A vehicle-mounted speech recognition simulation test method, system and storage medium

Abstract

The present application relates to the field of speech recognition technology, and in particular to a vehicle-mounted speech recognition simulation test method, system and storage medium, wherein the method comprises: setting simulation parameters, and judging whether a control instruction is detected; if a control instruction is detected, generating a recognition failure mark; if a control instruction is not detected, setting a test voice; based on the test voice, obtaining a target instruction; if the test voice is not detected, generating a recognition failure mark; if the test voice is detected, generating a control instruction; judging whether the control instruction matches the target instruction; if the target instruction matches the control instruction, generating a recognition success mark based on the simulation parameters and the test voice; if the target instruction does not match the control instruction, generating a recognition failure mark based on the simulation parameters and the test voice; generating a simulation test report based on the recognition failure mark and the recognition success mark. The present application helps to improve the accuracy of vehicle-mounted speech recognition results.

Description

Technical Field

[0001] The present application relates to the field of speech recognition technology, and in particular to a vehicle-mounted speech recognition simulation test method, system and storage medium. Background Art

[0002] With the development of automobile intelligence, the in-vehicle voice recognition system has become a key part of automobile human-computer interaction. It allows the driver to control various functions of the vehicle, such as navigation, multimedia playback, temperature adjustment, etc., through voice commands without taking their hands off the steering wheel or their eyes off the road, greatly improving the convenience and safety of driving. In order to ensure that the voice commands spoken by the user can be correctly interpreted by the in-vehicle voice recognition device (such as the in-vehicle smart control screen), it is necessary to test whether the in-vehicle control screen can correctly perform voice recognition.

[0003] Traditional in-vehicle speech recognition testing methods mainly rely on actual road testing, which requires driving the vehicle in various real environments, such as highways, city streets, mountainous areas, etc. This is not only costly, but also poses certain safety risks. In order to solve these problems of real-scene testing, simulation testing technology came into being.

[0004] The in-vehicle speech simulation test method in the related art simply simulates various external scene conditions by establishing a virtual test environment. However, it has major defects in speech recognition in some complex semantic environments, resulting in low accuracy of in-vehicle speech recognition results. Summary of the invention

[0005] In order to help improve the accuracy of in-vehicle speech recognition results, the present application provides an in-vehicle speech recognition simulation test method, system and storage medium.

[0006] In the first aspect, the present application provides a vehicle-mounted speech recognition simulation test method, which adopts the following technical solution:

[0007] A vehicle-mounted speech recognition simulation test method, comprising:

[0008] Set simulation parameters and determine whether control instructions are detected;

[0009] If the control instruction is detected, generating a recognition failure flag based on the simulation parameters;

[0010] If the control command is not detected, setting a test voice;

[0011] Based on the test voice, obtaining a target instruction;

[0012] Determining whether the test voice is detected;

[0013] If the test speech is not detected, generating a recognition failure mark based on the simulation parameters and the test speech;

[0014] If the test voice is detected, the test voice is recognized and a control instruction is generated;

[0015] Determining whether the control instruction matches the target instruction;

[0016] If the target instruction matches the control instruction, generating a recognition success mark based on the simulation parameters and the test speech;

[0017] If the target instruction does not match the control instruction, generating a recognition failure mark based on the simulation parameters and the test speech;

[0018] A simulation test report is generated based on the recognition failure mark and the recognition success mark.

[0019] By adopting the above technical solution, after setting the simulation parameters, it is determined whether a control instruction is detected. If a control instruction is detected, it indicates that the control instruction is detected before the test voice is set, which is very likely to be a misrecognition, so a recognition failure mark is generated; if the control instruction is not detected, a test voice is further set, and a target instruction is obtained based on the detection voice to determine whether the test voice is detected. If the detection voice is not detected, it indicates that the test voice is set but cannot be detected, which is very likely to be a missed recognition, so a recognition failure mark is generated; if the detection voice is detected, a control instruction is generated after the detection voice is recognized, and it is determined whether the control instruction matches the target instruction. If they match, it indicates that the recognition of the detection voice is successful, so a recognition success mark is generated; if they do not match, it indicates that the recognition of the detection voice fails, so a recognition failure mark is generated; finally, a simulation test report is generated according to all recognition failure marks and recognition success marks;

[0020] Combined with simulation parameters, multiple judgment operations are set to determine whether speech recognition fails or succeeds in different situations. Especially in the process of recognizing the detected speech, combined with the speech environment, the possibility of inaccurate recognition results caused by major defects in speech recognition in some complex semantic environments in related technologies is effectively reduced, thereby helping to improve the accuracy of in-vehicle speech recognition results.

[0021] Optionally, if the test voice is detected, recognizing the test voice and generating a control instruction includes:

[0022] If the test speech is detected, the test speech is analyzed and a semantic environment corresponding to the test speech is obtained;

[0023] Based on the test speech and the semantic environment, obtaining keywords;

[0024] Based on the keywords and preset mapping rules, control instructions are obtained.

