A vehicle model expansion evaluation method and system based on voice middleware

Abstract

The present invention relates to the field of automobile machine technology. The present invention discloses a vehicle model expansion evaluation method and system based on voice middleware, comprising first calculating the deviation distance between a demand solution and M pre-built solutions, determining an initial solution from the M solutions according to the deviation distance, comparing the demand solution with the initial solution, generating a solution comparison result, determining N target solutions based on the solution comparison result and the initial solution, and determining an optimal solution from the N target solutions according to development time, development cost and an intelligent optimization algorithm. The present invention generates a target solution by screening and difference comparison, and selects the optimal solution by an intelligent optimization algorithm, thereby effectively reducing the deep customization requirements for complex projects. In this way, the diversified needs of users can be flexibly met within a limited development cycle and cost, simplifying the development process and improving efficiency.

Description

Technical Field

[0001] The present invention relates to the technical field of automobile machines, and more specifically, to a vehicle model expansion evaluation method and system based on voice middleware. Background Art

[0002] With the in-depth promotion of car screens in domestic and foreign markets, facing the personalized needs of different customers, the customized demand for voice functions continues to increase. At the same time, as the range of compatible vehicle models expands, the overall development and maintenance costs are also rising. In order to meet the needs of the rapid iteration of the car screen market, it is particularly urgent to develop a universal car screen voice solution. The existing voice middleware is difficult to meet the complex needs of customized projects. In addition, the existing technology also lacks a universal voice solution that can flexibly adapt to different customized needs.

[0003] For example, the Chinese patent application with publication number CN116993118A provides a business customization method, apparatus, equipment and medium. The application is based on automatic machine learning technology, combined with the user's actual business customization needs and the collected sample data associated with the customized business needs, to generate an overall business flow program that can be uniformly deployed to the drone end and / or the handheld device end and / or the vehicle end; the overall business flow program can meet the user's customized business needs.

[0004] Although existing technologies can meet users' actual business needs and shorten the development cycle through customization, their reliance on customized requirements for complex projects has not been effectively resolved. With the gradual expansion of vehicle voice services, existing technologies are insufficient in reducing the complexity of customized projects. In particular, in the process of avoiding in-depth customization, it is difficult to flexibly meet the diverse needs of different users while controlling the development cycle and reducing costs.

[0005] In view of this, the present invention proposes a vehicle model expansion evaluation method and system based on voice middleware to solve the above problems. Summary of the invention

[0006] In order to overcome the above-mentioned defects of the prior art, the present invention provides a vehicle model expansion evaluation method and system based on voice middleware.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] In the first aspect, a vehicle model expansion evaluation method based on voice middleware includes:

[0009] Obtain the demand solution corresponding to the model to be expanded, calculate the deviation distance between the demand solution and M pre-built solutions, and determine the initial solution from the M solutions according to the deviation distance. The initial solution is one of the M solutions, where M is an integer greater than 0.

[0010] Compare the demand solution with the initial solution to generate a solution comparison result, determine N target solutions based on the solution comparison result and the initial solution, and calculate the development time and development cost corresponding to the N target solutions, where N is an integer greater than 0;

[0011] The optimal solution is determined from N target solutions based on development time, development cost and intelligent optimization algorithm. The optimal solution is one of the N target solutions.

[0012] Furthermore, the method for calculating the deviation distance between the demand solution and the M pre-built solutions includes:

[0013] The Euclidean distance between each first element in the demand solution and each second element in the solution is calculated, and the calculated Euclidean distances are weighted summed to obtain the deviation distance corresponding to each demand solution, where the first element corresponds to the second element.

[0014] Furthermore, the method of determining an initial solution from the M solutions according to the deviation distance includes:

[0015] The deviation distances corresponding to the M solutions are compared with the preset distance threshold, and the solutions corresponding to those greater than the preset distance threshold are eliminated. The remaining solutions are clustered around the required solution, and the solution closest to the required solution is determined as the initial solution.

[0016] Furthermore, the method for generating the scheme comparison result includes:

[0017] Each first element in the demand solution is reconstructed according to preset rules, and the Pearson correlation coefficient and information entropy are calculated based on each processed first element in the demand solution and the corresponding second element in the initial solution. The Pearson correlation coefficient and information entropy are fused to generate element comparison results, and the set of all element comparison results is used as the solution comparison result.

[0018] Further, the method of calculating the Pearson correlation coefficient includes:

[0019] ;

[0020] In the formula, R is the Pearson correlation coefficient, is the total number of first elements in the demand scenario, For the demand plan The eigenvector value corresponding to the first element, For demand solutions The average value of the eigenvectors corresponding to the first elements, The initial plan The eigenvector value corresponding to the second element, For the initial plan The average value of the eigenvectors corresponding to the second elements.

