reflects function stickiness and user immersion; the active usage rate reflects users' willingness to choose functions autonomously; and user feedback sentiment analysis directly captures users' emotional responses. These four indicators, from different dimensions, collectively constitute the function satisfaction index, which can comprehensively reflect users' emotional experience and subjective satisfaction with cockpit functions.

[0067] Based on any of the above embodiments, the above quantitative evaluation model is obtained by training a Long Short-Term Memory (LSTM) network with an attention fusion mechanism, including: A training dataset is obtained, consisting of key behavioral features corresponding to historical interaction data and experience evaluation result labels. Using the key behavioral features corresponding to historical interaction data as input and the experience evaluation result labels as target values, a weighted mean squared error loss function is employed. An optimizer is used to minimize the mean squared error loss function to train a Long Short-Term Memory (LSTM) network based on a fusion attention mechanism until the mean squared error on the validation set is below a preset threshold, thus obtaining the quantitative evaluation model. This allows the LSTM network to learn the mapping relationship between key behavioral features and experience evaluation results, achieving end-to-end automated evaluation from user interaction behavior to experience rating, avoiding individual bias and high-cost physiological equipment investment inherent in traditional manual evaluation. Furthermore, the weighted mean squared error loss function assigns different weights based on the importance of each key behavioral feature, allowing the quantitative evaluation model to focus on key behavioral features that contribute more significantly to the evaluation, thereby improving the accuracy and reliability of the evaluation results while ensuring the generalization ability of the quantitative evaluation model.

[0068] For example, when obtaining experience evaluation result labels, key behavioral features corresponding to historical interaction behavior data can be acquired. These key behavioral features are then input into a Gradient Boosting Tree (XGBoost) model to generate experience evaluation result labels. The XGBoost model is trained based on known experience evaluation results and corresponding key behavioral features. This efficient and low-cost automatic labeling of large-scale user behavior data without subjective ratings using the XGBoost model solves the problems of high cost and low efficiency associated with traditional manual labeling. Furthermore, the XGBoost model can learn the non-linear mapping relationship between key behavioral features and experience evaluation results, ensuring the accuracy of the generated labels and providing reliable supervision signals for the subsequent training of quantitative evaluation models.

[0069] For example, the aforementioned known experience evaluation results can be obtained in the following way: Select typical users covering different driving environments and scenarios, such as 200-500 people, and pop up a lightweight user experience rating questionnaire on the central control screen of the smart cockpit to collect subjective ratings from typical users on the ease of use, responsiveness, reliability, and satisfaction of the functions; then, invite cockpit experience design experts to perform secondary calibration on the collected subjective ratings, and select valid rating samples as known experience evaluation results.

[0070] For example, the Gradient Boosting Tree XGBoost model employs a differential regularization strategy during training: Based on the known importance of experience evaluation results, the corresponding key behavioral features are divided into core features and secondary features; a smaller regularization coefficient is set for the core features. Set a larger regularization coefficient for secondary features. ,and < Set a smaller split gain threshold for the core features. Set a larger split gain threshold for secondary features. ,and < Based on regularization coefficients and split gain thresholds, a gradient boosting tree (XGBoost) model is trained. This approach divides key behavioral features into core and secondary features according to their importance to the known experience evaluation results. Smaller regularization coefficients and split gain thresholds are set for core features, while larger ones are set for secondary features. This allows the XGBoost model to apply smaller penalties and lower split thresholds to core features during training, thus retaining more information that significantly impacts the experience evaluation. Simultaneously, the larger regularization coefficients and split gain thresholds suppress overfitting of secondary features. This ensures that the XGBoost model focuses more on core features that contribute significantly to the experience evaluation during training, reducing interference from secondary features and improving the accuracy of the experience evaluation result labels.