[0025] Optionally, the acquiring keywords based on the test speech and the semantic environment includes:

[0026] Segmenting the test speech and obtaining single sentences;

[0027] Determining whether a target object exists in the single sentence;

[0028] If the target object exists in the single sentence, obtaining a target verb corresponding to the target object;

[0029] Acquire keywords based on the target object and the target verb;

[0030] If the target object does not exist in the single sentence, obtaining a referent object based on the test speech and the semantic environment;

[0031] Obtaining a referential verb corresponding to the referent object;

[0032] Based on the reference object and the reference verb, a keyword is acquired.

[0033] Optionally, acquiring keywords based on the target object and the target verb includes:

[0034] Determining whether the target object contains multiple sub-objects;

[0035] If the target object does not include multiple sub-objects, acquiring the keyword based on the target object and the target verb;

[0036] If the target object contains multiple sub-objects, obtaining the target sub-object;

[0037] The keyword is obtained based on the target sub-object and the target verb.

[0038] Optionally, if the target object includes multiple sub-objects, obtaining the target sub-object includes:

[0039] If the target object includes multiple sub-objects, obtaining a test distance corresponding to the test speech;

[0040] Based on the test distance, obtaining a simulated sound position corresponding to the test speech;

[0041] Determining whether the simulated sound emission position is the driving position;

[0042] If the simulated sound emission position is the driving position, the multiple sub-objects are all used as target sub-objects;

[0043] If the simulated sound emission position is not the driving position, the target sub-object is acquired based on the simulated sound emission position.

[0044] Optionally, the acquiring the control instruction based on the keyword and the preset mapping rule includes:

[0045] Get the number of keywords;

[0046] Determine whether the number of keywords is greater than 1;

[0047] If the number of the keywords is less than or equal to 1, acquiring the control instruction based on the keywords and the preset mapping rules;

[0048] If the number of the keywords is greater than 1, determining whether there is a specified correlation between different keywords;

[0049] If the specified correlation does not exist between different keywords, obtaining the acquisition time nodes corresponding to the different keywords and generating a sorting list based on a preset sorting rule;

[0050] Based on the sorted list, the keywords and the preset mapping rules, obtaining a control instruction;

[0051] If the specified association exists between different keywords, obtaining the target keyword based on the sorting list and the preset selection rule;

[0052] Based on the target keyword and the preset mapping rule, the control instruction is acquired.

[0053] Optionally, if the number of the keywords is greater than 1, determining whether there is a specified correlation between different keywords includes:

[0054] If the number of the keywords is greater than 1, obtaining the target objects corresponding to the keywords based on the semantic environment;

[0055] Determining whether the target objects corresponding to different keywords are the same;

[0056] If the target objects corresponding to different keywords do not exist to be the same, it is determined that there is no designated association between the different keywords;

[0057] If the target objects corresponding to different keywords are the same, the keywords corresponding to the same target objects are marked as the same type of keywords;

[0058] Based on the semantic environment, determining whether the control instructions corresponding to the similar keywords are contradictory;

[0059] If the control instructions corresponding to the keywords of the same type are in conflict with each other, it is determined that there is a specified correlation between the different keywords.

[0060] In the second aspect, the present application also discloses a vehicle-mounted speech recognition simulation test system, which adopts the following technical solution:

[0061] A vehicle-mounted speech recognition simulation test system, comprising:

[0062] The parameter setting module is used to set the simulation parameters and determine whether the control instruction is detected;

[0063] a first generating module, wherein if the control instruction is detected, the first generating module is used to generate a recognition failure mark based on the simulation parameters;

[0064] a voice setting module, wherein if the control instruction is not detected, the voice setting module is used to set the test voice;

[0065] An instruction acquisition module, used for acquiring a target instruction based on the test voice;

[0066] A first judging module, used for judging whether the test voice is detected;

[0067] A second generating module, if the test speech is not detected, the second generating module is used to generate a recognition failure mark based on the simulation parameters and the test speech;

[0068] A third generating module, if the test voice is detected, the third generating module is used to recognize the test voice and generate a control instruction;

[0069] A second judgment module, used to judge whether the control instruction matches the target instruction;

[0070] A fourth generation module, if the target instruction matches the control instruction, the fourth generation module is used to generate a recognition success mark based on the simulation parameters and the test speech;

[0071] a fifth generating module, wherein if the target instruction does not match the control instruction, the fifth generating module is used to generate a recognition failure mark based on the simulation parameters and the test speech;

[0072] The sixth generating module is used to generate a simulation test report based on the recognition failure mark and the recognition success mark.