[0021] Furthermore, the method for calculating information entropy includes:

[0022] ;

[0023] In the formula, SE is information entropy, For the demand plan The weight corresponding to the first element, So The logarithmic function with base , is a natural constant.

[0024] Furthermore, the method for determining N target solutions based on the solution comparison result and the initial solution includes:

[0025] The solution comparison results and the initial solution are input into the pre-built solution generation model to obtain N target solutions output by the solution generation model.

[0026] The construction method of the solution generation model includes:

[0027] Acquire a sample data set, wherein the sample data set includes a historical scheme comparison result, a historical initial scheme, and a historical target scheme;

[0028] Divide the sample data set into a sample training set and a sample test set, and build a regression network;

[0029] The historical scheme comparison results and historical initial schemes in the sample training set are used as the input data of the regression network, and the historical target schemes in the sample training set are used as the output data of the regression network. The regression network is trained to obtain an initial regression network for predicting the target scheme.

[0030] The initial regression network is tested using a sample test set, and the output of the initial regression network that is less than the preset error value is used as the solution generation model.

[0031] Furthermore, the intelligent optimization algorithm is a sparrow search algorithm, and the method of obtaining the optimal solution includes:

[0032] S301: Randomly generate a group of sparrow individuals, each sparrow individual represents a target solution, and randomly generate an initial position for each sparrow individual , that is, the target solution, so that it satisfies:

[0033] ;

[0034] ;

[0035] in, is the development cost corresponding to the target solution, To preset the development cost threshold, is the development time corresponding to the target solution, To preset the development time threshold;

[0036] S302: Construct the fitness function value of each sparrow individual :

[0037] ;

[0038] In the formula, , All are weight factors;

[0039] S303: Perform exploration operations on some sparrow individuals to update their target initial ratio ;

[0040] ;

[0041] In the formula, For the A sparrow in The position of the generation, For the A sparrow in the The position of the generation, is the maximum number of iterations, is a random angle, It is an exponential function with e as the base, where e is a natural constant;

[0042] S304: Check the updated Is it less than If not, perform boundary processing and check the updated Is it less than , if not, perform boundary processing;

[0043] S305: Perform a follow-up operation on the remaining sparrow individuals to update their target initial ratio ;

[0044] S306: Calculate the fitness function value of each sparrow individual after the update , update the current optimal solution;

[0045] S307: Repeat the above steps S303 to S306 until the preset maximum number of iterations is reached. Stop searching when , and output the target solution corresponding to the current optimal solution, which is the optimal solution.

[0046] Furthermore, check the updated Is it less than If not, the boundary processing methods include:

[0047] = ;

[0048] In the formula, It refers to the development cost of the i-th sparrow individual after the t+1 generation update.

[0049] In a second aspect, a vehicle model expansion evaluation system based on voice middleware is used to implement the above-mentioned vehicle model expansion evaluation method based on voice middleware, including:

[0050] Screening module: used to obtain the demand solution corresponding to the model to be expanded, calculate the deviation distance between the demand solution and M pre-built solutions, and determine the initial solution from the M solutions according to the deviation distance. The initial solution is one of the M solutions, where M is an integer greater than 0.

[0051] Solution generation module: used to compare the demand solution with the initial solution, generate solution comparison results, determine N target solutions based on the solution comparison results and the initial solution, and calculate the development time and development cost corresponding to the N target solutions, where N is an integer greater than 0;

[0052] Solution optimization module: used to determine the best solution from N target solutions based on development time, development cost and intelligent optimization algorithm. The best solution is one of the N target solutions.

[0053] In a third aspect, an electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned vehicle model expansion evaluation method based on voice middleware when executing the computer program.

[0054] In a fourth aspect, a computer-readable storage medium stores a computer program, and when the computer program is executed, the vehicle model expansion evaluation method based on voice middleware is implemented.

[0055] Compared with the prior art, the present invention has the following beneficial effects:

[0056] The present invention first calculates the deviation distance between the demand solution and M pre-built solutions, determines the initial solution from the M solutions according to the deviation distance, compares the demand solution with the initial solution, generates a solution comparison result, determines N target solutions based on the solution comparison result and the initial solution, and determines the optimal solution from the N target solutions according to the development time, development cost and intelligent optimization algorithm. The present invention generates target solutions by screening and difference comparison, and selects the optimal solution through intelligent optimization algorithm, which effectively reduces the deep customization requirements for complex projects. In this way, it can flexibly meet the diverse needs of users within a limited development cycle and cost, simplify the development process and improve efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 A flow chart of a vehicle model expansion evaluation method based on voice middleware in the present invention;

[0058] Figure 2 It is a structural schematic diagram of a vehicle model expansion evaluation system based on voice middleware in the present invention;