[0071] For example, when training a quantitative evaluation model, a stratified sampling method can be used to divide the training dataset into a training set and a validation set at a preset ratio, such as 7:3, ensuring that the distribution of the training set and the validation set is consistent across key dimensions. A mean squared error loss function with dynamic weights is designed, and an optimizer, such as the Adam optimizer, is used to minimize the mean squared error between the predicted score and the experience evaluation result label. The optimizer combines momentum and adaptive learning rate to accelerate model training and improve stability. A grid search combined with cross-validation is used to select the optimal hyperparameter combination between the Long Short-Term Memory (LSTM) network and the attention mechanism algorithm until the mean squared error on the validation set is lower than a preset threshold, thus obtaining the quantitative evaluation model.

[0072] Understandably, the XGBoost gradient boosting tree model described above provides efficient and low-cost automatic annotation of large-scale user behavior data without subjective ratings, solving the problems of high cost and low efficiency of traditional manual annotation. It also provides reliable supervision signals for the subsequent training of the quantitative evaluation model. Through differentiated regularization strategies, the XGBoost model focuses more on core features that contribute significantly to the experience evaluation during training, thereby improving the accuracy of the generated labels for the experience evaluation results. Furthermore, by ensuring the consistency of the distribution between the training and validation sets through stratified sampling, and by optimizing hyperparameters through grid search and cross-validation, the model can improve the accuracy and reliability of the evaluation results while maintaining the generalization ability of the quantitative evaluation model.

[0073] Based on any of the above embodiments, for example, in some embodiments, after obtaining the experience scores corresponding to the functional usability index, functional responsiveness index, functional reliability index, and functional satisfaction index, the experience scores corresponding to the functional usability index, functional responsiveness index, functional reliability index, and functional satisfaction index can be further calculated using a hierarchical weighted average to obtain a comprehensive experience score; an automated evaluation report is output, which can display the experience scores and comprehensive experience scores corresponding to each index using bar charts or radar charts. This can intuitively present the performance advantages and disadvantages of the intelligent cockpit functions in different dimensions, helping R&D personnel quickly identify experience shortcomings; for low-scoring dimensions, combined with data to locate specific problems, it can provide a clear direction for targeted optimization of intelligent cockpit functions, realizing a closed-loop iteration of evaluation, positioning, and optimization.

[0074] For example, to ensure the accuracy of the experience evaluation result labels generated by the Gradient Boosting Tree (XGBoost) model, a consistency evaluation of the experience evaluation result labels can be performed. For example, the following methods can be used to perform consistency checks on the experience evaluation result labels: For large-scale user behavior data without subjective ratings, generate experience evaluation result labels using the Gradient Boosting Tree (XGBoost) model; use the Shapley Additive Explanations (SHAP) value visualization tool to generate a SHAP force-directed graph, showing the influence and distribution of each key behavioral feature; filter contradictory samples based on the SHAP force-directed graph and correct the bias of the Gradient Boosting Tree (XGBoost) model to ensure that the experience evaluation result labels generated by the Gradient Boosting Tree (XGBoost) model match the actual user experience logic.

[0075] The following describes the experience evaluation device for the intelligent cockpit function of automobiles provided in this application. The experience evaluation device for the intelligent cockpit function of automobiles described below can be referred to in correspondence with the experience evaluation method for the intelligent cockpit function of automobiles described above.

[0076] Figure 2 This application provides a schematic diagram of the structure of an experience evaluation device for automotive intelligent cockpit functions. For example, please refer to... Figure 2 As shown, the experience evaluation device 20 for the car's smart cockpit function may include: The acquisition unit 201 is used to acquire interaction behavior data generated when the user interacts with various cockpit functions in the intelligent cockpit of the car. Processing unit 202 is configured to determine key behavioral features based on the interactive behavior data; wherein the key behavioral features include at least one of the behavioral features of functional usability indicators, functional responsiveness indicators, functional reliability indicators, or functional satisfaction indicators. The processing unit 202 is further configured to input the key behavioral features into a quantitative evaluation model, and output the user's experience evaluation result of the intelligent cockpit function of the car through the quantitative evaluation model; wherein, the quantitative evaluation model is trained based on a long short-term memory LSTM network with an attention fusion mechanism.