[0073] By adopting the above technical solution, after setting the simulation parameters, it is determined whether a control instruction is detected. If a control instruction is detected, it indicates that the control instruction is detected before the test voice is set, which is very likely to be a misrecognition, so a recognition failure mark is generated; if the control instruction is not detected, a test voice is further set, and a target instruction is obtained based on the detection voice to determine whether the test voice is detected. If the detection voice is not detected, it indicates that the test voice is set but cannot be detected, which is very likely to be a missed recognition, so a recognition failure mark is generated; if the detection voice is detected, a control instruction is generated after the detection voice is recognized, and it is determined whether the control instruction matches the target instruction. If they match, it indicates that the recognition of the detection voice is successful, so a recognition success mark is generated; if they do not match, it indicates that the recognition of the detection voice fails, so a recognition failure mark is generated; finally, a simulation test report is generated according to all recognition failure marks and recognition success marks;

[0074] Combined with simulation parameters, multiple judgment operations are set to determine whether speech recognition fails or succeeds in different situations. Especially in the process of recognizing the detected speech, combined with the speech environment, the possibility of inaccurate recognition results caused by major defects in speech recognition in some complex semantic environments in related technologies is effectively reduced, thereby helping to improve the accuracy of in-vehicle speech recognition results.

[0075] In a third aspect, the present application provides a computer-readable storage medium, which adopts the following technical solution:

[0076] A computer-readable storage medium stores a computer program, and when the computer program is loaded by a processor, the method of the first aspect is executed.

[0077] By adopting the above technical solution, a computer program is generated based on the method of the first aspect and stored in a computer-readable storage medium so as to be loaded and executed by a processor. The computer-readable storage medium facilitates the readability and storage of the computer program.

[0078] In summary, this application includes the following beneficial technical effects:

[0079] Combined with simulation parameters, multiple judgment operations are set to determine whether speech recognition fails or succeeds in different situations. Especially in the process of recognizing the detected speech, combined with the speech environment, the possibility of inaccurate recognition results caused by major defects in speech recognition in some complex semantic environments in related technologies is effectively reduced, thereby helping to improve the accuracy of in-vehicle speech recognition results. BRIEF DESCRIPTION OF THE DRAWINGS

[0080] Figure 1This is a main flow chart of a vehicle-mounted speech recognition simulation test method according to an embodiment of the present application;

[0081] Figure 2 is a flowchart of steps S201 to S203;

[0082] Figure 3 is a flowchart of steps S301 to S307;

[0083] Figure 4 is a flowchart of steps S401 to S404;

[0084] Figure 5 is a flowchart of steps S501 to S505;

[0085] Figure 6 is a flowchart of steps S601 to S608;

[0086] Figure 7 is a flowchart of steps S701 to S706;

[0087] Figure 8 It is a module diagram of a vehicle-mounted speech recognition simulation test system according to an embodiment of the present application.

[0088] Description of reference numerals:

[0089] 1. Parameter setting module; 2. First generation module; 3. Voice setting module; 4. Instruction acquisition module; 5. First judgment module; 6. Second generation module; 7. Third generation module; 8. Second judgment module; 9. Fourth generation module; 10. Fifth generation module; 11. Sixth generation module. DETAILED DESCRIPTION

[0090] In a first aspect, the present application discloses a vehicle-mounted speech recognition simulation test method.

[0091] Reference Figure 1 , a vehicle-mounted speech recognition simulation test method, comprising steps S101 to S111:

[0092] Step S101: setting simulation parameters and determining whether a control instruction is detected.

[0093] Specifically, in this embodiment, the simulation parameters include noise parameters, vehicle parameters, road parameters, etc., wherein the noise parameters include wind noise, human voice and other noise in the car, etc.; the vehicle parameters include vehicle information and vehicle speed information, etc.; the road parameters include road properties, and the road properties include bumpy sections and smooth sections; the control instructions are instructions for controlling the vehicle to perform related operations, such as turning on the lights, turning on the air conditioner, and opening the sunroof, etc.

[0094] Step S102: If a control instruction is detected, a recognition failure flag is generated based on the simulation parameters.

[0095] Specifically, in this embodiment, if a control instruction is detected, it means that the control instruction is detected without setting the detection voice. Then it is very likely that misrecognition occurs, and other noises are identified as detection voices, and then the corresponding control instruction is generated based on the identified detection voice. Therefore, based on the simulation parameters at the current moment, a corresponding recognition failure mark is generated.

[0096] Step S103: If no control command is detected, a test voice is set.

[0097] Specifically, the test voice is the voice used for performing simulation experiments.

[0098] Step S104: Acquire target instructions based on the test voice.

[0099] Specifically, the target instruction is the control instruction corresponding to the test voice, and is also the control instruction that needs to be tested this time; in this embodiment, the test voice corresponds to the target instruction, so after determining the test voice, the corresponding target instruction can be known, and the target instruction can be multiple instructions.

[0100] Step S105: Determine whether a test voice is detected.