[0059] Figure 3 This is a functional diagram of the voice middleware in the present invention;

[0060] Figure 4 A schematic diagram of the process of generating scheme comparison results in the present invention. DETAILED DESCRIPTION

[0061] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0062] Example 1

[0063] See also Figure 1 As shown, this embodiment discloses a vehicle model expansion evaluation method based on voice middleware, including:

[0064] S10: Obtain the demand solution corresponding to the vehicle model to be expanded, calculate the deviation distance between the demand solution and the M pre-built solutions, and determine an initial solution from the M solutions according to the deviation distance, where the initial solution is one of the M solutions, and M is an integer greater than 0;

[0065] It is understandable that with the in-depth promotion of car screens in domestic and foreign markets, in the face of the personalized needs of different customers, the customized demand for voice functions continues to increase, and the requirements for customization of different models are also very different. For example, as shown in Table 1:

[0066] Table 1

[0067]

[0068] As shown in Table 1, different models have different requirements for voice customization, so each model will have a corresponding demand solution. For example, if the model is Geely Lynk & Co, then as shown in Table 1, the demand solution includes the need for a 09 series exclusive image, support for custom wake-up words / images, four-zone voice recognition, etc., then this embodiment can pre-set M solutions according to different model prices. The M solutions here are all developed based on voice middleware. Voice middleware is a general voice software, such as Figure 3 As shown, the main functions include voice wake-up, voice recognition, semantic understanding, etc. For different demand solutions, it is only necessary to expand on the original voice middleware without re-customizing it according to the demand solution. This not only saves costs, but also avoids problems such as long development time and difficulty in subsequent project maintenance. It can be understood that the solution is pre-set by technical personnel in this field and stored in the database. For example, the solution corresponding to the small ant model in Table 1 is used as the solution.

[0069] The method for calculating the deviation distance between the demand solution and the M pre-built solutions includes:

[0070] The Euclidean distance between each first element in the demand solution and each second element in the solution is calculated, and the calculated Euclidean distances are weighted summed to obtain the deviation distance corresponding to each demand solution, where the first element corresponds to the second element.

[0071] In this embodiment, the first element represents the feature items in the demand solution, such as the demand for voice customization of vehicle models, including "exclusive image", "custom support for wake-up words / images", "sound zone coverage", etc. Similarly, the second element represents the feature items in the solution. The weight corresponding to each feature item can be pre-set based on expert experience. For example, "sound zone coverage" is more important than "wake-up words", so the cost of changing the "sound zone coverage" is higher and the development time required is also longer.

[0072] Methods for determining an initial solution from M solutions based on the deviation distance include:

[0073] The deviation distances corresponding to the M solutions are compared with the preset distance threshold, and the solutions corresponding to those greater than the preset distance threshold are eliminated. The remaining solutions are clustered around the required solution, and the solution closest to the required solution is determined as the initial solution.

[0074] It should be added that, in this embodiment, after eliminating the solutions with large deviations, the focus is on the better solutions that meet the requirements, which can reduce the subsequent development adjustment and modification workload and reduce development costs. Clustering can then effectively narrow the range of candidate solutions, thereby reducing the computational burden. Clustering analysis further optimizes the selection process and makes the determination of the final solution more efficient. The clustering mentioned above can be K-means clustering or DBSCAN clustering, which will not be elaborated in this embodiment.

[0075] S20: Compare the demand solution with the initial solution to generate a solution comparison result, determine N target solutions based on the solution comparison result and the initial solution, and calculate the development time and development cost corresponding to the N target solutions, where N is an integer greater than 0;

[0076] In one embodiment of the present invention, the solution comparison result represents the degree of difference between the demand solution and the initial solution in various feature items. These differences reflect the parts of the initial solution that need to be adjusted or improved to meet the requirements of the demand solution. Figure 4 As shown, the method for generating scheme comparison results includes:

[0077] Each first element in the demand solution is reconstructed according to preset rules, and the Pearson correlation coefficient and information entropy are calculated based on each processed first element in the demand solution and the corresponding second element in the initial solution. The Pearson correlation coefficient and information entropy are fused to generate element comparison results, and the set of all element comparison results is used as the solution comparison result.

[0078] It should be added that the preset rules mentioned above refer to the arrangement rules of text-related descriptions. For example, taking voice image as an example, the text sent by the customer is "09 series exclusive image", then the reconstructed text is "The @voice image@ of the demand plan should be the 09 series exclusive style", and the corresponding text description of the voice image in the initial plan is "The #voice image# of the initial plan is the ordinary QQ style". It can be understood that the purpose of such reconstruction is to reconstruct the text descriptions in the demand plan and the initial plan into a standardized structure (such as "The @voice image@ of the demand plan should be..."), which can ensure that the description form of the feature items is consistent, so that the subsequently calculated Pearson correlation coefficient and information entropy are more accurate and comparable. This structured description avoids the matching error caused by non-standardized text and improves the reliability of the difference measurement.