[0077] For example, in an embodiment of this application, the processing unit 202 is used to input the key behavioral features into a quantitative evaluation model, and output the user's experience evaluation result of the intelligent cockpit function of the car through the quantitative evaluation model, including: The key behavioral features are input into the Long Short-Term Memory (LSTM) network in the quantitative evaluation model. The temporal dependencies of the key behavioral features are extracted through the LSTM network, and the hidden feature matrix is ​​output. The hidden feature matrix is ​​input into the attention mechanism module in the quantitative evaluation model. The attention mechanism module calculates dynamic weights on the hidden feature matrix to obtain the weighted fused global features. The weighted and fused global features are input into the fully connected network in the quantitative evaluation model, and the experience evaluation result is output through the fully connected network.

[0078] For example, in an embodiment of this application, the quantitative evaluation model is trained based on a Long Short-Term Memory (LSTM) network with an attention-integrated mechanism, and includes: Obtain a training dataset, which includes key behavioral features and experience evaluation result labels corresponding to historical interaction behavior data; The key behavioral features corresponding to the historical interaction data are used as input, and the experience evaluation result label is used as the target value. A weighted mean squared error loss function is adopted, and the mean squared error loss function is minimized by an optimizer. The Long Short-Term Memory (LSTM) network based on the fusion attention mechanism is trained until the mean squared error on the validation set is lower than a preset threshold, thus obtaining the quantitative evaluation model.

[0079] For example, in this embodiment of the application, the acquisition unit 201 is further configured to acquire key behavioral features corresponding to the historical interaction behavior data; The processing unit 202 is further configured to input the key behavioral features corresponding to the historical interaction behavior data into the gradient boosting tree XGBoost model, and generate the experience evaluation result label through the gradient boosting tree XGBoost model; wherein, the gradient boosting tree XGBoost model is trained based on the known experience evaluation results and the corresponding key behavioral features.

[0080] For example, in an embodiment of this application, the gradient boosting tree XGBoost model employs a differential regularization strategy during training: Based on the importance of the known experience evaluation results, the corresponding key behavioral features are divided into core features and secondary features; Set a small regularization coefficient for the core feature. Set a larger regularization coefficient for the secondary features. ,and < ; Set a small split gain threshold for the core feature. Set a larger split gain threshold for the secondary feature. ,and < ; The gradient boosting tree XGBoost model is trained based on the regularization coefficient and the split gain threshold.

[0081] For example, in the embodiments of this application, the functional usability indicators include at least one of the following: task completion rate, average task completion time, number of operation steps, or first-time use success rate; The task completion rate is used to measure the user's ability to successfully complete the cockpit functions; The average task completion time is used to measure the average efficiency of the user in completing the cockpit functions; The number of operation steps is used to measure the average number of operation steps required for the user to complete the cockpit function. The first-time use success rate is used to measure the proportion of users who successfully complete the first use of the cockpit function.

[0082] For example, in the embodiments of this application, the functional responsiveness index includes at least one of system response time, task flow time, or task switching time; The system response time is used to measure the time required for the cockpit function to respond to user commands; The task flow time is used to measure the total time required for the user to complete a complex task that includes multiple steps. The task switching time is used to measure the time required for the user to switch between different cockpit functions and restore the context.

[0083] For example, in the embodiments of this application, the functional reliability index includes at least one of the following: functional abnormal exit rate, recognition accuracy rate, or interaction failure rate; The abnormal exit rate is used to measure the percentage of times the cockpit function is abnormally terminated during use without the user's active operation. The recognition accuracy rate is used to measure the proportion of times that the cockpit function correctly recognizes user commands or gestures and executes the expected operation. The interaction failure rate is used to measure the proportion of times that the user's interaction intent is incorrectly recognized or fails to be correctly executed by the cockpit function.

