Vehicle-mounted voice test method and device, electronic equipment and storage medium

Abstract

The invention provides a vehicle-mounted voice test method and device, electronic equipment and a storage medium, and the method comprises the steps: setting a corresponding grammar rule based on the control of voice on a vehicle function, and extracting an original text corpus from the grammar rule based on a preset regular expression; inputting the original text corpora into an artificial intelligence generation content AIGC model to generalize the original text corpora, and outputting a plurality of test text corpora similar to the original text corpora in terms of semantics; converting the test text corpus into a corresponding test voice corpus, and playing the test voice corpus in a cabin of the vehicle; after the test voice corpus is played, a test result of vehicle-mounted voice is determined according to a state response and a voice interaction response of the vehicle, the state response represents a vehicle machine control response after the vehicle receives the test voice corpus, and the voice interaction response represents a vehicle machine control response after the vehicle receives the test voice corpus. The voice interaction response represents the interaction response of the vehicle machine to the test voice corpus. According to the invention, the problem that the test corpus of the vehicle-mounted voice is single in the prior art is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of vehicle-mounted voice testing, and in particular to a vehicle-mounted voice testing method and device, electronic equipment and storage medium. Background Art

[0002] With the rapid progress and development of smart cockpit technology for new energy vehicles, voice interaction has become an important way of human-computer interaction, and the degree of intelligence of voice interaction is also an important indicator for measuring the level of intelligence of car cockpits. In actual use, everyone has different speaking habits and ways, and the voice commands issued are also different. The voice commands received by the vehicle voice system are also different, resulting in incorrect parsing of the user's voice commands and thus executing incorrect operations. The test corpus of the existing vehicle voice test is formed by analyzing the user's daily interaction system after limiting the test scope and extracting common sentence patterns. The test corpus formed by this method is limited and single, and cannot cover the voice commands issued by each user for the same function.

[0003] There is currently no effective solution to the above problems in the prior art. Summary of the invention

[0004] One of the purposes of the present invention is to provide a method and device for testing in-vehicle speech, an electronic device and a storage medium to solve the problem of single test corpus for in-vehicle speech in the prior art.

[0005] In a first aspect, the present application provides a method for testing in-vehicle voice, comprising: setting corresponding grammar rules based on voice control of vehicle computer functions, and extracting original text corpus from the grammar rules based on preset regular expressions; inputting the original text corpus into an artificial intelligence generated content AIGC model to generalize the original text corpus, and outputting multiple test text corpora with similar semantics to the original text corpus; converting the test text corpus into corresponding test voice corpus, and playing the test voice corpus in a vehicle cabin; after playing the test voice corpus, determining the test result of the in-vehicle voice according to the state response and voice interaction response of the vehicle, wherein the state response represents the vehicle computer control response of the vehicle after receiving the test voice corpus, and the voice interaction response represents the vehicle computer's interactive response to the test voice corpus.

[0006] Furthermore, the corresponding grammar rules are set based on the control of vehicle computer functions by voice, including: determining the vehicle control scenario and system application scenario of the in-vehicle voice interaction; wherein the vehicle control scenario refers to the control scenario of the vehicle physical equipment by the in-vehicle voice, and the system application scenario refers to the control scenario of the vehicle system software by the in-vehicle voice; the grammar rules are set according to the vehicle control scenario, the system application scenario and the user's language habits.

[0007] Furthermore, the original text corpus is input into an artificial intelligence generated content (AIGC) model to generalize the original text corpus, and multiple test text corpora that are semantically similar to the original text corpus are output, including: transforming and expanding the original text corpus based on the AIGC model to generate new speech text data; fusing the new speech text data with the original text corpus, and processing the fused data through data enhancement to obtain multiple test text corpora that are semantically similar to the original text corpus.

[0008] Furthermore, the test text corpus is converted into a corresponding test speech corpus, including: performing natural language processing on the test text corpus to obtain a phoneme sequence and linguistic features corresponding to the test text corpus; using the linguistic features and phoneme sequence to predict the acoustic features of the speech, and inputting the predicted acoustic features into a WaveNet model to predict the audio data of the next time step; wherein the test speech corpus includes audio data corresponding to multiple time steps predicted by the WaveNet model.

[0009] Furthermore, the test voice corpus is played in the cockpit of the vehicle, including: recording noise samples of different scenes, and mixing the noise samples in different scenes according to a preset ratio to obtain a target noise sample; fixing a bionic artificial mouth in a seat in the cockpit, and playing the target noise sample through an audio device while the bionic artificial mouth plays the test voice corpus; collecting the decibels of the test voice corpus and the target noise sample, and adjusting the volume of different sound source playback devices in real time to simulate the playback volume of the test voice corpus and the target noise sample in different scenes; and playing the test voice corpus and the target noise sample based on the playback volume.