[0079] Methods for calculating the Pearson correlation coefficient include:

[0080] ;

[0081] In the formula, R is the Pearson correlation coefficient, is the total number of first elements in the demand scenario, For the demand plan The eigenvector value corresponding to the first element, For demand solutions The average value of the eigenvectors corresponding to the first elements, The initial plan The eigenvector value corresponding to the second element, For the initial plan The average value of the eigenvectors corresponding to the second elements.

[0082] In this embodiment, It is both the total number of the first elements in the demand solution and the total number of the second elements in the initial solution. In the above, both the first element and the second element are text data. Then the first element and the second element are converted into feature vectors. The feature vector value can be the value after encoding the feature vector. The range of R is between [-1,1]. If R is closer to 1, it means that the feature items of the demand solution and the initial solution are more similar. On the contrary, if R is closer to -1, it means that the difference between the feature items of the demand solution and the initial solution is greater.

[0083] In the above content, information entropy refers to the difference and uncertainty between the demand plan and the initial plan in each feature item. The larger the information entropy, the greater the uncertainty of the feature item, and the more significant the difference between the two, which means that there is a big difference between the demand plan and the initial plan in this feature item. Conversely, the smaller the information entropy, the relatively close or consistent the demand plan and the initial plan in this feature item.

[0084] Methods for calculating information entropy include:

[0085] ;

[0086] In the formula, SE is information entropy, For the demand plan The weight corresponding to the first element, So The logarithmic function with base , is a natural constant.

[0087] In the above, the fusion processing of the Pearson correlation coefficient and the information entropy may be a weighted summation or the Pearson correlation coefficient and the information entropy may be converted into vector form and then the vector inner product may be performed, which is not limited in this embodiment.

[0088] The initial solution selected in this embodiment has been screened by the deviation distance to ensure that it has a high similarity in the demand solution. When entering step S20, through detailed feature item comparison (such as calculating the Pearson correlation coefficient and information entropy), it is possible to more accurately identify the differences between the demand solution and the initial solution in specific feature items, avoid errors caused by large deviations, further improve the accuracy of the matching degree, and help provide targeted solutions in personalized environments with different vehicle models and needs. This sequential combination ensures the targeted nature of the solution selection and can more flexibly adapt to changes in customer demand on different feature items.

[0089] Methods for determining N target solutions based on solution comparison results and initial solutions include:

[0090] The solution comparison results and the initial solution are input into the pre-built solution generation model to obtain N target solutions output by the solution generation model.

[0091] The construction method of the solution generation model includes:

[0092] Acquire a sample data set, wherein the sample data set includes a historical scheme comparison result, a historical initial scheme, and a historical target scheme;

[0093] Divide the sample data set into a sample training set and a sample test set, and build a regression network;

[0094] The historical scheme comparison results and historical initial schemes in the sample training set are used as the input data of the regression network, and the historical target schemes in the sample training set are used as the output data of the regression network. The regression network is trained to obtain an initial regression network for predicting the target scheme.

[0095] The initial regression network is tested using a sample test set, and the output of the initial regression network that is less than the preset error value is used as the solution generation model. The initial regression network is a deep neural network model.

[0096] It is not difficult to understand that the logic of generating N target solutions is that the number of items that differ between the initial solution and the required solution is determined to be k according to the solution comparison results, and then N for generating the target solution is 2 k-1, for example, following the content in Table 1 above, the demand solution includes the need for a 09 series exclusive image, support for custom wake-up words / images, four-tone zone voice recognition, etc., and the initial solution includes QQ, support for custom wake-up words, but not support for customized images, single-tone zone recognition, etc., then the number of items that differ between the demand solution and the initial solution is 6, and the N for generating the target solution is 63. It can be understood that the reason why the N for generating the target solution is 2 k -1 is based on the principles of combinatorial mathematics and the flexibility requirements of feature item combinations. For the k difference items between the initial solution and the required solution, each difference item can be chosen to be "changed" or "not changed", so each difference item has two options. The case where all difference items do not change is the initial solution itself, so this unique case needs to be removed. This can cover all possible difference combinations between the required solution and the initial solution, thereby providing multiple options for subsequent optimization.

