Vehicle-mounted voice self-defined wake-up word activation method and device, and vehicle

Abstract

The invention provides a vehicle-mounted voice self-defined wake-up word activation method and device and a vehicle. The method comprises the following steps: acquiring a user-defined wake-up word input by a user; inputting the user-defined wake-up word into a multi-dimensional evaluation model, and performing comprehensive scoring on the user-defined wake-up word by using the multi-dimensional evaluation model to generate a comprehensive score of the user-defined wake-up word; and determining the application state of the custom wake-up word according to the comprehensive score. According to the method provided by the invention, the problem of system failure or frequent false triggering caused by improper setting of the user-defined wake-up word is reduced, the risk of wake-up failure and false wake-up is reduced, and more intelligent and personalized voice interaction experience is provided for the user.

Description

Technical Field

[0001] The present invention relates to the technical field of in-vehicle voice interaction, and particularly to a method, a device, and a vehicle for activating a custom wake-up word for in-vehicle voice. Background Art

[0002] With the continuous progress of intelligent vehicle technology, in-vehicle voice interaction systems have become increasingly popular and have become an important part of modern travel. In terms of the wake-up mechanism, mainstream in-vehicle platforms generally adopt a fixed wake-up word design strategy. With just a simple wake-up word, users can easily activate the voice assistant and then conveniently operate various functions. Although using a fixed wake-up word can ensure basic wake-up stability, since each user has their own preferences and styles, in order to meet the personalized needs of users, the industry allows users to set unique custom wake-up words according to their preferences, thereby enhancing the personalization of the usage experience.

[0003] However, there are many problems with the custom wake-up words of in-vehicle voice in actual applications, such as wake-up failure and false wake-up, which lead to problems such as system failures or frequent false triggers in the in-vehicle voice interaction system, thus reducing the user experience. Summary of the Invention

[0004] The present invention provides a method, a device, and a vehicle for activating a custom wake-up word for in-vehicle voice, so as to solve the defect of poor user experience caused by wake-up failure and false wake-up in the prior art.

[0005] The present invention provides a method for activating a custom wake-up word for in-vehicle voice, including:

[0006] Obtaining a custom wake-up word input by a user;

[0007] Inputting the custom wake-up word into a multi-dimensional evaluation model, and using the multi-dimensional evaluation model to comprehensively evaluate the custom wake-up word to generate a comprehensive score of the custom wake-up word; at least the evaluation items of the multi-dimensional evaluation model include: wake-up word length evaluation and conflict word matching evaluation;

[0008] Determining the application status of the custom wake-up word according to the comprehensive score; the application status at least includes: activating the custom wake-up word.

[0009] According to the method for activating a custom wake-up word for in-vehicle voice provided by the present invention, the comprehensive evaluation of the custom wake-up word by using the multi-dimensional evaluation model includes:

[0010] Check whether the length of the custom wake word is within the set range. If it is not within the set range, deduct points accordingly based on the deviation degree between the wake word length and the set range, and determine whether to perform a conflict word matching evaluation based on the deducted points. If it is within the set range, perform the conflict word matching evaluation.

[0011] The conflict word matching evaluation includes:

[0012] Match and detect the custom wake word with the words in the conflict word library to determine whether the conflict word library includes the custom wake word. If it is included, record the score of the corresponding custom wake word as 0 and prompt the user to modify it.

[0013] The conflict word library includes: a sensitive word library and a restricted word library; the conflict word library is a word library obtained by regularly updating the sensitive word library and the restricted word library.

[0014] According to an activation method for a vehicle-mounted voice custom wake word provided by the present invention, the evaluation items of the multi-dimensional evaluation model further include: speech instruction similarity evaluation.

[0015] The comprehensive scoring of the custom wake word using the multi-dimensional evaluation model includes:

[0016] First, perform a wake word length evaluation and a conflict word matching evaluation on the custom wake word in sequence, and determine whether to perform a speech instruction similarity evaluation based on the evaluation results.

[0017] The speech instruction similarity evaluation includes:

[0018] Determine the similarity between the custom wake word and the instructions in the pre-constructed speech instruction library, and deduct points accordingly based on the similarity. The lower the similarity, the higher the deduction.

[0019] According to an activation method for a vehicle-mounted voice custom wake word provided by the present invention, the evaluation items of the multi-dimensional evaluation model further include: symbol evaluation.

[0020] The comprehensive scoring of the custom wake word using the multi-dimensional evaluation model includes:

[0021] First, perform a wake word length evaluation, a conflict word matching evaluation, and a speech instruction similarity evaluation on the custom wake word in sequence, and determine whether to perform a symbol evaluation based on the evaluation results.

[0022] The symbol evaluation includes:

[0023] Identify whether the custom wake word contains non-alphanumeric characters. If it contains non-alphanumeric characters, deduct points accordingly based on the number of non-alphanumeric characters.