[0010] Furthermore, after playing the test voice corpus, the test result of the in-vehicle voice is determined according to the status response and voice interaction response of the vehicle, including: after playing the test voice corpus, the status response of each physical component inside and outside the vehicle and the soft switches of each component and the system application on the central control screen of the vehicle computer are recorded to obtain the status response; the voice interaction response of the vehicle computer is collected to obtain the voice interaction response; and the test result of the in-vehicle voice is determined according to the status response, the instructions executed by the vehicle and the audio response result of the voice interaction.

[0011] In a second aspect, the present application provides a vehicle-mounted voice testing device, comprising: a first processing module, used to set corresponding grammar rules based on voice control of vehicle-machine functions, and extract original text corpus from the grammar rules based on preset regular expressions; a second processing module, used to input the original text corpus into an artificial intelligence generated content AIGC model to generalize the original text corpus, and output multiple test text corpora with similar semantics to the original text corpus; a third processing module, used to convert the test text corpus into corresponding test voice corpus, and play the test voice corpus in the cabin of the vehicle; a testing module, used to determine the test result of the vehicle-mounted voice according to the state response and voice interaction response of the vehicle after playing the test voice corpus, wherein the state response represents the vehicle-machine control response of the vehicle after receiving the test voice corpus, and the voice interaction response represents the vehicle-machine interaction response to the test voice corpus.

[0012] Furthermore, the first processing module includes: a determination unit, used to determine the vehicle control scenario and system application scenario of in-vehicle voice interaction; wherein the vehicle control scenario refers to the control scenario of the vehicle physical equipment by the in-vehicle voice, and the system application scenario refers to the control scenario of the vehicle system software by the in-vehicle voice; a setting unit, used to set the grammar rules according to the vehicle control scenario, the system application scenario and the user's language habits.

[0013] In a third aspect, the present application provides an electronic device comprising: at least one communication interface; at least one bus connected to the at least one communication interface; at least one processor connected to the at least one bus; and at least one memory connected to the at least one bus, wherein the processor is configured to execute the in-vehicle voice testing method described in the first aspect of the present application.

[0014] In a fourth aspect, the present application also provides a computer storage medium storing computer executable instructions, wherein the computer executable instructions are used to execute the vehicle-mounted voice testing method described in the first aspect of the present application.

[0015] In the embodiment of the present application, after setting the corresponding grammar rules based on the control of the vehicle function by voice, the original text corpus is extracted from the grammar rules based on the preset regular expression, and the text corpus of the vehicle voice is expanded for the first time. After that, the original text corpus is generalized by the AIGC model, and multiple test text corpora with similar semantics to the original text corpus are output, that is, multiple text corpora with similar semantics can be obtained for the same text corpus, because different users may have different expressions for the output control of the same vehicle voice, but the actual control content is consistent. Therefore, after generalizing the original text corpus by the AIGC model, the needs of different users can be met, and the test scenario of the vehicle voice is enriched. Finally, the voice corpus is tested in the vehicle cabin, and the test result of the vehicle voice is determined according to the vehicle state response and voice interaction response, and further real-time adjustment can be made to ensure the accuracy of the vehicle voice test result. Through the above-mentioned embodiment of the present application, the test corpus of the vehicle voice is enriched, so that the test of different voice commands issued by each user for the same function can be covered, and the control efficiency and experience of the subsequent vehicle voice are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a flow chart of the vehicle voice testing method in the embodiment of the present application;

[0017] Figure 2 This is a schematic diagram of the environment layout of the in-vehicle voice test in an embodiment of the present application;

[0018] Figure 3 This is a schematic diagram of the voice test structure in the specific implementation mode of this application;

[0019] Figure 4 This is a flow chart of the in-vehicle speech testing method based on grammar regularization generalization and AIGC corpus generalization in the specific implementation manner of this application;

[0020] Figure 5 It is a structural schematic diagram of a vehicle-mounted voice test device in an embodiment of the present application;

[0021] Figure 6 It is a schematic diagram of the structure of an electronic device in an embodiment of the present application. DETAILED DESCRIPTION

[0022] The following will describe the embodiments of the present invention with reference to the accompanying drawings and preferred embodiments. Those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be understood that the preferred embodiments are only for illustrating the present invention, not for limiting the scope of protection of the present invention.

[0023] It should be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present invention, and thus the drawings only show components related to the present invention rather than being drawn according to the number, shape and size of components in actual implementation. In actual implementation, the type, quantity and proportion of each component may be changed arbitrarily, and the component layout may also be more complicated.