[0097] It should be added that, to calculate the development time and development cost corresponding to N target solutions, the development sub-time and development sub-cost corresponding to each difference item between the target solution and the initial solution can be calculated respectively, and then the development time and development cost corresponding to the target solution are obtained by summing them up. Taking Table 1 as an example, the voice image in the initial solution is QQ, and the voice image in the target solution is the exclusive image of the 09 series. Therefore, it takes time and cost to change the voice image from QQ to the exclusive image of the 09 series, and due to the different difference items, the corresponding time and cost are different. For example, the cost of changing the coverage of the sound zone is greater than the cost of changing the voice image.

[0098] S30: Determine an optimal solution from N target solutions according to development time, development cost and intelligent optimization algorithm, where the optimal solution is one of the N target solutions.

[0099] The intelligent optimization algorithm can be a sparrow search algorithm, and the methods for obtaining the optimal solution include:

[0100] S301: Randomly generate a group of sparrow individuals, each sparrow individual represents a target solution, and randomly generate an initial position for each sparrow individual , that is, the target solution, so that it satisfies:

[0101] ;

[0102] ;

[0103] in, is the development cost corresponding to the target solution, To preset the development cost threshold, is the development time corresponding to the target solution, The preset development time threshold.

[0104] S302: Construct the fitness function value of each sparrow individual :

[0105] ;

[0106] In the formula, , All are weight factors;

[0107] It should be noted that the target solution generated by this embodiment should be optimized in terms of development time and development cost. Development time and development cost are two key constraints in the project. By calculating the sum of development cost and development time as the fitness, the comprehensive performance of each solution in these two aspects can be measured. The smaller the fitness value, the more the solution meets the requirements in terms of development time and cost, and the closer it is to the optimal solution. It should be added that the development time and development cost are first evaluated by technical personnel in this field and then set after communicating with customers. The purpose of this embodiment is to complete product development at a faster speed at a limited cost.

[0108] S303: Perform exploration operations on some sparrow individuals to update their target initial ratio ;

[0109] ;

[0110] In the formula, For the A sparrow in the The position of the generation, For the A sparrow in the The position of the generation, is the maximum number of iterations, is a random angle, It is an exponential function with e as base, where e is a natural constant.

[0111] Among them, through exploration operations, some sparrow individuals are moved randomly to better cover the solution space, find potential optimal solutions, increase the diversity of solutions, prevent sparrow individuals from falling into local optimality, and provide more solution paths for the group.

[0112] S304: Check the updated Is it less than If not, perform boundary processing and check the updated Is it less than , if not, perform boundary processing;

[0113] This embodiment ensures that all sparrow individuals still meet the constraints of development cost and development time after updating, performs boundary processing on solutions that do not meet the conditions, and ensures that all solutions are within the range of feasible solutions to ensure the effectiveness of subsequent optimization operations.

[0114] S305: Perform a follow-up operation on the remaining sparrow individuals to update their target initial ratio ;

[0115] The above-mentioned following operation means that the remaining sparrow individuals will follow the direction of the current optimal solution for update. By continuously following the optimal solution, the follower individuals have the opportunity to gradually approach the optimal solution, thereby improving the overall fitness of the population.

[0116] S306: Calculate the fitness function value of each sparrow individual after the update , update the current optimal solution;

[0117] After updating the target initial ratio of the individual sparrows, the fitness of each individual sparrow is recalculated to determine its pros and cons under the new ratio scheme. Through this operation, the algorithm can evaluate the effect of each update and select the optimal target scheme.

[0118] S307: Repeat the above steps S303 to S306 until the preset maximum number of iterations is reached. Stop searching when , and output the target solution corresponding to the current optimal solution, which is the optimal solution.

[0119] Check for updated Is it less than If not, the boundary processing methods include:

[0120] = ;

[0121] In the formula, It refers to the development cost of the i-th sparrow individual after the t+1 generation update.

[0122] It should be added that the above steps are and This ensures that the target solution that exceeds the preset development cost threshold is forced to be corrected to the feasible range, ensuring that all solutions in the optimization process can be actually applied and avoiding infeasible solutions.

[0123] Similarly, check the updated Is it less than If not, the boundary processing methods include:

[0124] = ;

[0125] In the formula, It refers to the development time of the i-th sparrow individual after the t+1 generation update.

[0126] Update its target initial ratio The methods include:

[0127] ;

[0128] In the formula, is a random number, is the location of the current optimal solution.

[0129] In this embodiment, the above steps indicate that the individual sparrows follow and update based on the current optimal solution. In this way, the individual sparrows will gradually move closer to the optimal solution and converge to a better solution.

[0130] This embodiment preliminarily screens a set of feasible solutions that meet the requirements through the solution comparison results of step S20 and the generation of N target solutions, ensuring that all target solutions have a high degree of functional match with the required solutions. The subsequent step S30 further optimizes under the constraints of development cost and development time to ensure that the optimal solution finally selected not only meets the functional requirements, but also achieves the best in cost and time, and finally achieves the goal of quickly completing development at a limited cost. It is understandable that after the optimal solution is generated, communication will be conducted with the customer. When the customer is not satisfied, the optimal solution will be removed and the optimal solution will be regenerated until the customer is satisfied.