[0084] For example, in the embodiments of this application, the function satisfaction index includes at least one of the following: function usage frequency, single usage duration, active usage rate, or user feedback sentiment analysis; The frequency of use of the function is used to measure the average number of times the user actively uses the cockpit function per unit of time. The single-use duration is used to measure the average duration for which the user starts and uses the cockpit function each time; The active usage rate is used to measure the proportion of times the user actively activates the cockpit function to the total number of times it is used. The user feedback sentiment analysis is used to measure the proportion of positive sentiment among users during the use of the cockpit functions.

[0085] For example, in this embodiment of the application, the experience evaluation device 20 for the intelligent cockpit function of the car further includes an output unit; The processing unit 202 is further configured to perform hierarchical weighted calculation on the experience scores corresponding to the functional usability index, the functional responsiveness index, the functional reliability index and the functional satisfaction index to obtain a comprehensive experience score; The output unit is used to output an automated evaluation report, which uses a bar chart or radar chart to display the experience score corresponding to each indicator and the overall experience score.

[0086] For example, in this embodiment of the application, the acquisition unit 201 is used to acquire interaction behavior data generated when a user interacts with various cockpit functions in the car's smart cockpit, including: Acquire initial interaction behavior data generated when users interact with various cockpit functions in the car's smart cockpit; The initial interaction behavior data is sequentially cleaned and standardized to obtain the interaction behavior data. The data cleaning includes outlier detection, duplicate data detection, and missing value imputation; the standardization process includes structured feature transformation and normalization.

[0087] The experience evaluation device 20 for the intelligent cockpit function of automobiles provided in this application embodiment can execute the technical solution of the experience evaluation method for the intelligent cockpit function of automobiles in any of the above embodiments. Its implementation principle and beneficial effects are similar to those of the above-mentioned experience evaluation method for the intelligent cockpit function of automobiles. Please refer to the implementation principle and beneficial effects of the above-mentioned experience evaluation method for the intelligent cockpit function of automobiles, which will not be repeated here.

[0088] Figure 3 This is a schematic diagram of the physical structure of an electronic device provided in an embodiment of this application, such as... Figure 3 As shown, the electronic device may include: a processor 310, a communication interface 320, a memory 330, and a communication bus 340, wherein the processor 310, the communication interface 320, and the memory 330 communicate with each other through the communication bus 340. The processor 310 can call logical instructions in the memory 330 to execute an experience evaluation method for automotive intelligent cockpit functions. This method includes: acquiring interaction behavior data generated when a user interacts with various cockpit functions in the automotive intelligent cockpit; determining key behavioral features based on the interaction behavior data; wherein the key behavioral features include at least one of the behavioral features of functional usability indicators, functional responsiveness indicators, functional reliability indicators, or functional satisfaction indicators; inputting the key behavioral features into a quantitative evaluation model, and outputting the user's experience evaluation result for the automotive intelligent cockpit functions through the quantitative evaluation model; wherein the quantitative evaluation model is trained based on a Long Short-Term Memory (LSTM) network with an attention fusion mechanism.

[0089] Furthermore, the logical instructions in the aforementioned memory 330 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0090] On the other hand, this application also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the experience evaluation method for the intelligent cockpit function of an automobile provided by the above methods. The method includes: acquiring interaction behavior data generated when a user interacts with various cockpit functions in the intelligent cockpit of an automobile; determining key behavioral features based on the interaction behavior data; wherein the key behavioral features include at least one of the behavioral features of functional usability indicators, functional responsiveness indicators, functional reliability indicators, or functional satisfaction indicators; inputting the key behavioral features into a quantitative evaluation model, and outputting the user's experience evaluation result for the intelligent cockpit function of the automobile through the quantitative evaluation model; wherein the quantitative evaluation model is trained based on a Long Short-Term Memory (LSTM) network with an attention fusion mechanism.