[0024] According to an activation method of an in-vehicle voice custom wake-up word provided by the present invention, the evaluation items of the multi-dimensional evaluation model further include: reduplicated word and phoneme repetition evaluation;

[0025] The comprehensive scoring of the custom wake-up word by using the multi-dimensional evaluation model includes:

[0026] First, perform wake-up word length evaluation, conflicting word matching evaluation, voice command similarity evaluation, and symbol evaluation on the custom wake-up word in sequence, and then perform reduplicated word and phoneme repetition evaluation;

[0027] The reduplicated word and phoneme repetition evaluation includes:

[0028] Detect whether there are reduplicated words in the custom wake-up word. If there are, deduct corresponding points according to the number of reduplicated words;

[0029] Detect whether there is a phoneme repetition rate in the custom wake-up word. If the phoneme repetition rate exceeds a preset ratio, deduct corresponding points according to the magnitude of the exceeded ratio value. The greater the exceeded value, the higher the deduction.

[0030] According to an activation method of an in-vehicle voice custom wake-up word provided by the present invention, determining the application status of the custom wake-up word according to the comprehensive score includes:

[0031] When the comprehensive score is greater than or equal to a first threshold, activate the custom wake-up word;

[0032] When the comprehensive score is greater than or equal to a second threshold and less than the first threshold, adjust the custom wake-up word and display the specific deduction reasons;

[0033] When the comprehensive score is less than the second threshold, replace the custom wake-up word and give detailed deduction reasons and improvement directions.

[0034] According to an activation method of an in-vehicle voice custom wake-up word provided by the present invention, when the application status of the custom wake-up word is to replace the custom wake-up word, update the custom wake-up word, and use the multi-dimensional evaluation model to evaluate the updated custom wake-up word again until the comprehensive score reaches the state of activating the custom wake-up word.

[0035] The present invention also provides an activation device for an in-vehicle voice custom wake-up word, including:

[0036] An acquisition unit, configured to acquire a custom wake-up word input by a user;

[0037] A scoring unit for inputting the custom wake word into a multi-dimensional evaluation model, and using the multi-dimensional evaluation model to comprehensively score the custom wake word to generate a comprehensive score of the custom wake word; the evaluation items of the multi-dimensional evaluation model at least include: wake word length evaluation, conflict vocabulary matching evaluation;

[0038] A determination unit for determining the application status of the custom wake word according to the comprehensive score; the application status at least includes: activating the custom wake word.

[0039] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, and when the processor executes the program, it implements the activation method of the in-vehicle voice custom wake word as described in any one of the above.

[0040] The present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the activation method of the in-vehicle voice custom wake word as described in any one of the above.

[0041] The present invention also provides a computer program product, including a computer program, and when the computer program is executed by a processor, it implements the activation method of the in-vehicle voice custom wake word as described in any one of the above.

[0042] The activation method, device, and vehicle of the in-vehicle voice custom wake word provided by the present invention comprehensively evaluate multiple aspects of the wake word through a multi-dimensional evaluation model, so as to be able to screen out wake words that better meet the system requirements, which helps to reduce system failures or frequent false triggers caused by improper setting of custom wake words, thereby reducing the risks of wake-up failure and false wake-up, and providing a more intelligent and personalized voice interaction experience for users. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 is a flowchart of the activation method of the in-vehicle voice custom wake word provided by the present invention;

[0044] Figure 2 is a structural diagram of the activation device of the in-vehicle voice custom wake word provided by the present invention;

[0045] Figure 3 is a structural diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0046] To make the objectives, technical solutions, and advantages of the present invention clearer, the following will clearly and completely describe the technical solutions in the present invention with reference to the accompanying drawings in the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of them. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present invention without creative efforts belong to the scope of protection of the present invention.

[0047] The execution subject of the present invention is an in-vehicle voice control system.

[0048] Figure 1 It is a flowchart of the activation method for the in-vehicle voice custom wake-up word provided by the present invention. As Figure 1 shown, the method includes the following steps:

[0049] Step 101: Obtain the custom wake-up word input by the user.

[0050] Specifically, the user initiates a custom wake-up word setting operation in the in-vehicle voice interaction system. The user can, according to their own preferences and needs, enter the custom wake-up word in the input box and click confirm after completion. These wake-up words are phrases or words that the user hopes to use to activate the voice assistant and implement various functional operations.

[0051] Step 102: Input the custom wake-up word into a multi-dimensional evaluation model, and use the multi-dimensional evaluation model to comprehensively score the wake-up word to generate a comprehensive score for the custom wake-up word; the evaluation items of the multi-dimensional evaluation model at least include: wake-up word length evaluation, conflicting word matching evaluation.