[0131] In this embodiment, the deviation distance between the demand solution and M pre-built solutions is first calculated, and an initial solution is determined from the M solutions based on the deviation distance. The demand solution and the initial solution are compared to generate a solution comparison result, and N target solutions are determined based on the solution comparison result and the initial solution. The optimal solution is determined from the N target solutions based on the development time, development cost and intelligent optimization algorithm. This embodiment generates target solutions by screening and difference comparison, and selects the optimal solution through an intelligent optimization algorithm, which effectively reduces the need for deep customization of complex projects. In this way, it is possible to flexibly meet the diverse needs of users within a limited development cycle and cost, simplify the development process and improve efficiency.

[0132] Example 2

[0133] See also Figure 2 As shown, based on the same inventive concept, this embodiment discloses a vehicle model expansion evaluation system based on voice middleware. For details not provided in this embodiment, please refer to the description of the relevant parts in Embodiment 1. The system includes:

[0134] Screening module: used to obtain the demand solution corresponding to the model to be expanded, calculate the deviation distance between the demand solution and M pre-built solutions, and determine the initial solution from the M solutions according to the deviation distance. The initial solution is one of the M solutions, where M is an integer greater than 0.

[0135] The method for calculating the deviation distance between the demand solution and the M pre-built solutions includes:

[0136] The Euclidean distance between each first element in the demand solution and each second element in the solution is calculated, and the calculated Euclidean distances are weighted summed to obtain the deviation distance corresponding to each demand solution, where the first element corresponds to the second element.

[0137] Methods for determining an initial solution from M solutions based on the deviation distance include:

[0138] The deviation distances corresponding to the M solutions are compared with the preset distance threshold, and the solutions corresponding to those greater than the preset distance threshold are eliminated. The remaining solutions are clustered around the required solution, and the solution closest to the required solution is determined as the initial solution.

[0139] Solution generation module: used to compare the demand solution with the initial solution, generate solution comparison results, determine N target solutions based on the solution comparison results and the initial solution, and calculate the development time and development cost corresponding to the N target solutions, where N is an integer greater than 0;

[0140] In one embodiment of the present invention, the solution comparison result represents the degree of difference between the demand solution and the initial solution in various feature items. These differences reflect the parts of the initial solution that need to be adjusted or improved to meet the requirements of the demand solution. Then, the method for generating the solution comparison result includes:

[0141] Each first element in the demand solution is reconstructed according to preset rules, and the Pearson correlation coefficient and information entropy are calculated based on each processed first element in the demand solution and the corresponding second element in the initial solution. The Pearson correlation coefficient and information entropy are fused to generate element comparison results, and the set of all element comparison results is used as the solution comparison result.

[0142] Methods for calculating the Pearson correlation coefficient include:

[0143] ;

[0144] In the formula, R is the Pearson correlation coefficient, is the total number of first elements in the demand scenario, For the demand plan The eigenvector value corresponding to the first element, For demand solutions The average value of the eigenvectors corresponding to the first elements, The initial plan The eigenvector value corresponding to the second element, For the initial plan The average value of the eigenvectors corresponding to the second elements.

[0145] Methods for calculating information entropy include:

[0146] ;

[0147] In the formula, SE is information entropy, For the demand plan The weight corresponding to the first element, So The logarithmic function with base , is a natural constant.

[0148] Methods for determining N target solutions based on solution comparison results and initial solutions include:

[0149] The solution comparison results and the initial solution are input into the pre-built solution generation model to obtain N target solutions output by the solution generation model.

[0150] The construction method of the solution generation model includes:

[0151] Acquire a sample data set, wherein the sample data set includes a historical scheme comparison result, a historical initial scheme, and a historical target scheme;

[0152] Divide the sample data set into a sample training set and a sample test set, and build a regression network;

[0153] The historical scheme comparison results and historical initial schemes in the sample training set are used as the input data of the regression network, and the historical target schemes in the sample training set are used as the output data of the regression network. The regression network is trained to obtain an initial regression network for predicting the target scheme.

[0154] The initial regression network is tested using a sample test set, and the output of the initial regression network that is less than the preset error value is used as the solution generation model. The initial regression network is a deep neural network model.

[0155] Solution optimization module: used to determine the best solution from N target solutions based on development time, development cost and intelligent optimization algorithm. The best solution is one of the N target solutions.

[0156] Example 3

[0157] This embodiment provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, a vehicle model expansion evaluation method based on voice middleware provided by the above methods is implemented.