[0024] In the prior art, the test corpus of in-vehicle speech is formed by analyzing the user's daily interaction system after limiting the test scope and extracting common sentence patterns. The test corpus formed by this method is limited and single, and the test cannot cover the speaking habits of more users. At the same time, the audio corpus is recorded by a selected recording person, and due to the limited number of recording people, it is also impossible to better test the generalization ability of the speech system for different corpora and different sound features.

[0025] Based on this, the present application embodiment provides a method for testing vehicle-mounted voice, such as Figure 1 As shown, the steps of the method include:

[0026] Step 101, setting corresponding grammar rules based on voice control of vehicle computer functions, and extracting original text corpus from the grammar rules based on preset regular expressions;

[0027] In the embodiment of the present application, the control of the vehicle function based on voice may specifically refer to the voice control of the physical equipment of the vehicle, such as voice control of vehicle windows, lights, seats, air conditioners and other vehicle-mounted equipment, or voice control of vehicle system applications, such as voice control of navigation, music, telephone, third-party applications, etc. Therefore, the voice corresponding to the control setting of the vehicle function based on voice may specifically be, "turn on Bluetooth", "turn on high beam", etc.

[0028] In addition, the grammar rules in the embodiments of the present application refer to the rules used to describe the sentence structure in the language, and these rules define the grammatical structure of the language, including aspects such as lexical, syntactic and paragraph structure. In the specific examples of the present application, the corresponding rules can be represented by corresponding symbols, for example, parentheses () must be included when using the grammar to construct the corpus, and brackets [] are used to indicate that the content in the brackets [] can be used or not when using the grammar to construct the corpus, and angle brackets <> are used to indicate a dictionary or word slot, in which the content can be replaced by multiple words. Therefore, based on the grammar rules in the embodiments of the present application, the corresponding speech can be: <open|open|open> (fishing|camping) mode.

[0029] In addition, a regular expression is a text pattern used to match text that conforms to a certain rule in a string. It uses a single string to describe and match a series of strings that conform to a certain syntactic rule, and is usually used for text search and replacement. Therefore, in the embodiment of the present application, all options of the voice corpus in the grammar rule can be found through regular expressions, that is, based on the text corresponding to the voice control of the vehicle function, multiple semantically similar text corpora can be derived to constitute the original text corpus for the vehicle voice test.

[0030] Step 102, inputting the original text corpus into the AIGC model to generalize the original text corpus and output multiple test text corpora that are semantically similar to the original text corpus;

[0031] In a specific example, the original text corpus can be preprocessed first, and the preprocessing may include segmentation, normalization, augmentation and other operations to better adapt to subsequent processing procedures. In a specific example of an embodiment of the present application, the artificial intelligence generated content (AIGC) model can be a GAN (Generative Adversarial Network), a VAE (Variational auto Encoder) and other models.

[0032] Step 103, converting the test text corpus into corresponding test voice corpus, and playing the test voice corpus in the vehicle cabin;

[0033] In an embodiment of the present application, in order to be able to better adapt to the real in-vehicle voice control environment, the test voice corpus can be played through corresponding equipment at various positions where users can sit in the cabin, that is, it can be played at the main driver's seat, the co-driver's seat, the right side of the rear seat, the left side of the back seat, etc., thereby meeting the needs of various users riding in the vehicle and making the test of in-vehicle voice more realistic and effective.

[0034] Step 104, after playing the test voice corpus, determine the test result of the vehicle voice according to the vehicle's state response and voice interaction response, wherein the state response represents the vehicle computer control response after receiving the test voice corpus, and the voice interaction response represents the vehicle computer's interactive response to the test voice corpus.

[0035] In the embodiment of the present application, corresponding audio and video acquisition equipment can be set around the vehicle to collect the vehicle's state response and voice interaction response after the test voice material is played, so as to determine the test result of the vehicle voice. Figure 2 As shown, an audio and video acquisition device can be set up around the vehicle to collect the responses of physical devices and system applications in the vehicle after playing the test voice corpus, so as to test the in-vehicle voice. Only when the played test voice corpus is consistent with the vehicle's state response and / or voice interaction response, it indicates that the in-vehicle voice interaction response is correct.

[0036] It can be known from the above steps 101 to 104 that in the embodiment of the present application, after the corresponding grammar rules are set based on the control of the vehicle function by voice, the original text corpus is extracted from the grammar rules based on the preset regular expression, and the text corpus of the vehicle voice is expanded for the first time. After that, the original text corpus is generalized by the AIGC model, and multiple test text corpora with similar semantics to the original text corpora are output, that is, multiple text corpora with similar semantics can be obtained for the same text corpus, because different users may have different expressions for the output control of the same vehicle voice, but the actual control content is the same. Therefore, after generalizing the original text corpus by the AIGC model, the needs of different users can be met, and the test scenarios of vehicle voice are enriched. Finally, the voice corpus is tested in the vehicle cabin, and the test results of the vehicle voice are determined according to the vehicle status response and voice interaction response, and further real-time adjustments can be made to ensure the accuracy of the vehicle voice test results. Through the above-mentioned embodiments of the present application, the test corpus of in-vehicle voice is enriched, so that the test of different voice commands issued by each user for the same function can be covered, thereby improving the control efficiency and experience of subsequent in-vehicle voice.