[0052] Specifically, in the embodiment of the present invention, the multi-dimensional evaluation model includes 2 evaluation items, namely: wake-up word length evaluation, conflicting word matching evaluation. Among them, for the wake-up word length evaluation, the lower the length of the custom wake-up word, the easier it is to be mis-awakened, and the higher the length of the wake-up word, the lower the wake-up rate. Therefore, if the length of the custom wake-up word is too long or too short, it will affect the recognition accuracy of the custom wake-up word and ultimately affect the wake-up rate. For the conflicting word matching evaluation, wake-up words with inappropriate words will pose safety hazards or social impacts. Therefore, it is necessary to establish conflicting words and perform matching detection on the wake-up word through the conflicting words to avoid using inappropriate wake-up words. Therefore, the multi-dimensional evaluation model provided by the present application can comprehensively and meticulously evaluate the wake-up word, so as to screen out wake-up words that better meet the system requirements. When the wake-up word input by the user is sent into the multi-dimensional evaluation model, a comprehensive score is generated by this model. This score reflects the comprehensive performance of the wake-up word in multiple dimensions and is an important basis for determining whether to activate the wake-up word.

[0053] The present invention comprehensively evaluates the length of the wake-up word, the matching degree with conflicting words, etc. through a multi-dimensional evaluation model, and can screen out wake-up words that better meet the system requirements. This helps to reduce system failures or frequent false triggers caused by improper setting of wake-up words, reduce the risks of wake-up failure and false wake-up, and provide users with a more intelligent and personalized voice interaction experience.

[0054] Step 103: Determine the application status of the wake-up word according to the comprehensive score; the application status at least includes: activating the custom wake-up word.

[0055] Specifically, the application status of the custom wake-up word of the present invention can be classified into corresponding levels according to actual needs, and the classified levels are associated with the comprehensive score, so as to achieve the purpose of determining the application status of the custom wake-up word according to the comprehensive score. In the embodiment of the present invention, the application status of the custom wake-up word is to activate the custom wake-up word. When the final score of the custom wake-up word input by the user is to activate the custom wake-up word, the system can activate and use the custom wake-up word.

[0056] The activation method of the in-vehicle voice custom wake-up word provided by this application evaluates the custom wake-up word from at least two aspects through a multi-dimensional evaluation model and obtains a comprehensive score accordingly. Then, according to this comprehensive score, it is determined whether the custom wake-up word input by the user is activated. This method effectively avoids problems such as system failures or frequent false triggers of the in-vehicle voice interaction system caused by wake-up failure or false wake-up of the custom wake-up word, and improves the user experience.

[0057] In another embodiment of the present invention, the comprehensive scoring of the custom wake-up word by using the multi-dimensional evaluation model specifically includes the following scheme:

[0058] Detect whether the length of the custom wake-up word is within the set range. If it is not within the set range, corresponding deductions are made according to the deviation degree of the custom wake-up word length from the set range; and it is determined whether to perform a conflicting word matching evaluation according to the deduction situation; if it is within the set range, a conflicting word matching evaluation is performed; the conflicting word matching evaluation includes: matching the custom wake-up word with the words in the conflict word library to determine whether the conflict word library includes the custom wake-up word; if it is included, the score of the corresponding custom wake-up word is recorded as 0, and the user is prompted to modify it; the conflict word library includes: a sensitive word library and a restricted word library; and the conflict word library is a word library obtained by regularly updating the sensitive word library and the restricted word library.

[0059] Specifically, set the length range of the custom wake word, for example, 2 - 5 Chinese characters or 3 - 6 English letters. Score according to the deviation degree of the length of the wake word from the set range. The greater the deviation, the higher the deduction, and the lower the comprehensive score. For example, after the user inputs a custom wake word, the system detects whether its length is within the set range of the shortest [X] characters and the longest [Y] characters. If it is within this range, proceed to the next step of "conflict word library evaluation". If it is not within this range, further determine the deviation degree of the length of the custom wake word from the set range. If the deviation is too large, resulting in a very low deduction and thus the comprehensive score is lower than a certain preset threshold, then do not conduct the conflict word library evaluation and directly suggest that the user change the custom wake word.

[0060] The following explains the conflict vocabulary matching evaluation:

[0061] Since wake words with inappropriate vocabulary may pose safety hazards or social impacts, this application constructs a sensitive word library and a restricted word library to ensure that the custom wake word conforms to social ethics. This application compares the custom wake word input by the user with the sensitive word library and the restricted word library respectively to check whether the custom wake word is a word in the sensitive word library or the restricted word library. If it is, the multi - dimensional evaluation model records the score of the custom wake word as zero and prompts the user to modify it. Since sensitive words and restricted words are strictly prohibited in the custom wake word, as long as they appear, the user is required to modify it until the modified custom wake word meets the requirements. If only a deduction operation is performed when the custom wake word is a conflict vocabulary, it will cause the comprehensive score to be continuously pulled down, instead reducing the accuracy of the comprehensive score and being unfavorable for the effective activation of the custom wake word.