[0158] Since the electronic device introduced in this embodiment is an electronic device used to implement a vehicle model expansion evaluation method based on voice middleware in the embodiment of the present application, based on the vehicle model expansion evaluation method based on voice middleware introduced in the embodiment of the present application, technical personnel in this field can understand the specific implementation mode of the electronic device of this embodiment and its various variations, so how the electronic device implements the method in the embodiment of the present application is no longer introduced in detail. As long as technical personnel in this field implement the electronic device used in the vehicle model expansion evaluation method based on voice middleware in the embodiment of the present application, it belongs to the scope of protection of this application.

[0159] Example 4

[0160] The present embodiment discloses a computer-readable storage medium, on which a computer program is stored. When the computer program is executed, the above-mentioned vehicle model expansion evaluation method based on voice middleware is implemented.

[0161] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters, weights and thresholds in the formula are set by technicians in this field according to actual conditions.

[0162] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present invention is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from one website, computer, server or data center to another website, computer, server or data center via a wired network or a wireless network. The computer-readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server or data center that includes one or more available media sets. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD) or a semiconductor medium. The semiconductor medium may be a solid-state hard disk.

[0163] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed in the present invention can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0164] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0165] In the several embodiments provided by the present invention, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of the units is only one, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0166] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0167] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0168] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

[0169] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A vehicle model expansion evaluation method based on voice middleware, characterized in that: include: Obtain the demand solution corresponding to the model to be expanded, calculate the deviation distance between the demand solution and M pre-built solutions, and determine the initial solution from the M solutions according to the deviation distance. The initial solution is one of the M solutions, where M is an integer greater than 0. Compare the demand solution with the initial solution to generate a solution comparison result, determine N target solutions based on the solution comparison result and the initial solution, and calculate the development time and development cost corresponding to the N target solutions, where N is an integer greater than 0; Determine the best solution from N target solutions based on development time, development cost and intelligent optimization algorithm. The best solution is one of the N target solutions. The method for generating a scheme comparison result comprises: Reconstruct each first element in the demand solution according to preset rules, calculate the Pearson correlation coefficient and information entropy based on each processed first element in the demand solution and the corresponding second element in the initial solution, fuse the Pearson correlation coefficient and information entropy to generate element comparison results, and take the set of all element comparison results as the solution comparison result; The method for determining N target solutions based on the solution comparison result and the initial solution includes: Input the scheme comparison results and the initial scheme into the pre-built scheme generation model to obtain N target schemes output by the scheme generation model; The construction method of the solution generation model includes: Acquire a sample data set, wherein the sample data set includes a historical scheme comparison result, a historical initial scheme, and a historical target scheme; Divide the sample data set into a sample training set and a sample test set, and build a regression network; The historical scheme comparison results and historical initial schemes in the sample training set are used as the input data of the regression network, and the historical target schemes in the sample training set are used as the output data of the regression network. The regression network is trained to obtain an initial regression network for predicting the target scheme. The initial regression network is tested using the sample test set, and the output of the initial regression network with a smaller error value than the preset error value is used as the solution generation model; The intelligent optimization algorithm is a sparrow search algorithm, and the method for obtaining the optimal solution includes: S301: Randomly generate a group of sparrow individuals, each sparrow individual represents a target solution, and randomly generate an initial position for each sparrow individual , that is, the target solution, so that it satisfies: ∈ ; ∈ ; in, is the development cost corresponding to the target solution, To preset the development cost threshold, is the development time corresponding to the target solution, To preset the development time threshold; S302: Construct the fitness function value of each sparrow individual : = ; In the formula, , All are weight factors; S303: Perform exploration operations on some sparrow individuals to update their target initial ratio ; ; In the formula, For the A sparrow in the The position of the generation, For the A sparrow in the The position of the generation, is the maximum number of iterations, is a random angle, It is an exponential function with e as the base, where e is a natural constant; S304: Check the updated Is it less than If not, perform boundary processing and check the updated Is it less than , if not, perform boundary processing; S305: Perform a follow-up operation on the remaining sparrow individuals to update their target initial ratio ; S306: Calculate the fitness function value of each sparrow individual after the update , update the current optimal solution; S307: Repeat S303 to S306 until the preset maximum number of iterations is reached. Stop searching when , and output the target solution corresponding to the current optimal solution, which is the optimal solution.

2. According to the method for vehicle model expansion evaluation based on voice middleware in claim 1, it is characterized in that: The method for calculating the deviation distance between the demand solution and the M pre-built solutions includes: The Euclidean distance between each first element in the demand solution and each second element in the solution is calculated, and the calculated Euclidean distances are weighted summed to obtain the deviation distance corresponding to each demand solution, where the first element corresponds to the second element.