[0101] Step S106: If the test speech is not detected, a recognition failure mark is generated based on the simulation parameters and the test speech.

[0102] Specifically, in this embodiment, if the test voice is not detected, it means that the test voice has been set, but the voice recognition device has not detected it. Therefore, it is very likely that the voice recognition device has missed recognition. Therefore, based on the simulation parameters at the current moment, a corresponding recognition failure mark is generated.

[0103] Step S107: If the test voice is detected, the test voice is recognized based on the semantic environment and a control instruction is generated.

[0104] Specifically, in this embodiment, if a test voice is detected, the test voice is recognized and analyzed according to the voice environment, and then a control instruction is generated.

[0105] Step S108: Determine whether the control instruction matches the target instruction.

[0106] Specifically, in this embodiment, judging whether the control instruction matches the target instruction is judging whether the control instruction is consistent with the target instruction. If they are consistent, they match.

[0107] Step S109: If the target instruction matches the control instruction, a recognition success mark is generated based on the simulation parameters and the test speech.

[0108] Specifically, in this embodiment, if the target instruction matches the control instruction, it means that the control instruction obtained by the speech recognition device based on the test speech is consistent with the target instruction that the test speech wants to detect. Therefore, based on the simulation parameters at the current moment, a corresponding recognition success mark is generated.

[0109] Step S110: If the target instruction does not match the control instruction, a recognition failure mark is generated based on the simulation parameters and the test speech.

[0110] Specifically, in this embodiment, if the target instruction does not match the control instruction, it means that the control instruction obtained by the speech recognition device based on the test speech is inconsistent with the target instruction that the test speech wants to detect. Therefore, based on the simulation parameters at the current moment, a corresponding recognition failure mark is generated.

[0111] Step S111: Generate a simulation test report based on the recognition failure mark and the recognition success mark.

[0112] Specifically, the simulation detection report includes all recognition failure marks and the corresponding simulation parameters and detection voices, and also includes all recognition success marks and the corresponding simulation parameters and detection voices.

[0113] The vehicle-mounted speech recognition simulation test method provided in this embodiment, after setting the simulation parameters, determines whether a control instruction is detected. If a control instruction is detected, it indicates that the control instruction is detected before the test voice is set, which is very likely to be a misrecognition, so a recognition failure mark is generated; if the control instruction is not detected, a test voice is further set, and a target instruction is obtained based on the detection voice to determine whether the test voice is detected. If the detection voice is not detected, it indicates that the test voice is set but cannot be detected, which is very likely to be a missed recognition, so a recognition failure mark is generated; if the detection voice is detected, a control instruction is generated after recognizing the detection voice, and it is determined whether the control instruction matches the target instruction. If they match, it indicates that the recognition of the detection voice is successful, so a recognition success mark is generated; if they do not match, it indicates that the recognition of the detection voice fails, so a recognition failure mark is generated; finally, a simulation test report is generated based on all recognition failure marks and recognition success marks.

[0114] Combined with simulation parameters, multiple judgment operations are set to determine whether speech recognition fails or succeeds in different situations. Especially in the process of recognizing the detected speech, combined with the speech environment, the possibility of inaccurate recognition results caused by major defects in speech recognition in some complex semantic environments in related technologies is effectively reduced, thereby helping to improve the accuracy of in-vehicle speech recognition results.

[0115] Reference Figure 2 In one implementation of this embodiment, if the test voice is detected in step S107, the test voice is recognized based on the semantic environment, and the control instruction is generated, including steps S201 to S203:

[0116] Step S201: If a test speech is detected, the test speech is analyzed and the semantic environment corresponding to the test speech is obtained.

[0117] Specifically, in this embodiment, the semantic environment refers to the environment in which a language unit (such as a word, sentence, chapter, etc.) is located during use, which includes the contextual context inside the language and the situational context outside the language; the contextual context refers to the front and back parts of the language unit in the text. For example, Zhang San: "Should the air conditioner continue to be turned on?" Li Si: "No need". If the contextual context is not considered, it is difficult to obtain the correct control instruction after recognizing "no need" alone, and the control instruction obtained after combining the contextual context is: turn off the air conditioner. The situational context involves external factors when the language is used, including time, place, participants, topics, etc. For example, Zhang San: "I'm sweating, but I have a cold, so it's better to blow natural wind." According to the situational context, it can be analyzed that Zhang San is sweating, which means that he is currently hot, but he has a cold, so he can't blow air conditioning, but can only blow natural wind, so the control instruction can be analyzed to be: open the car window, turn off the air conditioner.

[0118] Step S202: Acquire keywords based on the test speech and semantic environment.

[0119] Specifically, in this embodiment, the keywords are important words used to represent the core content or theme of each test speech.

[0120] Step S203: Acquire control instructions based on keywords and preset mapping rules.