[0062] In addition, since no deduction is made during the conflict vocabulary evaluation process and it does not affect the comprehensive score, if the conflict vocabulary matching evaluation is performed first and then the wake word length evaluation is performed, then if the wake word length evaluation is unqualified later, the process of repeatedly modifying the conflict vocabulary matching to meet the requirements earlier is in vain. This not only wastes computing resources and time, but also reduces the user experience. Therefore, the embodiments of this application adopt the order of first performing the wake word length evaluation and then performing the conflict vocabulary matching evaluation.

[0063] Among them, to ensure that the conflict word library contains the latest sensitive vocabulary and restricted vocabulary, the conflict word library of this application is a word library that regularly updates the sensitive word library and the restricted word library. The update methods include: the system actively obtains, and the user actively gives feedback.

[0064] The activation method of the in-vehicle voice custom wake-up word provided by the embodiments of the present invention, on the one hand, by gradually detecting the length and conflicting words of the custom wake-up word, enables the system to quickly eliminate the wake-up words that do not meet the requirements, reduces the burden of subsequent detection, and improves the overall evaluation efficiency; on the other hand, since the length of the wake-up word is one of the important factors affecting the user experience and technical implementation, it is used as a basic condition for priority evaluation. On the basis of ensuring that the length of the wake-up word meets the requirements, the conflicting words are evaluated and optimized, which can more effectively improve the evaluation efficiency and user experience.

[0065] In another embodiment of the present invention, the evaluation items of the multi-dimensional evaluation model further include: voice command similarity evaluation; the embodiments of the present invention use the multi-dimensional evaluation model to comprehensively score the custom wake-up word, including: first, evaluating the length of the wake-up word and matching the conflicting words of the custom wake-up word in sequence, and determining whether to perform voice command similarity evaluation according to the evaluation results; the voice command similarity evaluation includes: determining the similarity between the custom wake-up word and the commands in the pre-constructed voice command library, and deducting corresponding points according to the similarity. The lower the similarity, the higher the deduction.

[0066] Specifically, if the custom wake-up word is similar to the common voice command, it will cause mis-triggering. Therefore, this application calculates the similarity between the wake-up word and the voice command. The higher the similarity, the lower the comprehensive score, and the user is prompted to select a wake-up word with a greater difference. In the embodiments of the present invention, the length of the wake-up word is a basic condition. Prioritized detection can quickly eliminate overly long or short wake-up words and reduce the burden of subsequent detection. Sensitive words and restricted words involve security and social impacts, and wake-up words that do not meet the requirements should be detected and excluded as early as possible. On the basis of ensuring the length and content security of the wake-up word, further detecting its similarity to the common voice command can effectively avoid mis-triggering. If the voice command similarity evaluation is performed first and then the conflicting word evaluation is performed, there may be a risk that even if the custom wake-up word input by the user is not similar to the common command, it may involve sensitive words or restricted words, resulting in the voice command similarity evaluation being in vain, thus reducing the evaluation efficiency and user experience.

[0067] The activation method of the in-vehicle voice custom wake-up word provided by the embodiments of the present invention, on the one hand, evaluates the custom wake-up word in three aspects: wake-up word length evaluation, conflicting word matching evaluation, and voice command similarity evaluation through the multi-dimensional evaluation model, making the finally selected custom wake-up word more in line with the system requirements and further reducing the risks of wake-up failure and mis-wake-up; on the other hand, by evaluating the custom wake-up word in the order of wake-up word length evaluation, conflicting word matching evaluation, and voice command similarity evaluation in sequence, not only improves the evaluation efficiency, but also further enhances the user experience.

[0068] In another embodiment of the present invention, in the voice command similarity detection, a speech recognition algorithm can be used, such as a voice similarity model based on deep learning, to improve the accuracy of similarity detection, thereby further reducing the risk of wake-up failure and false wake-up, and enhancing the user experience.

[0069] In another embodiment of the present invention, the evaluation items of the multi-dimensional evaluation model further include: symbol evaluation. The embodiment of the present invention uses the multi-dimensional evaluation model to comprehensively score the custom wake-up word, including: first, sequentially performing wake-up word length evaluation, conflicting word matching evaluation, and voice command similarity evaluation on the custom wake-up word, and determining whether to perform symbol evaluation according to the evaluation results; the symbol evaluation includes: identifying whether the custom wake-up word contains non-alphanumeric characters, and if so, deducting corresponding points according to the number of non-alphanumeric characters.