[0037] In the embodiment of the present application, the method of setting corresponding grammar rules for controlling the vehicle machine function based on voice involved in the above step 101 may further include:

[0038] Step 11, determining the vehicle control scenario and system application scenario of the vehicle voice interaction; wherein the vehicle control scenario refers to the scenario in which the vehicle voice controls the vehicle physical equipment, and the system application scenario refers to the scenario in which the vehicle voice controls the vehicle system software;

[0039] Step 12: Set grammar rules according to vehicle control scenarios, system application scenarios, and user language habits.

[0040] In this regard, the vehicle control scenario in the embodiment of the present application refers to the control of the vehicle-mounted physical device by voice, and the physical device may include: windows, lights, seats, air conditioners, etc. The system application scenario refers to the control of the system application of the vehicle, and the system application may include: music, phone, navigation, third-party applications, etc. In a specific example, the grammar rules can be formulated according to the scenario and combined with the user's language habits as follows: using small brackets () to indicate that the content in the small brackets () must be included when using the grammar to construct the corpus, using square brackets [] to indicate that the content in the square brackets [] can be used or not when using the grammar to construct the corpus, using angle brackets <> to indicate a dictionary or word slot, the content of which can be replaced by multiple words, but when using the grammar to construct the corpus, the content in an angle bracket <> must be included, and finally using "|" to represent the relationship between the words before and after it is "or", that is, any one of the words before and after can be selected. Through the setting of grammar rules, the same text corpus can obtain multiple options, and then more semantically close sentences can be generated by combining them through regular expressions in the subsequent process, that is, the original text corpus is expanded to meet the language habits of different users.

[0041] In an optional implementation of the embodiment of the present application, the method of inputting the original text corpus into the artificial intelligence generated content AIGC model involved in the above step 102 to generalize the original text corpus and output a plurality of test text corpora with semantics similar to the original text corpus may further include:

[0042] Step 21, transforming and expanding the original text corpus based on the AIGC model to generate new speech text data;

[0043] Step 22, the new speech text data is merged with the original text corpus, and the merged data is processed by data enhancement to obtain multiple test text corpora that are semantically similar to the original text corpus.

[0044] In this regard, in a specific example, new speech text data is generated by the AIGC model, and the generation ability of the model can be used, that is, the original data is transformed and expanded to increase the diversity of the speech text corpus. Then the generated speech text data is fused with the original data, that is, data enhancement methods can be used, including but not limited to back translation, synonym replacement, semi-supervised learning, unsupervised data enhancement, etc., to mix the generated data with the original data for training, so as to improve the generalization ability of the speech recognition system. Finally, the AIGC model is used to generate generalized speech text corpus and increase the sample size of the corpus. It can be seen that the AIGC model can further enrich the in-vehicle corpus to meet the needs of more scenarios. Compared with the in-vehicle speech of a single corpus in the prior art, the in-vehicle corpus can be tested in different scenarios through the embodiments of the present application to meet the needs of different users.

[0045] In an optional implementation of the embodiment of the present application, the method of converting the test text corpus into the corresponding test voice corpus involved in the above step 103 may further include:

[0046] Step 31, performing natural language processing on the test text corpus to obtain the phoneme sequence and linguistic features corresponding to the test text corpus;

[0047] Step 32, using linguistic features and phoneme sequences to predict acoustic features of speech, and inputting the predicted acoustic features into the WaveNet model to predict the audio data of the next time step; wherein the test speech corpus includes audio data corresponding to multiple time steps predicted by the WaveNet model.

[0048] In this regard, in a specific example, natural language processing of the test text corpus can be specifically sentence segmentation, word segmentation, text normalization, part-of-speech tagging and phoneme conversion. The linguistic features and phoneme sequences in the text are used to predict the acoustic characteristics of the speech, such as fundamental frequency (F0) and phoneme duration. The predicted linguistic features are input into the WaveNet model as conditions. WaveNet is an autoregressive model that can predict the audio sample of the next time step based on the input of all previous time steps. In this way, the test speech corpus corresponding to the test text corpus can be obtained, so that the test speech corpus corresponding to the test text corpus can be played during the subsequent in-vehicle voice test.