[0091] In another aspect, this application also provides a non-transitory computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program implements an experience evaluation method for automotive intelligent cockpit functions provided by the methods described above. This method includes: acquiring interaction behavior data generated when a user interacts with various cockpit functions in the automotive intelligent cockpit; determining key behavioral features based on the interaction behavior data; wherein the key behavioral features include at least one of behavioral features of functional usability indicators, functional responsiveness indicators, functional reliability indicators, or functional satisfaction indicators; inputting the key behavioral features into a quantitative evaluation model, and outputting the user's experience evaluation result for the automotive intelligent cockpit functions through the quantitative evaluation model; wherein the quantitative evaluation model is trained based on a Long Short-Term Memory (LSTM) network with an attention fusion mechanism.

[0092] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0093] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0094] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for evaluating the experience of intelligent cockpit functions in automobiles, characterized in that, include: Acquire interaction behavior data generated when users interact with various cockpit functions in the car's smart cockpit; Based on the interaction behavior data, key behavioral features are determined; wherein, the key behavioral features include at least one of the behavioral features of functional usability indicators, functional responsiveness indicators, functional reliability indicators, or functional satisfaction indicators. The key behavioral features are input into a quantitative evaluation model, which then outputs the user's evaluation of the car's smart cockpit functions. The quantitative evaluation model is trained based on a Long Short-Term Memory (LSTM) network with an attention-integrated mechanism.

2. The method for evaluating the experience of automotive intelligent cockpit functions according to claim 1, characterized in that, The step of inputting the key behavioral features into a quantitative evaluation model and outputting the user's experience evaluation results of the car's intelligent cockpit functions through the quantitative evaluation model includes: The key behavioral features are input into the Long Short-Term Memory (LSTM) network in the quantitative evaluation model. The temporal dependencies of the key behavioral features are extracted through the LSTM network, and the hidden feature matrix is ​​output. The hidden feature matrix is ​​input into the attention mechanism module in the quantitative evaluation model. The attention mechanism module calculates dynamic weights on the hidden feature matrix to obtain the weighted fused global features. The weighted and fused global features are input into the fully connected network in the quantitative evaluation model, and the experience evaluation result is output through the fully connected network.

3. The method for evaluating the experience of automotive intelligent cockpit functions according to claim 1 or 2, characterized in that, The quantitative evaluation model is trained based on a Long Short-Term Memory (LSTM) network with an attention-integrated mechanism, and includes: Obtain a training dataset, which includes key behavioral features and experience evaluation result labels corresponding to historical interaction behavior data; The key behavioral features corresponding to the historical interaction data are used as input, and the experience evaluation result label is used as the target value. A weighted mean squared error loss function is adopted, and the mean squared error loss function is minimized by an optimizer. The Long Short-Term Memory (LSTM) network based on the fusion attention mechanism is trained until the mean squared error on the validation set is lower than a preset threshold, thus obtaining the quantitative evaluation model.

4. The method for evaluating the experience of automotive intelligent cockpit functions according to claim 3, characterized in that, The method further includes: Obtain the key behavioral features corresponding to the historical interaction behavior data; The key behavioral features corresponding to the historical interaction behavior data are input into the Gradient Boosting Tree (XGBoost) model, and the experience evaluation result label is generated through the Gradient Boosting Tree (XGBoost) model; wherein, the Gradient Boosting Tree (XGBoost) model is trained based on the known experience evaluation results and the corresponding key behavioral features.

5. The method for evaluating the experience of intelligent cockpit functions in automobiles according to claim 4, characterized in that, The gradient boosting tree XGBoost model employs a differential regularization strategy during training: Based on the importance of the known experience evaluation results, the corresponding key behavioral features are divided into core features and secondary features; Set a small regularization coefficient for the core feature. Set a larger regularization coefficient for the secondary features. ,and < ; Set a small split gain threshold for the core feature. Set a larger split gain threshold for the secondary feature. ,and < ; The gradient boosting tree XGBoost model is trained based on the regularization coefficient and the split gain threshold.