[0121] Specifically, in this embodiment, the preset mapping rule is a preset rule for obtaining corresponding control instructions according to keywords, that is, the correspondence between keywords and control instructions.

[0122] The in-vehicle speech recognition simulation test method provided in this embodiment analyzes the test speech and obtains the semantic environment corresponding to the test speech. Keywords are obtained based on the test speech and the semantic environment, which helps to improve the accuracy of keyword acquisition. Finally, control instructions are obtained based on the keywords and preset mapping rules, which helps to improve the accuracy of control instructions.

[0123] Reference Figure 3 In one implementation of this embodiment, step S202 obtains keywords based on the test speech and semantic environment, including steps S301 to S307:

[0124] Step S301: segment the test speech and obtain single sentences.

[0125] Step S302: Determine whether there is a target object in the single sentence.

[0126] Specifically, the target object is an object in the vehicle that can perform related operations after generating control instructions through voice recognition. It can be a real object or a virtual object, such as a car window, air conditioner, music player, and navigation.

[0127] Step S303: If there is a target object in the single sentence, a target verb corresponding to the target object is obtained.

[0128] Specifically, in this embodiment, the target verb is the verb corresponding to the target object. For example, in the single sentence "I'm so bored, let's turn on the music player and listen to some songs", the target object is "music player" and the verb corresponding to it is "open", so the target verb corresponding to the target object "music player" is "open".

[0129] Step S304: Acquire keywords based on the target object and the target verb.

[0130] Specifically, in this embodiment, the keyword may be a combination of a target object and a target verb. For example, if the target verb corresponding to the target object “music player” is “open”, the keyword may be “open music player”.

[0131] Step S305: If the target object does not exist in the single sentence, the referent object is obtained based on the test speech and semantic environment.

[0132] Specifically, in this embodiment, the reference object is a target object generated according to the voice environment when there is no target object in a single sentence; for example, in the single sentence "I am sweating, but I have a little cold, so it is better to enjoy the natural wind", there is no target object, but analysis of the single sentence shows that the user wants to open the car window, so the car window is used as the reference object.

[0133] Step S306: Obtain the referential verb corresponding to the referent object.

[0134] Specifically, in this embodiment, the referential verb is a verb corresponding to the referential object.

[0135] Step S307: Acquire keywords based on the referent object and the referent verb.

[0136] Specifically, in this embodiment, the keyword may be a combination of a reference object and a reference verb.

[0137] The in-vehicle speech recognition simulation test method provided in this embodiment determines whether there is a target object in a single sentence. If so, keywords are obtained based on the target object and the target verb; if not, keywords are obtained based on the referent object and the referent verb; according to different situations, corresponding methods are selected to obtain keywords, which helps to improve the accuracy of keyword acquisition, thereby helping to improve the accuracy of speech recognition results.

[0138] Reference Figure 4 In one implementation of this embodiment, step S304 obtains keywords based on the target object and the target verb, including steps S401 to S404:

[0139] Step S401: Determine whether the target object contains multiple sub-objects.

[0140] Specifically, in this embodiment, the sub-object is a subordinate object of the target object, for example, the car windows include the left front window, left rear window, right front window and right rear window, wherein the car window is the target object, and the left front window, left rear window, right front window and right rear window are sub-objects.

[0141] Step S402: If the target object does not contain multiple sub-objects, obtain keywords based on the target object and the target verb.

[0142] Specifically, in this embodiment, if the target object does not contain multiple sub-objects, it indicates that the target object is unique, so the keyword is directly obtained based on the referent object and the referent verb.

[0143] Step S403: If the target object contains multiple sub-objects, the target sub-objects are obtained.

[0144] Specifically, in this embodiment, the target sub-object is the sub-object that is actually desired to be controlled in the test speech, for example, opening the left front window, wherein the left front window is the target sub-object.

[0145] Step S404: Acquire keywords based on the target child object and the target verb.

[0146] Specifically, in this embodiment, the keyword may be a combination of a target sub-object and a target verb.

[0147] The in-vehicle speech recognition simulation test method provided in this embodiment determines whether the target object contains multiple sub-objects. If not, the keywords are directly obtained based on the referent object and the referent verb; if not, the target sub-object is obtained and the keywords are obtained based on the target sub-object and the target verb; according to different situations, the corresponding method is selected to obtain keywords, which helps to improve the accuracy of keyword acquisition, thereby helping to improve the accuracy of speech recognition results.

[0148] Reference Figure 5 In one implementation of this embodiment, if the target object includes multiple sub-objects in step S403, obtaining the target sub-objects includes steps S501 to S505:

[0149] Step S501: If the target object includes multiple sub-objects, a test distance corresponding to the test speech is obtained.

[0150] Specifically, the test distance is the distance from the location where the test voice is emitted to the voice recognition device. In this embodiment, the voice recognition device is set at the steering wheel position, and the test distance can be measured by echo ranging method.