[0049] In an optional implementation of the embodiment of the present application, the method of playing the test voice corpus in the cabin of the vehicle involved in the above step 104 may further include:

[0050] Step 41, recording noise samples of different scenes, and mixing the noise samples of different scenes according to a preset ratio to obtain a target noise sample;

[0051] Step 42, fix the bionic artificial mouth to the seat in the cockpit, and play the target noise sample through the audio equipment while the bionic artificial mouth plays the test speech corpus;

[0052] Step 43, collecting the decibels of the test speech corpus and the target noise sample, and adjusting the volume of different sound source playback devices in real time to simulate the playback volume of the test speech corpus and the target noise sample in different scenarios;

[0053] Step 44: Play the test speech corpus and the target noise sample based on the playback volume.

[0054] In this regard, in a specific example of an embodiment of the present application, the noise samples recorded in different scenes may refer to the noise in different scenes such as urban streets, highways, underground parking lots, bridges and tunnels, etc. These are all areas that the vehicle may pass through during driving. Therefore, in order to be able to simulate the in-vehicle voice control process more realistically, it is also necessary to simulate the noise in each different scene realistically, so that the in-vehicle voice test process is more realistic. After the test is completed, the user's control voice can still be correctly recognized during the in-vehicle voice control process. In addition, during the test process, it is also necessary to adjust the decibels of the test voice corpus and the target noise sample in real time to better fit the noise and in-vehicle voice decibel volume in the real scene.

[0055] In an optional implementation of the embodiment of the present application, the method of determining the test result of the in-vehicle voice according to the state response and voice interaction response of the vehicle after playing the test voice corpus involved in the above step 104 may further include:

[0056] Step 51, after playing the test voice corpus, the status response of each physical component inside and outside the vehicle and each component soft switch and system application on the vehicle computer central control screen is recorded to obtain the status response;

[0057] Step 52, collecting the voice interaction response of the vehicle computer to obtain the voice interaction response;

[0058] Step 53, determining the vehicle voice test result based on the state response to determine the command executed by the vehicle and the audio response result of the voice interaction.

[0059] For the above steps 51 to 53, it is possible to Figure 2The state response and voice interaction response are collected in the manner described in , and then the instructions executed by the vehicle are determined according to the state response and the audio response results of the voice interaction are determined to determine the test results of the in-vehicle voice. Only when the audio and video results are judged to be in line with expectations, can the current round of human-computer voice interaction response be determined to be correct, that is, the played test corpus is consistent with the vehicle's response. For example, if the test voice corpus is "turn on the radio", the vehicle feedback "the radio is turned on", and then further confirmation is made as to whether the radio is really turned on, and then the test is confirmed to be correct.

[0060] The present application is further explained below in conjunction with the specific implementation of the embodiment of the present application, which provides a vehicle-mounted speech testing method based on grammar regularization generalization and AIGC corpus generalization. Figure 3 , and as Figure 4 As shown, the steps of the method include:

[0061] Step 401, formulate voice grammar rules for the vehicle computer function based on voice control, write regular expressions, and apply regular expressions to extract and generate original corpus;

[0062] Specifically, the main scenarios of vehicle-machine voice interaction are first determined, including vehicle control scenarios (including but not limited to windows, lights, seats, air conditioning, etc.) and system application scenarios (including but not limited to music, telephone, navigation, third-party applications, etc.). Then, according to the scenarios and in combination with user language habits, grammar rules are formulated as follows: Use small brackets () to indicate that the content in the small brackets () must be included when using the grammar to construct the corpus; use square brackets [] to indicate that the content in the square brackets [] can be used or not when using the grammar to construct the corpus; use angle brackets <> to indicate a dictionary or word slot, the content of which can be replaced by multiple words, but the content in an angle bracket <> must be included when using the grammar to construct the corpus; finally, use “|” to represent that the relationship between the words before and after is “or”, that is, either of the words before and after can be selected. Based on the above rules, a grammar for constructing the corpus can be formulated, such as: <open|on|turn on> (fishing|camping) mode, [one-click] all closed [rear] (spare|tail) (box|door); after formulating the grammar, use regular expressions to find all options in the grammar, then separate the options into a list, and then use Python's itertools.product library to generate all possible combinations, and finally insert the combinations into the original grammar to generate sentences. At this point, the original text corpus covering the main scenarios of car-machine voice interaction has been generated.