6. The method for evaluating the experience of automotive intelligent cockpit functions according to claim 1 or 2, characterized in that, The usability metrics include at least one of the following: task completion rate, average task completion time, number of operation steps, or first-time use success rate. The task completion rate is used to measure the user's ability to successfully complete the cockpit functions; The average task completion time is used to measure the average efficiency of the user in completing the cockpit functions; The number of operation steps is used to measure the average number of operation steps required for the user to complete the cockpit function. The first-time use success rate is used to measure the proportion of users who successfully complete the first use of the cockpit function.

7. The method for evaluating the experience of intelligent cockpit functions in automobiles according to claim 1 or 2, characterized in that, The functional responsiveness metrics include at least one of system response time, task flow time, or task switching time. The system response time is used to measure the time required for the cockpit function to respond to user commands; The task flow time is used to measure the total time required for the user to complete a complex task that includes multiple steps. The task switching time is used to measure the time required for the user to switch between different cockpit functions and restore the context.

8. The method for evaluating the experience of intelligent cockpit functions in automobiles according to claim 1 or 2, characterized in that, The functional reliability indicators include at least one of the following: functional abnormal exit rate, recognition accuracy rate, or interaction failure rate. The abnormal exit rate is used to measure the percentage of times the cockpit function is abnormally terminated during use without the user's active operation. The recognition accuracy rate is used to measure the proportion of times that the cockpit function correctly recognizes user commands or gestures and executes the expected operation. The interaction failure rate is used to measure the proportion of times that the user's interaction intent is incorrectly recognized or fails to be correctly executed by the cockpit function.

9. The method for evaluating the experience of intelligent cockpit functions in automobiles according to claim 1 or 2, characterized in that, The function satisfaction index includes at least one of the following: function usage frequency, single usage duration, active usage rate, or user feedback sentiment analysis. The frequency of use of the function is used to measure the average number of times the user actively uses the cockpit function per unit of time. The single-use duration is used to measure the average duration for which the user starts and uses the cockpit function each time; The active usage rate is used to measure the proportion of times the user actively activates the cockpit function to the total number of times it is used. The user feedback sentiment analysis is used to measure the proportion of positive sentiment among users during the use of the cockpit functions.

10. The method for evaluating the experience of intelligent cockpit functions in automobiles according to claim 1 or 2, characterized in that, The method further includes: The experience scores corresponding to the functional usability index, functional responsiveness index, functional reliability index, and functional satisfaction index are calculated using a hierarchical weighted method to obtain a comprehensive experience score. Output an automated evaluation report, which uses a bar chart or radar chart to display the experience score corresponding to each indicator and the overall experience score.

11. The method for evaluating the experience of intelligent cockpit functions in automobiles according to claim 1 or 2, characterized in that, The data obtained when a user interacts with various cockpit functions in the car's smart cockpit includes: Acquire initial interaction behavior data generated when users interact with various cockpit functions in the car's smart cockpit; The initial interaction behavior data is sequentially cleaned and standardized to obtain the interaction behavior data. The data cleaning includes outlier detection, duplicate data detection, and missing value imputation; the standardization process includes structured feature transformation and normalization.

12. A device for evaluating the experience of intelligent cockpit functions in automobiles, characterized in that, include: The acquisition unit is used to acquire interaction behavior data generated when the user interacts with various cockpit functions in the car's smart cockpit; The processing unit is configured to determine key behavioral features based on the interactive behavior data; wherein the key behavioral features include at least one of the behavioral features of functional usability indicators, functional responsiveness indicators, functional reliability indicators, or functional satisfaction indicators. The processing unit is further configured to input the key behavioral features into a quantitative evaluation model, and output the user's experience evaluation result of the intelligent cockpit function of the car through the quantitative evaluation model; wherein, the quantitative evaluation model is trained based on a long short-term memory LSTM network with an attention fusion mechanism.

Images