[0070] Specifically, the so-called symbol evaluation refers to checking whether the custom wake-up word contains special symbols (such as "@", "#", "$", etc.). If the custom wake-up word includes special symbols, points will be deducted according to the number of special symbols. Each symbol will be deducted [P] points according to the scoring rules, and the deduction situation will be recorded, and the final comprehensive score will be determined according to the deduction situation. Since custom wake-up words containing special symbols will increase the recognition difficulty, therefore, in order to reduce the recognition difficulty of the custom wake-up word, improve the wake-up rate of the custom wake-up word, and improve the user experience, the symbol evaluation is also added to the evaluation items of the multi-dimensional evaluation model. Wake-up words containing special symbols will increase the difficulty of the system to recognize the custom wake-up word, resulting in a lower comprehensive score. Therefore, this application recommends that users use pure text or letters.

[0071] The activation method of the in-vehicle voice custom wake-up word provided by the embodiment of the present invention evaluates the custom wake-up word in four aspects: wake-up word length evaluation, conflicting word matching evaluation, voice command similarity evaluation, and symbol evaluation through the multi-dimensional evaluation model, making the finally selected custom wake-up word more in line with the system requirements, further reducing the risk of wake-up failure and false wake-up, and enhancing the user experience.

[0072] In another embodiment of the present invention, the evaluation items of the multi-dimensional evaluation model further include: reduplicated word and phoneme repetition evaluation. The embodiment of the present invention uses the multi-dimensional evaluation model to comprehensively score the custom wake-up word, including: first, sequentially performing wake-up word length evaluation, conflicting word matching evaluation, voice command similarity evaluation, and symbol evaluation on the custom wake-up word, and then performing reduplicated word and phoneme repetition evaluation; the reduplicated word and phoneme repetition evaluation includes: detecting whether there are reduplicated words in the custom wake-up word, and if so, deducting corresponding points according to the number of reduplicated words; detecting whether there is a phoneme repetition rate in the custom wake-up word, and if the phoneme repetition rate exceeds the preset ratio, deducting corresponding points according to the size of the exceeded value, and the larger the exceeded value, the higher the deduction.

[0073] Specifically, overly repeated phonemes will affect the recognition effect of the custom wake-up word, resulting in a decrease in the comprehensive score. Therefore, it is recommended that users reduce repetition. The embodiments of the present invention utilize speech recognition and analysis technologies to detect the occurrence of reduplicated words and phoneme repetitions in the custom wake-up word. If there are reduplicated words, each reduplicated word will be deducted [Q] points according to the scoring rules, and the deduction situation will be recorded. If the phoneme repetition rate exceeds [Z]%, then according to the value of the exceeding ratio, [S] points will be deducted according to the scoring rules, and the greater the exceeding value, the higher the deduction. If there are no reduplicated words and phoneme repetitions, no points will be deducted.

[0074] The activation method of the in-vehicle voice custom wake-up word provided by the embodiments of the present invention evaluates the custom wake-up word in five aspects: wake-up word length evaluation, conflicting word matching evaluation, voice command similarity evaluation, and symbol evaluation through a multi-dimensional evaluation model, making the finally selected custom wake-up word more in line with system requirements, further reducing the risks of wake-up failure and false wake-up, and enhancing the user experience.

[0075] In another embodiment of the present invention, by introducing speech analysis technology and combining with a machine learning model, the phoneme repetition situation in the wake-up word can be detected more accurately, thereby avoiding the recognition difficulties caused by excessive repetition, further reducing the risks of wake-up failure and false wake-up, and enhancing the user experience.

[0076] In the embodiments of the present invention, the design of the detection order is crucial for improving the wake-up rate and user experience. Through a reasonable detection order, inappropriate wake-up words can be gradually excluded, reducing the situations of false wake-up and wake-up failure. Specifically, on the one hand, by gradually detecting the length, sensitive words, voice command similarity, symbol usage, and phoneme repetition of the wake-up word, the system can quickly exclude the wake-up words that do not meet the requirements, reducing the burden of subsequent detection and improving the overall efficiency. On the other hand, placing the detection of sensitive words and restricted words in the front can exclude unsafe wake-up words as early as possible, avoiding unnecessary subsequent detections. At the same time, as a basic condition, length detection is prioritized to quickly screen out the wake-up words that do not meet the length requirements, thereby reducing misjudgments. On the other hand, through gradual detection and scoring, the system can provide detailed feedback to users, helping users understand the problems of the wake-up word, so that they can more accurately identify the custom wake-up word, further improving the wake-up rate of the custom wake-up word, and ultimately enhancing the user experience.