[0063] Step 402, inputting the generated original corpus into the AIGC model for processing to generalize multiple test corpora with similar semantics;

[0064] Specifically, the speech corpus data generated above is first preprocessed, including segmentation, normalization, augmentation and other operations, so as to better adapt to the subsequent processing flow. Use AIGC models including but not limited to GAN, VAE, etc. for training, and enable the model to generate similar speech text data by learning the characteristics and distribution of speech text data. Then use the trained AIGC model to generate new speech text data. The generation ability of the model can be used to transform and expand the original data to increase the diversity of the speech text corpus, and then merge the generated speech text data with the original data. Data enhancement methods, including but not limited to back translation, synonym replacement, semi-supervised learning, unsupervised data enhancement, etc., are used to mix the generated data with the original data for training, so as to improve the generalization ability of the speech recognition system. Finally, the trained model is used to generate generalized speech text corpus to increase the sample size of the corpus.

[0065] Step 403, using speech synthesis technology to convert the generated text-type test corpus into audio-type test corpus, and send it to the bionic artificial mouth device;

[0066] Further, the generated text corpus data is subjected to a series of natural language processing, including sentence segmentation, word segmentation, text normalization, part-of-speech tagging, and phoneme conversion. This series of processing converts the text into a phoneme sequence and provides relevant linguistic information. Then, the acoustic characteristics of the speech, such as fundamental frequency (F0) and phoneme duration, are predicted using linguistic features in the text (such as phoneme identity, syllable stress, number of syllables in a word, position of the current syllable in a phrase, etc.), as well as additional frame position and phoneme duration features. The predicted linguistic features are input as conditions into the WaveNet model. WaveNet is an autoregressive model that predicts the audio sample of the next time step based on the input of all previous time steps. The WaveNet model uses dilated causal convolutions to increase the receptive field, allowing the model to capture long-term temporal dependencies in audio signals. The model generates audio samples time-step by time step, which are probability distributions of quantized audio values, usually output through a softmax layer. The audio samples generated by WaveNet are then converted into audible speech signals through a vocoder.

[0067] Step 404, fix the bionic artificial mouth on the real car, simulate the use scenario of the real user through the artificial mouth, and input the generated audio test corpus into it to test the corpus generalization ability and intelligence degree of the smart cockpit voice system;

[0068] In this regard, the bionic artificial mouth can be fixed on the main driving seat, the co-pilot seat, the right side of the back row, and the left side of the back row to simulate the real use environment of the main and co-pilot seats and the back row passengers, and then a microphone is used to record noise samples under different road conditions, such as urban streets, highways, underground parking lots, bridge tunnels, etc. At the same time, audio editing software is used to combine the already recorded noise samples and generate new noise samples according to a specific ratio, so as to build a noise database to meet the needs of the real test environment. Then, the voice test corpus generated in step S3 is played through the bionic artificial mouth fixed in the car using a voice interaction test control system. While playing the voice test corpus for voice testing, the speakers fixed near the four windows are controlled to play specific noise audio in the noise database, and the voice test control device can analyze the ambient noise collected by the decibel collection control system installed on the car and the decibel of the simulated human voice, so as to adjust the volume of different sound source playback devices in real time, meet the needs of the test environment simulating different road conditions, and try to cover various test environments.

[0069] Step 405, automatically analyzing the test results, including analyzing the intent of the collected video during the test and comparing the response broadcast audio with the defined response broadcast, to determine whether the test results are correct;

[0070] In this regard, after the test is completed, the test results recorded in the test record analysis system are analyzed. The test record analysis system consists of three parts: a 360-degree HD video acquisition device fixed around and inside the vehicle, an audio acquisition device for voice interaction response broadcasting inside and outside the vehicle, and an audio and video test result analysis device. Figure 3 As shown, the high-definition video acquisition device can record the status response of each physical component inside and outside the car and the soft switches and system applications of each component on the central control screen of the car computer after the voice command is issued. At the same time, the voice interaction response broadcast audio acquisition device inside and outside the car can collect the voice interaction response of the car computer after the voice command is issued. Finally, the collected status response video and the response broadcast audio are sent to the audio and video test result analysis device for test result judgment. The result analysis device uses the 3DCNN model to understand the video intent of the collected video content and judge the specific command operation. At the same time, the collected response audio is compared with the defined audio response to judge whether the audio broadcast after the human-computer voice interaction is correct. Only when the audio and video result judgments are in line with expectations, the current round of human-computer voice interaction response is judged to be correct.

[0071] Corresponding to the above Figure 1 , the present application embodiment provides a vehicle-mounted voice test device, such as Figure 5 As shown, the device comprises:

[0072] The first processing module 502 is used to set corresponding grammar rules for controlling the vehicle computer function based on voice, and extract original text corpus from the grammar rules based on a preset regular expression;

[0073] The second processing module 504 is used to input the original text corpus into the AIGC model to generalize the original text corpus and output multiple test text corpora that are semantically similar to the original text corpus;

[0074] The third processing module 506 is used to convert the test text corpus into corresponding test voice corpus and play the test voice corpus in the vehicle cabin;

[0075] The test module 508 is used to determine the test result of the vehicle voice according to the vehicle's state response and voice interaction response after playing the test voice corpus, wherein the state response represents the vehicle computer control response after receiving the test voice corpus, and the voice interaction response represents the vehicle computer's interactive response to the test voice corpus.