3. A vehicle model expansion evaluation method based on voice middleware according to claim 2, characterized in that: The method for determining an initial solution from M solutions according to the deviation distance includes: The deviation distances corresponding to the M solutions are compared with the preset distance threshold, and the solutions corresponding to those greater than the preset distance threshold are eliminated. The remaining solutions are clustered around the required solution, and the solution closest to the required solution is determined as the initial solution.

4. The vehicle model expansion evaluation method based on voice middleware according to claim 3 is characterized in that: The method for calculating the Pearson correlation coefficient includes: R= ; In the formula, R is the Pearson correlation coefficient, is the total number of first elements in the demand scenario, For the demand plan The eigenvector value corresponding to the first element, For demand solutions The average value of the eigenvectors corresponding to the first elements, The initial plan The eigenvector value corresponding to the second element, For the initial plan The average value of the eigenvectors corresponding to the second elements.

5. The vehicle model expansion evaluation method based on voice middleware according to claim 4 is characterized in that: The method for calculating the information entropy includes: SE= ; In the formula, SE is information entropy, For the demand plan The weight corresponding to the first element, So The logarithmic function with base , is a natural constant.

6. The vehicle model expansion evaluation method based on voice middleware according to claim 1 is characterized in that: The check updated Is it less than If not, the boundary processing methods include: = ; In the formula, It refers to the development cost of the i-th sparrow individual after the t+1 generation update.

7. A vehicle model expansion evaluation system based on voice middleware, which is used to implement a vehicle model expansion evaluation method based on voice middleware as described in any one of claims 1 to 6, characterized in that: include: Screening module: used to obtain the demand solution corresponding to the model to be expanded, calculate the deviation distance between the demand solution and M pre-built solutions, and determine the initial solution from the M solutions according to the deviation distance. The initial solution is one of the M solutions, where M is an integer greater than 0. Solution generation module: used to compare the demand solution with the initial solution, generate solution comparison results, determine N target solutions based on the solution comparison results and the initial solution, and calculate the development time and development cost corresponding to the N target solutions, where N is an integer greater than 0; Solution optimization module: used to determine the best solution from N target solutions based on development time, development cost and intelligent optimization algorithm. The best solution is one of the N target solutions. The method for generating a scheme comparison result comprises: Reconstruct each first element in the demand solution according to preset rules, calculate the Pearson correlation coefficient and information entropy based on each processed first element in the demand solution and the corresponding second element in the initial solution, fuse the Pearson correlation coefficient and information entropy to generate element comparison results, and take the set of all element comparison results as the solution comparison result; The method for determining N target solutions based on the solution comparison result and the initial solution includes: Input the scheme comparison results and the initial scheme into the pre-built scheme generation model to obtain N target schemes output by the scheme generation model; The construction method of the solution generation model includes: Acquire a sample data set, wherein the sample data set includes a historical scheme comparison result, a historical initial scheme, and a historical target scheme; Divide the sample data set into a sample training set and a sample test set, and build a regression network; The historical scheme comparison results and historical initial schemes in the sample training set are used as the input data of the regression network, and the historical target schemes in the sample training set are used as the output data of the regression network. The regression network is trained to obtain an initial regression network for predicting the target scheme. The initial regression network is tested using the sample test set, and the output of the initial regression network with a smaller error value than the preset error value is used as the solution generation model; The intelligent optimization algorithm is a sparrow search algorithm, and the method for obtaining the optimal solution includes: S301: Randomly generate a group of sparrow individuals, each sparrow individual represents a target solution, and randomly generate an initial position for each sparrow individual , that is, the target solution, so that it satisfies: ∈ ; ∈ ; in, is the development cost corresponding to the target solution, To preset the development cost threshold, is the development time corresponding to the target solution, To preset the development time threshold; S302: Construct the fitness function value of each sparrow individual : = ; In the formula, , All are weight factors; S303: Perform exploration operations on some sparrow individuals to update their target initial ratio ; ; In the formula, For the A sparrow in the The position of the generation, For the A sparrow in the The position of the generation, is the maximum number of iterations, is a random angle, It is an exponential function with e as base, where e is a natural constant; S304: Check the updated Is it less than If not, perform boundary processing and check the updated Is it less than , if not, perform boundary processing; S305: Perform a follow-up operation on the remaining sparrow individuals to update their target initial ratio ; S306: Calculate the fitness function value of each sparrow individual after the update , update the current optimal solution; S307: Repeat S303 to S306 until the preset maximum number of iterations is reached. Stop searching when , and output the target solution corresponding to the current optimal solution, which is the optimal solution.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the vehicle model expansion evaluation method based on voice middleware as described in any one of claims 1 to 6 is implemented.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed, the method for expanding and evaluating vehicle models based on voice middleware according to any one of claims 1 to 6 is implemented.

Images