[0077] In another embodiment of the present invention, determining the application status of the custom wake-up word according to the comprehensive score specifically includes the following solutions:

[0078] When the comprehensive score is greater than or equal to the first threshold, activate the custom wake-up word; when the comprehensive score is greater than or equal to the second threshold and less than the first threshold, adjust the custom wake-up word and display the specific deduction reasons; when the comprehensive score is less than the second threshold, replace the custom wake-up word and give the detailed deduction reasons and improvement directions.

[0079] Specifically, when the comprehensive score ≥ 80 points, prompt the user that the custom wake-up word has good applicability and can be directly used. When 60 points ≤ comprehensive score < 80 points, it is recommended that the user appropriately adjust the wake-up word according to the deduction reasons to improve the wake-up effect, and display the specific deduction reasons. When the comprehensive score < 60 points, it is strongly recommended that the user reset the wake-up word and elaborate on the deduction reasons and improvement directions. Finally, when the scoring process ends, wait for the user's further operations (such as accepting the current wake-up word, adjusting the wake-up word or resetting, etc.).

[0080] The activation method of the in-vehicle voice custom wake-up word provided by the embodiment of the present invention divides the application status of the custom wake-up word into 3 levels according to the comprehensive score. One level is to directly activate the custom wake-up word; one level is to suggest modifying the custom wake-up word; one level is to must modify the custom wake-up word, which not only reduces the situations of false wake-up and wake-up failure, but also for users, the custom wake-up word input by the user will not be directly disabled due to unqualified, but it is recommended that the user replace or modify it, making the user experience better.

[0081] In another embodiment of the present invention, when the application status of the custom wake-up word is to replace the custom wake-up word, update the custom wake-up word, and use the multi-dimensional evaluation model to evaluate the updated custom wake-up word again until the comprehensive score reaches the state of activating the custom wake-up word.

[0082] The activation method of the in-vehicle voice custom wake-up word provided by the embodiment of the present invention further reduces the situations of false wake-up and wake-up failure by repeatedly updating the custom wake-up word until the comprehensive score reaches the state of activating the custom wake-up word.

[0083] In another embodiment of the present invention, during the process of the user inputting the wake-up word, the system can display the score and potential problems of the current wake-up word in real time and help the user adjust in time.

[0084] The following gives an overall example introduction to the activation method of the in-vehicle voice custom wake-up word provided by the present invention:

[0085] Start: The user initiates the custom wake-up word setting operation in the in-vehicle voice interaction system.

[0086] Input wake-up word: The user inputs the custom wake-up word in the input box and clicks confirm after inputting.

[0087] Length detection: The scoring module obtains the wake-up word input by the user and detects whether its length is within the range of the set minimum of [X] characters and maximum of [Y] characters.

[0088] Yes: Proceed to the next step, "sensitive word detection".

[0089] No: Depending on whether the length is exceeded or insufficient, deduct [M] points (if the length is less than [X]) or [N] points (if the length is greater than [Y]) according to the scoring rules, record the deduction situation, and then proceed to the next step, "sensitive word detection".

[0090] Sensitive word detection: The scoring module compares the wake-up word with the sensitive word and restricted word library to check if it contains the words in the library.

[0091] Yes: The wake-up word containing sensitive words or restricted words gets a score of zero, and the user is prompted to modify it.

[0092] No: Directly proceed to the next step, "comparison with voice commands".

[0093] Comparison with voice commands: Using a voice similarity algorithm (such as the dynamic time warping algorithm), match the similarity between the wake-up word and the commands in the common voice command comparison library.

[0094] Same: Deduct [G] points, record the deduction situation, and proceed to the next step, "symbol detection".

[0095] Similar: Depending on the degree of similarity, deduct [3 - 8] points according to the scoring rules, record the deduction situation, and proceed to the next step, "symbol detection".

[0096] Different: Directly proceed to the next step, "symbol detection".

[0097] Symbol detection: The scoring module checks if the wake-up word contains special symbols (such as "@", "#", "$", etc.).

[0098] Yes: Deduct [P] points for each symbol according to the scoring rules, record the deduction situation, and proceed to the next step, "detection of repeated words and phonemes".

[0099] No: Directly proceed to the next step, "detection of repeated words and phonemes".

[0100] Detection of repeated words and phonemes: Using speech recognition and analysis technology, detect the repeated words and phoneme repetition in the wake-up word.

[0101] Existence of repeated words: Deduct [Q] points for each repeated word according to the scoring rules, record the deduction situation.

[0102] Phoneme repetition rate exceeds [Z]%: Depending on the exceeding ratio, deduct [S] points according to the scoring rules, record the deduction situation.

[0103] No above situation: no deduction of points.

[0104] Calculate the total score: The scoring module summarizes the deduction of points in the previous steps to obtain the final wake word score (with a full score of 100 points, subtracting the sum of the points deducted in each step).