[0076] In an optional implementation manner of an embodiment of the present application, the first processing module in the embodiment of the present application may further include: a determination unit, used to determine the vehicle control scenario and system application scenario of the in-vehicle voice interaction; wherein the vehicle control scenario refers to the control scenario of the vehicle physical equipment by the in-vehicle voice, and the system application scenario refers to the control scenario of the vehicle system software by the in-vehicle voice; a setting unit, used to set grammar rules according to the vehicle control scenario, system application scenario and user language habits.

[0077] In an optional implementation manner of an embodiment of the present application, the second processing module in the embodiment of the present application may further include: a first processing unit, used to transform and expand the original text corpus based on the AIGC model to generate new speech text data; a second processing unit, used to fuse the new speech text data with the original text corpus, and process the fused data through data enhancement to obtain multiple test text corpora that are semantically similar to the original text corpus.

[0078] In an optional implementation of an embodiment of the present application, the third processing module in the embodiment of the present application may further include: a third processing unit, used to perform natural language processing on a test text corpus to obtain a phoneme sequence and linguistic features corresponding to the test text corpus; a fourth processing unit, used to predict the acoustic features of speech using linguistic features and phoneme sequences, and input the predicted acoustic features into a WaveNet model to predict audio data for the next time step; wherein the test speech corpus includes audio data corresponding to multiple time steps predicted by the WaveNet model.

[0079] In an optional implementation manner of the embodiment of the present application, the third processing module in the embodiment of the present application may further include: a fifth processing unit, used to record noise samples of different scenes, and mix the noise samples in different scenes according to a preset ratio to obtain a target noise sample; a sixth processing unit, used to fix the bionic artificial mouth in a seat in the cockpit, and play the target noise sample through an audio device while the bionic artificial mouth plays the test voice corpus; a seventh processing unit, used to collect the decibels of the test voice corpus and the target noise sample, and adjust the volume of different sound source playback devices in real time to simulate the playback volume of the test voice corpus and the target noise sample in different scenes; an eighth processing unit, used to play the test voice corpus and the target noise sample based on the playback volume.

[0080] In an optional implementation manner of the embodiment of the present application, the test module in the embodiment of the present application may further include: a recording unit, which is used to record the status response of various physical components inside and outside the vehicle and the soft switches of various components and system applications on the central control screen of the vehicle computer after playing the test voice corpus to obtain a status response; a collection unit, which is used to collect the voice interaction response of the vehicle computer to obtain a voice interaction response; a testing unit, which is used to determine the test result of the vehicle voice based on the status response to determine the instructions executed by the vehicle and the audio response result of the voice interaction.

[0081] like Figure 6 As shown, an embodiment of the present application provides an electronic device, including a processor 611, a communication interface 612, a memory 613 and a communication bus 614, wherein the processor 611, the communication interface 612, and the memory 613 communicate with each other through the communication bus 614.

[0082] Memory 613, used for storing computer programs;

[0083] In one embodiment of the present application, the processor 611 is used to execute the program stored in the memory 613 to implement the vehicle voice test control method provided by any of the aforementioned method embodiments, and its role is similar and will not be repeated here.

[0084] An embodiment of the present application also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the vehicle-mounted voice testing method provided in any of the aforementioned method embodiments are implemented.

[0085] The device embodiments described above are merely illustrative, wherein 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 modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0086] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a general hardware platform, and of course, by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the relevant technology can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0087] It should be understood that the terms used herein are only for the purpose of describing specific example embodiments and are not intended to be limiting. Unless the context clearly indicates otherwise, the singular forms "one", "an" and "said" as used herein may also be meant to include plural forms. The terms "include", "comprise", "contain", and "have" are inclusive, and therefore specify the existence of stated features, steps, operations, elements and / or parts, but do not exclude the existence or addition of one or more other features, steps, operations, elements, parts, and / or combinations thereof. The method steps, processes, and operations described herein are not interpreted as necessarily requiring them to be performed in the specific order described or illustrated, unless the execution order is clearly indicated. It should also be understood that additional or alternative steps may be used.