[0151] Step S502: Based on the test distance, obtain the simulated sound position corresponding to the test speech.

[0152] Specifically, the simulated sound position is the position where the test voice is emitted. By comparing the test distance with the distance from the steering wheel to different positions, the simulated position can be known. The simulated position includes the driver's seat, the co-driver's seat, the left rear position and the right rear position. In this embodiment, the distance from the steering wheel to each position is pre-set.

[0153] Step S503: Determine whether the simulated sound emission position is the driving position.

[0154] Step S504: if the simulated sound generation position is the driving position, then multiple sub-objects are all used as target sub-objects.

[0155] Specifically, in this embodiment, if the simulated sound position is the driving position, all sub-objects are generally taken into consideration when controlling the vehicle. Therefore, if no target sub-object is specified, all sub-objects are regarded as target sub-objects; for example, if the test voice emitted from the driving position is to open the left front window, the corresponding target instruction is to open the left front window, but if the test voice emitted from the driving position is to open the window, the corresponding target instruction is to open all windows.

[0156] Step S505: If the simulated sound emission position is not the driving position, a target sub-object is acquired based on the simulated sound emission position.

[0157] Specifically, if the simulated sound position is not the driving seat, the corresponding target sub-object is obtained according to the simulated sound position. For example, the test voice emitted by the passenger seat is to open the window, and the corresponding target instruction is to open the right front window. In this embodiment, the driving seat is on the left front side as an example.

[0158] The in-vehicle speech recognition simulation test method provided in this embodiment, if the target object contains multiple sub-objects, obtains the test distance corresponding to the test speech, and obtains the simulated sound position based on the target distance, and then determines whether the simulated sound position is the driving position. If so, all of the multiple sub-objects are used as target sub-objects. If not, the target sub-object is obtained according to the simulated sound position; according to the simulation location, the corresponding rule is selected to obtain the corresponding target sub-object, which not only helps to improve the accuracy of speech recognition, but also helps to improve the user experience.

[0159] Reference Figure 6 In one implementation of this embodiment, step S203 obtains the control instruction based on the keyword and the preset mapping rule, including steps S601 to S608:

[0160] Step S601: Obtain the number of keywords.

[0161] Specifically, in this embodiment, the number of keywords refers to the number of keywords.

[0162] Step S602: Determine whether the number of keywords is greater than 1.

[0163] Step S603: If the number of keywords is less than or equal to 1, a control instruction is obtained based on the keywords and preset mapping rules.

[0164] Specifically, in this embodiment, if the number of keywords is less than 1, it means that the keyword is not recognized, or there is no word related to the control instruction in the single sentence.

[0165] Step S604: If the number of keywords is greater than 1, determine whether there is a specified correlation between different keywords.

[0166] Specifically, if the number of keywords is greater than 1, it indicates that there are multiple keywords. In order to reduce the situation where the vehicle repeatedly executes opposite control instructions due to the antagonistic relationship between different keywords, it is further determined whether there is a specified association between different keywords. In this embodiment, the specified association means that there is an antagonistic relationship. For example, if there is an antagonistic relationship between the control instructions to turn on navigation and turn off navigation, then there is a specified association between the control instructions to turn on navigation and turn off navigation.

[0167] Step S605: If there is no specified correlation between different keywords, the acquisition time nodes corresponding to the different keywords are acquired and a sorting list is generated based on a preset sorting rule.

[0168] Specifically, the acquired time node is the time node at which the keyword is acquired; the preset sorting rule is the preset sorting rule, and in this embodiment, the keywords are arranged in order from the earliest to the latest; and the sorting list is a list of keywords sorted according to the preset sorting rule.

[0169] Step S606: Acquire control instructions based on the sorting list, keywords and preset mapping rules.

[0170] Specifically, in this embodiment, when there is no specified association between different keywords, multiple control instructions will be generated according to different keywords, sorting lists and preset mapping rules.

[0171] Step S607: If there is a specified correlation between different keywords, the target keyword is obtained based on the sorting list and the preset selection rules.

[0172] Specifically, if there is a specified correlation between different keywords, the target keyword will be obtained according to the preset selection rules and the sorting list. The preset selection rules are pre-set rules for selecting the target keywords. In this embodiment, the last keyword in the sorting list is selected as the target keyword, and the target keyword is the keyword that generates the final control instruction.

[0173] Step S608: Acquire control instructions based on the target keyword and preset mapping rules.

[0174] The vehicle-mounted speech recognition simulation test method provided in this embodiment determines whether the number of keywords is greater than 1. If the number of keywords is less than or equal to 1, control instructions are obtained based on the keywords and preset mapping rules; if the number of keywords is greater than 1, it is further determined whether there is a specified association between different keywords. If not, control instructions are obtained based on the sorting list, keywords and preset mapping rules; if so, control instructions are obtained based on the target keywords and preset mapping rules; according to different situations corresponding to the number of keywords, the corresponding method is selected to obtain control instructions, which helps to reduce the repeated execution of opposite control instructions by the vehicle due to the adversarial relationship between different keywords, thereby helping to improve the accuracy of the control instructions.