[0105] Feedback suggestions: Provide feedback to the user according to the final scoring result:

[0106] Score ≥ 80 points: Prompt the user that the applicability of this wake word is good and can be used directly.

[0107] 60 points ≤ score < 80 points: Suggest that the user appropriately adjust the wake word according to the reasons for the deduction of points to improve the wake-up effect, and display the specific reasons for the deduction of points.

[0108] Score < 60 points: Strongly suggest that the user reset the wake word, and elaborate on the reasons for the deduction of points and the improvement direction.

[0109] End: The scoring process ends, waiting for the user's further operations (such as accepting the current wake word, adjusting the wake word, or resetting, etc.).

[0110] An embodiment of the present invention further provides an activation device for a vehicle-mounted voice custom wake word, as Figure 2 shown. This device includes:

[0111] An acquisition unit 201, configured to acquire a custom wake word input by a user;

[0112] A scoring unit 202, configured to input the custom wake word into a multi-dimensional evaluation model, and use the multi-dimensional evaluation model to comprehensively score the custom wake word to generate a comprehensive score of the custom wake word; at least the evaluation items of the multi-dimensional evaluation model include: wake word length evaluation, conflict vocabulary matching evaluation;

[0113] A determination unit 203, configured to determine the application status of the custom wake word according to the comprehensive score; the application status at least includes: activating the custom wake word.

[0114] Figure 3 Illustrated is a schematic physical structure diagram of an electronic device, as Figure 3 shown. This electronic device may include: a processor 310, a communication interface 320, a memory 330, and a communication bus 340. Among them, the processor 310, the communication interface 320, and the memory 330 communicate with each other through the communication bus 340. The processor 310 can call the logical instructions in the memory 330 to execute the activation method of the vehicle-mounted voice custom wake word, and this method includes:

[0115] Obtain a custom wake-up word input by the user;

[0116] Input the custom wake-up word into a multi-dimensional evaluation model, and use the multi-dimensional evaluation model to comprehensively evaluate the custom wake-up word to generate a comprehensive score for the custom wake-up word; the evaluation items of the multi-dimensional evaluation model at least include: wake-up word length evaluation, conflicting vocabulary matching evaluation;

[0117] Determine the application status of the custom wake-up word according to the comprehensive score; the application status at least includes: activating the custom wake-up word.

[0118] In addition, when the logical instructions in the above-mentioned memory 330 are implemented in the form of software functional units and sold or used as an independent product, they can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that makes a contribution to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs that can store program codes.

[0119] On the other hand, the present invention also provides a computer program product, the computer program product includes a computer program, the computer program can be stored on a non-transitory computer-readable storage medium, and when the computer program is executed by a processor, the computer can execute the activation method of the in-vehicle voice custom wake-up word provided by the above-mentioned various methods. The method includes:

[0120] Obtain a custom wake-up word input by the user;

[0121] Input the custom wake-up word into a multi-dimensional evaluation model, and use the multi-dimensional evaluation model to comprehensively evaluate the custom wake-up word to generate a comprehensive score for the custom wake-up word; the evaluation items of the multi-dimensional evaluation model at least include: wake-up word length evaluation, conflicting vocabulary matching evaluation;

[0122] Determine the application status of the custom wake-up word according to the comprehensive score; the application status at least includes: activating the custom wake-up word.

[0123] In another aspect, the present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the activation method of the vehicle-mounted voice custom wake-up word provided by the above-mentioned various methods. The method includes:

[0124] Obtain a custom wake-up word input by a user;

[0125] Input the custom wake-up word into a multi-dimensional evaluation model, and use the multi-dimensional evaluation model to comprehensively score the custom wake-up word to generate a comprehensive score of the custom wake-up word; at least the evaluation items of the multi-dimensional evaluation model include: wake-up word length evaluation, conflict vocabulary matching evaluation;

[0126] Determine the application status of the custom wake-up word according to the comprehensive score; the application status at least includes: activating the custom wake-up word.

[0127] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated. The components shown as units may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. A person of ordinary skill in the art can understand and implement it without creative labor.

[0128] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, also by hardware. Based on such an understanding, the above technical solution essentially or the part that contributes to the prior art can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions for causing 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.

[0129] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of each embodiment of the present invention.

[0130] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for activating a vehicle-mounted voice custom wake-up word, characterized in that, Including: Obtain a custom wake-up word input by the user; Input the custom wake-up word into a multi-dimensional evaluation model, and use the multi-dimensional evaluation model to comprehensively score the custom wake-up word to generate a comprehensive score of the custom wake-up word; The evaluation items of the multi-dimensional evaluation model at least include: wake-up word length evaluation, conflicting vocabulary matching evaluation; Determine the application status of the custom wake-up word according to the comprehensive score; the application status at least includes: activating the custom wake-up word.