[0088] The foregoing is merely a specific embodiment of the present invention, which enables those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A vehicle voice testing method, characterized in that: include: Setting corresponding grammar rules for the control of the vehicle computer function based on voice, and extracting original text corpus from the grammar rules based on a preset regular expression; Inputting the original text corpus into an artificial intelligence generated content (AIGC) model to generalize the original text corpus and output multiple test text corpora that are semantically similar to the original text corpus; Converting the test text corpus into corresponding test voice corpus, and playing the test voice corpus in the cockpit of the vehicle; After playing the test voice corpus, the test result of the in-vehicle voice is determined according to the state response and voice interaction response of the vehicle, wherein the state response represents the vehicle computer control response of the vehicle after receiving the test voice corpus, and the voice interaction response represents the vehicle computer interaction response to the test voice corpus.

2. The method according to claim 1, characterized in that The grammar rules corresponding to the voice-based control settings of vehicle functions include: Determine the vehicle control scenario and system application scenario of vehicle voice interaction; wherein the vehicle control scenario refers to the scenario in which the vehicle voice controls the vehicle physical equipment, and the system application scenario refers to the scenario in which the vehicle voice controls the vehicle system software; The grammar rules are set according to the vehicle control scenario, the system application scenario and the user's language habits.

3. The method according to claim 1, characterized in that Inputting the original text corpus into the AIGC model to generalize the original text corpus, and outputting a plurality of test text corpora with semantic similarity to the original text corpus include: Transforming and expanding the original text corpus based on the AIGC model to generate new speech text data; The new speech text data is fused with the original text corpus, and the fused data is processed by data enhancement to obtain a plurality of test text corpora that are semantically similar to the original text corpus.

4. The method according to claim 1, characterized in that: Converting the test text corpus into corresponding test speech corpus includes: Performing natural language processing on the test text corpus to obtain a phoneme sequence and linguistic features corresponding to the test text corpus; The linguistic features and the phoneme sequence are used to predict the acoustic features of the speech, and the predicted acoustic features are input into the WaveNet model to predict the audio data of the next time step; wherein the test speech corpus includes audio data corresponding to multiple time steps predicted by the WaveNet model.

5. The method according to claim 1, characterized in that Playing the test voice corpus in the vehicle cabin includes: Record noise samples of different scenes, and mix the noise samples of different scenes according to a preset ratio to obtain the target noise sample; Fixing the bionic artificial mouth to the seat in the cockpit, and playing the target noise sample through the audio equipment while the bionic artificial mouth plays the test speech corpus; Collecting the decibels of the test voice corpus and the target noise sample, and adjusting the volume of different sound source playback devices in real time to simulate the playback volume of the test voice corpus and the target noise sample in different scenarios; The test speech corpus and the target noise sample are played based on the playback volume.

6. The method according to claim 1, characterized in that After playing the test voice corpus, determining the test result of the vehicle voice according to the state response and the voice interaction response of the vehicle, including: After playing the test voice corpus, the status responses of the soft switches of the various physical components inside and outside the vehicle and the system applications on the central control screen of the vehicle are recorded to obtain the status responses; Collecting the voice interaction response of the vehicle computer to obtain the voice interaction response; The command executed by the vehicle is determined according to the state response and the audio response result of the voice interaction is determined to determine the test result of the in-vehicle voice.

7. A vehicle-mounted voice testing device, characterized in that: include: A first processing module, configured to set corresponding grammar rules for controlling the vehicle computer functions based on voice, and extract original text corpus from the grammar rules based on a preset regular expression; The second processing module is used to input the original text corpus into the AIGC model to generalize the original text corpus and output a plurality of test text corpora that are semantically similar to the original text corpus; A third processing module, configured to convert the test text corpus into corresponding test voice corpus, and play the test voice corpus in a vehicle cabin; A test module is used to determine the test result of the vehicle voice according to the state response and voice interaction response of the vehicle after playing the test voice corpus, wherein the state response represents the vehicle computer control response of the vehicle after receiving the test voice corpus, and the voice interaction response represents the vehicle computer's interactive response to the test voice corpus.

8. The device according to claim 7, characterized in that The first processing module comprises: A determination unit, used to determine a vehicle control scenario and a system application scenario of the vehicle-mounted voice interaction; wherein the vehicle control scenario refers to a scenario in which the vehicle-mounted voice controls the vehicle's physical equipment, and the system application scenario refers to a scenario in which the vehicle-mounted voice controls the vehicle's system software; A setting unit is used to set the grammar rules according to the vehicle control scenario, the system application scenario and the user's language habits.

9. An electronic device, comprising: at least one communication interface; at least one bus connected to the at least one communication interface; at least one processor coupled to the at least one bus; At least one memory connected to the at least one bus, wherein the processor is configured to execute the vehicle-mounted voice testing method as described in any one of claims 1 to 6.

10. A computer storage medium storing computer executable instructions, wherein the computer executable instructions are used to execute the vehicle-mounted voice testing method according to any one of claims 1 to 6.

Images