[0175] Reference Figure 7 In one implementation of this embodiment, if the number of keywords is greater than 1, step S604 determines whether there is a specified correlation between different keywords, including steps S701 to S706:

[0176] Step S701: If the number of keywords is greater than 1, the target object corresponding to the keyword is obtained based on the semantic environment.

[0177] Step S702: Determine whether there are target objects corresponding to different keywords that are the same.

[0178] Step S703: If there are no identical target objects corresponding to different keywords, it is determined that there is no designated association between the different keywords.

[0179] Specifically, in this embodiment, if there are no identical target objects corresponding to different keywords, that is, the target objects corresponding to different keywords are different, then there is no designated association between different keywords.

[0180] Step S704: If there are different keywords corresponding to the same target object, the keywords corresponding to the same target object are marked as the same type of keywords.

[0181] Specifically, in this embodiment, the same type of keywords are keywords corresponding to the same target object.

[0182] Step S705: Based on the semantic environment, determine whether the control instructions corresponding to the same type of keywords are contradictory.

[0183] Specifically, in this embodiment, if there are opposite instructions immediately, for example, if there are opposite instructions for the control instructions to turn on navigation and to turn off navigation, then the control instructions corresponding to the same type of keywords are opposite.

[0184] Step S706: If the control instructions corresponding to the same type of keywords are in conflict with each other, it is determined that there is a specified correlation between different keywords.

[0185] The in-vehicle speech recognition simulation test method provided in this embodiment obtains the target object corresponding to the keyword, and determines whether the target objects corresponding to different keywords are the same. If not, it is determined that there is no specified association between different keywords; if so, it is further determined whether the control instructions corresponding to the same type of keywords are opposite. If the relationship is opposite, it is determined that there is a specified association between different keywords; through multiple judgments, based on the judgment results, it is determined whether there is a specified association between different keywords, which helps to improve the accuracy of the judgment results.

[0186] In a second aspect, the present application also discloses a vehicle-mounted speech recognition simulation test system.

[0187] Reference Figure 8 , a vehicle-mounted speech recognition simulation test system, comprising:

[0188] The parameter setting module is used to set the simulation parameters and determine whether the control instruction is detected;

[0189] A first generating module, if the control instruction is detected, the first generating module is used to generate a recognition failure mark based on the simulation parameters;

[0190] A voice setting module, which is used to set the test voice if no control command is detected;

[0191] An instruction acquisition module, used for acquiring target instructions based on test speech;

[0192] A first judgment module, used to judge whether a test voice is detected;

[0193] A second generation module, if the test speech is not detected, the second generation module is used to generate a recognition failure mark based on the simulation parameters and the test speech;

[0194] A third generating module, if the test voice is detected, the third generating module is used to recognize the test voice and generate a control instruction;

[0195] A second judgment module is used to judge whether the control instruction matches the target instruction;

[0196] A fourth generation module, if the target instruction matches the control instruction, the fourth generation module is used to generate a recognition success mark based on the simulation parameters and the test speech;

[0197] A fifth generation module, if the target instruction does not match the control instruction, the fifth generation module is used to generate a recognition failure mark based on the simulation parameters and the test speech;

[0198] The sixth generating module is used to generate a simulation test report based on the recognition failure mark and the recognition success mark.

[0199] In a third aspect, an embodiment of the present application discloses a computer-readable storage medium, and a computer program is stored in the computer-readable storage medium, wherein when the computer program is loaded by a processor, an in-vehicle speech recognition simulation test method of the above embodiment is executed.

[0200] The above are all preferred embodiments of the present application, and the protection scope of the present application is not limited thereto. Therefore, any equivalent changes made according to the structure, shape, and principle of the present application should be included in the protection scope of the present application.

Claims

1. A vehicle-mounted speech recognition simulation test method, characterized in that: include: Set simulation parameters and determine whether control instructions are detected; If the control instruction is detected, generating a recognition failure flag based on the simulation parameters; If the control command is not detected, setting a test voice; Based on the test voice, obtaining a target instruction; Determining whether the test voice is detected; If the test speech is not detected, generating a recognition failure mark based on the simulation parameters and the test speech; If the test voice is detected, the test voice is recognized based on the semantic environment and a control instruction is generated; Determining whether the control instruction matches the target instruction; If the target instruction matches the control instruction, generating a recognition success mark based on the simulation parameters and the test speech; If the target instruction does not match the control instruction, generating a recognition failure mark based on the simulation parameters and the test speech; A simulation test report is generated based on the recognition failure mark and the recognition success mark.