2. The activation method of the in-vehicle voice custom wake-up word according to claim 1, wherein The comprehensive scoring of the custom wake-up word using the multi-dimensional evaluation model includes: Detect whether the length of the custom wake-up word is within a set range. If it is not within the set range, make corresponding deductions according to the deviation degree of the custom wake-up word length from the set range, and determine whether to perform conflicting vocabulary matching evaluation according to the deduction situation; if it is within the set range, perform the evaluation of conflicting vocabulary matching; The conflicting vocabulary matching evaluation includes: Match and detect the custom wake-up word with the vocabulary in the conflicting word library to determine whether the conflicting word library includes the custom wake-up word; if it is included, record the score of the corresponding custom wake-up word as 0 and prompt the user to modify; The conflicting word library includes: a sensitive word library and a restricted word library; the conflicting word library is a word library obtained by regularly updating the sensitive word library and the restricted word library.

3. The activation method of the in-vehicle voice custom wake-up word according to claim 2, characterized in that, The evaluation items of the multi-dimensional evaluation model also include: voice command similarity evaluation; The comprehensive scoring of the custom wake-up word using the multi-dimensional evaluation model includes: First, perform wake-up word length evaluation and conflicting vocabulary matching evaluation on the custom wake-up word in sequence, and determine whether to perform voice command similarity evaluation according to the evaluation results; The voice command similarity evaluation includes: Determine the similarity between the custom wake-up word and the commands in the pre-constructed voice command library, and make corresponding deductions according to the similarity. The lower the similarity, the higher the deduction.

4. The activation method of the in-vehicle voice custom wake-up word according to claim 3, wherein, The evaluation items of the multi-dimensional evaluation model also include: symbol evaluation; The comprehensive scoring of the custom wake-up word using the multi-dimensional evaluation model includes: First, perform wake-up word length evaluation, conflicting vocabulary matching evaluation, voice command similarity evaluation on the custom wake-up word in sequence, and determine whether to perform symbol evaluation according to the evaluation results; The symbol evaluation includes: Identify whether the custom wake-up word contains non-alphanumeric characters. If it contains non-alphanumeric characters, make corresponding deductions according to the number of non-alphanumeric characters.

5. The activation method of the in-vehicle voice custom wake-up word according to claim 4, wherein The evaluation items of the multi-dimensional evaluation model also include: reduplicated word and phoneme repetition evaluation; The comprehensive scoring of the custom wake-up word using the multi-dimensional evaluation model includes: First, perform wake-up word length evaluation, conflicting vocabulary matching evaluation, voice command similarity evaluation, symbol evaluation on the custom wake-up word in sequence, and then perform reduplicated word and phoneme repetition evaluation; The reduplicated word and phoneme repetition evaluation includes: Detect whether there are reduplicated words in the custom wake-up word. If there are reduplicated words, make corresponding deductions according to the number of reduplicated words; Detect whether there is a phoneme repetition rate for the custom wake-up word. If the phoneme repetition rate exceeds a preset ratio, corresponding deductions are made according to the magnitude of the exceeded ratio. The larger the exceeded value, the higher the deduction.

6. The activation method of the vehicle-mounted voice custom wake-up word according to any one of claims 1-5, characterized in that The determining the application status of the custom wake-up word according to the comprehensive score includes: When the comprehensive score is greater than or equal to the first threshold, activate the custom wake-up word; When the comprehensive score is greater than or equal to the second threshold and less than the first threshold, adjust the custom wake-up word and display the specific reason for deduction; When the comprehensive score is less than the second threshold, replace the custom wake-up word and give the detailed reason for deduction and the improvement direction.

7. The activation method of the in-vehicle voice custom wake-up word according to claim 6, characterized in that When the application status of the custom wake-up word is to replace the custom wake-up word, update the custom wake-up word, and use the multi-dimensional evaluation model to evaluate the updated custom wake-up word again until the comprehensive score reaches the state of activating the custom wake-up word.

8. An activation device for a vehicle-mounted voice custom wake-up word, characterized in that, including: An acquisition unit for acquiring the custom wake-up word input by the user; A scoring unit for inputting the custom wake-up word into the multi-dimensional evaluation model and using the multi-dimensional evaluation model to comprehensively score the custom wake-up word to generate the comprehensive score of the custom wake-up word; The evaluation items of the multi-dimensional evaluation model at least include: wake-up word length evaluation, conflicting vocabulary matching evaluation; A determination unit for determining the application status of the custom wake-up word according to the comprehensive score; the application status at least includes: activating the custom wake-up word.

9. A vehicle, characterized in that, Equipped with the in-vehicle voice custom wake-up word adaptation and activation device as described in claim 8.

10. An electronic device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the activation method of the in-vehicle voice custom wake-up word as described in any one of claims 1 to 7.

Images