2. The vehicle-mounted speech recognition simulation test method according to claim 1, characterized in that: If the test voice is detected, the test voice is recognized based on the semantic environment, and a control instruction is generated, including: If the test speech is detected, the test speech is analyzed and a semantic environment corresponding to the test speech is obtained; Based on the test speech and the semantic environment, obtaining keywords; Based on the keywords and preset mapping rules, control instructions are obtained.

3. The vehicle-mounted speech recognition simulation test method according to claim 2, characterized in that: The acquiring of keywords based on the test speech and the semantic environment comprises: Segmenting the test speech and obtaining single sentences; Determining whether a target object exists in the single sentence; If the target object exists in the single sentence, obtaining a target verb corresponding to the target object; Acquire keywords based on the target object and the target verb; If the target object does not exist in the single sentence, obtaining a referent object based on the test speech and the semantic environment; Obtaining a referential verb corresponding to the referent object; Based on the reference object and the reference verb, a keyword is acquired.

4. The vehicle-mounted speech recognition simulation test method according to claim 3 is characterized in that: The acquiring of keywords based on the target object and the target verb includes: Determining whether the target object contains multiple sub-objects; If the target object does not include multiple sub-objects, acquiring the keyword based on the target object and the target verb; If the target object contains multiple sub-objects, obtaining the target sub-object; The keyword is obtained based on the target sub-object and the target verb.

5. The vehicle-mounted speech recognition simulation test method according to claim 4, characterized in that: If the target object includes multiple sub-objects, obtaining the target sub-objects includes: If the target object includes multiple sub-objects, obtaining a test distance corresponding to the test speech; Based on the test distance, obtaining a simulated sound position corresponding to the test speech; Determining whether the simulated sound emission position is the driving position; If the simulated sound emission position is the driving position, the multiple sub-objects are all used as target sub-objects; If the simulated sound emission position is not the driving position, the target sub-object is acquired based on the simulated sound emission position.

6. The vehicle-mounted speech recognition simulation test method according to claim 2, characterized in that: The acquiring of the control instruction based on the keyword and the preset mapping rule includes: Get the number of keywords; Determine whether the number of keywords is greater than 1; If the number of the keywords is less than or equal to 1, acquiring the control instruction based on the keywords and the preset mapping rules; If the number of the keywords is greater than 1, determining whether there is a specified correlation between different keywords; If the specified correlation does not exist between different keywords, obtaining the acquisition time nodes corresponding to the different keywords and generating a sorting list based on a preset sorting rule; Based on the sorted list, the keywords and the preset mapping rules, obtaining a control instruction; If the specified association exists between different keywords, obtaining the target keyword based on the sorting list and the preset selection rule; Based on the target keyword and the preset mapping rule, the control instruction is acquired.

7. The vehicle-mounted speech recognition simulation test method according to claim 6, characterized in that: If the number of the keywords is greater than 1, determining whether there is a specified correlation between different keywords includes: If the number of the keywords is greater than 1, obtaining the target objects corresponding to the keywords based on the semantic environment; Determining whether the target objects corresponding to different keywords are the same; If the target objects corresponding to different keywords do not exist to be the same, it is determined that there is no designated association between the different keywords; If the target objects corresponding to different keywords are the same, the keywords corresponding to the same target objects are marked as the same type of keywords; Based on the semantic environment, determining whether the control instructions corresponding to the similar keywords are contradictory; If the control instructions corresponding to the keywords of the same type are in conflict with each other, it is determined that there is a specified correlation between the different keywords.

8. A vehicle-mounted speech recognition simulation test system, characterized in that: include: The parameter setting module is used to set the simulation parameters and determine whether the control instruction is detected; a first generating module, wherein if the control instruction is detected, the first generating module is used to generate a recognition failure mark based on the simulation parameters; a voice setting module, wherein if the control instruction is not detected, the voice setting module is used to set the test voice; An instruction acquisition module, used for acquiring a target instruction based on the test voice; A first judging module, used for judging whether the test voice is detected; A second generating module, if the test speech is not detected, the second generating module is used to generate a recognition failure mark based on the simulation parameters and the test speech; A third generating module, if the test voice is detected, the third generating module is used to recognize the test voice and generate a control instruction; A second judgment module, used to judge whether the control instruction matches the target instruction; A fourth generation module, if the target instruction matches the control instruction, the fourth generation module is used to generate a recognition success mark based on the simulation parameters and the test speech; a fifth generating module, wherein if the target instruction does not match the control instruction, the fifth generating module is used to generate a recognition failure mark based on the simulation parameters and the test speech; The sixth generating module is used to generate a simulation test report based on the recognition failure mark and the recognition success mark.

9. A computer-readable storage medium having a computer program stored therein, characterized in that: When the computer program is loaded by a processor, the method according to any one of claims 1 to 7 is executed.

Images