Custom wake-up word detection method, electronic equipment, storage medium and program product

Abstract

The embodiment of the invention provides a user-defined wake-up word detection method, electronic equipment, a storage medium and a program product, and the method comprises the steps: obtaining a user-defined first wake-up word, and determining a first score according to a first parameter of the first wake-up word, the first parameter comprises a first wake-up rate of the first wake-up word and first response time of the first wake-up word; determining a first historical wake-up word meeting a matching degree condition with the first wake-up word in the stored historical wake-up words; determining a target score of the first wake-up word according to the first score, a second score of the first historical wake-up word and a matching degree between the first wake-up word and the first historical wake-up word, the second score of the first historical wake-up word is related to the user adoption frequency of the first historical wake-up word, the second wake-up rate of the first historical wake-up word and the second response time of the first historical wake-up word. Therefore, the wake-up effect of the first wake-up word is accurately estimated, and a user can intuitively perceive the wake-up effect.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a custom wake-up word detection method, electronic device, storage medium and program product. Background Art

[0002] The voice interaction system can provide voice interaction functions in a timely manner based on wake-up words. For example, when the voice interaction system detects that the user says a preset wake-up word, it can respond to the user's voice command and provide services corresponding to the voice command. The user does not need to manually start the voice interaction system, which is convenient and fast.

[0003] In order to improve the user experience of the voice interaction system, the user can freely set the wake-up word of the voice interaction system according to his or her own preferences. However, there may be a large number of words with similar pronunciations in the customized wake-up words. For example, the word "extraordinary" has similar words such as "feifan" and "feifan". The voice interaction system may recognize the specific content of the wake-up word as other words, causing the user to mistakenly recognize "feifan" when he or she correctly says "extraordinary", resulting in the failure of the voice interaction system to wake up and fail to provide voice interaction services to users in a timely manner.

[0004] From this, we can see that in actual application, the content of the wake-up word is strongly correlated with the recognition success rate of the voice interaction system, and the user-defined wake-up word may be difficult to wake up the voice interaction system. Summary of the invention

[0005] The embodiments of the present application provide a custom wake-up word detection method, an electronic device, a storage medium, and a program product, which solve the problem that it is difficult to wake up a voice interaction system using a custom wake-up word.

[0006] In a first aspect, the present application provides a custom wake-up word detection method, which is applied to a first electronic device, and the method includes: obtaining a first wake-up word customized by a user, and determining a first score based on a first parameter of the first wake-up word, wherein the first parameter includes a first wake-up rate of the first wake-up word and a first response time of the first wake-up word; determining a first historical wake-up word among the stored historical wake-up words that meets a matching condition with the first wake-up word; determining a target score for the first wake-up word based on the first score, the second score of the first historical wake-up word, and the matching degree between the first wake-up word and the first historical wake-up word, wherein the second score of the first historical wake-up word is related to the number of times the first historical wake-up word is adopted by users, the second wake-up rate of the first historical wake-up word, and the second response time of the first historical wake-up word.

[0007] It can be understood that the first wake-up rate is the wake-up rate of the voice interaction system by the speech synthesized by the first wake-up word (for example, the first audio data), and the first response time is the response time (for example, the average response time obtained from the test) required for the speech synthesized by the first wake-up word (for example, the first audio data) to wake up the voice interaction system. Furthermore, based on the above-mentioned first wake-up rate and first response time, the first score representing the test wake-up effect of the synthesized speech of the first wake-up word can be determined.

[0008] Furthermore, based on the first score, the second score representing the actual wake-up effect of the speech synthesized by the first historical wake-up word that meets the matching conditions with the first wake-up word on the voice interaction system, and the degree of match between the two, the target score representing the expected wake-up effect of the first wake-up word in the actual wake-up application can be accurately estimated. In this way, the expected wake-up effect of the first wake-up word can be accurately estimated, and the method based on the target score can allow the user to intuitively perceive the wake-up effect. Furthermore, the user can intuitively modify, replace, delete or adopt the first wake-up word according to the target score, thereby improving the wake-up efficiency of the wake-up word set by the user and reducing the number of cases where the user fails to wake up the voice interaction system.

[0009] In a possible implementation of the first aspect above, a method for obtaining the first parameter of the first wake-up word includes: generating first audio data according to the first wake-up word; inputting the first audio data into a preset wake-up model to perform a preset number of cyclic wake-up tests to obtain a first wake-up rate and a first response time, wherein the first wake-up rate is determined based on a preset number of successful wake-up times and a preset number of times of the cyclic wake-up tests, and the first response time is determined based on an average response time of the response times of the preset number of cyclic wake-up tests.

[0010] It can be understood that a cyclic wake-up test of the voice interaction system can be performed with a preset number of times as the total number of tests, and a more accurate first wake-up rate and first response time can be recorded through multiple tests. For example, the ratio of the number of successful wake-up times determined by the cyclic wake-up test to the preset number of times (such as the total number of tests) can be the first wake-up rate. The wake-up response duration obtained by each test in the cyclic wake-up test is combined and accumulated to obtain the total response duration. The ratio of the total response duration to the preset number of times (such as the total number of tests) can be the first response time, that is, the average value obtained by dividing the total response time by the preset number of times can be used as the first response time.

[0011] In a possible implementation of the first aspect above, a first score is determined according to a first parameter of the first wake-up word, including: determining the first score based on a first wake-up rate and a first response time, and a first weight coefficient corresponding to the first wake-up rate and a second weight coefficient corresponding to the first response time.

[0012] It can be understood that the first weight coefficient and the second weight coefficient are used to characterize the importance of the first wake-up rate and the first response time relative to the first score.

[0013] In a possible implementation of the first aspect, determining a target score for the first wake-up word according to the first score, the second score for the first historical wake-up word, and a degree of match between the first wake-up word and the first historical wake-up word includes: obtaining the target score by the following formula:

[0014] Score new =BaseScore+ln(StandardScore*M)

[0015] Among them, Score new is the target score, BaseScore is the first score, StandardScore is the second score, and M is the matching degree.

[0016] In a possible implementation of the first aspect above, the method for obtaining the second score of the first historical wake-up word includes: obtaining the third score of the first historical wake-up word based on the second wake-up rate and the second response time; determining the second score based on the number of user adoptions of the first historical wake-up word and the third score, as well as the third weight coefficient corresponding to the third score and the fourth weight coefficient corresponding to the number of user adoptions.

[0017] It can be understood that the third weight coefficient and the fourth weight coefficient are used to characterize the number of user adoptions and the importance of the third score relative to the second score.

[0018] In a possible implementation of the first aspect, determining the first score based on the first wake-up rate and the first response time, and a first weight coefficient corresponding to the first wake-up rate and a second weight coefficient corresponding to the first response time includes: obtaining the first score by the following formula:

[0019]

[0020] Among them, w 1 is the first weight coefficient, w 2 is the second weight coefficient, used to balance the importance of the first wake-up rate and the first response time relative to the first score; ∈ is a constant; BaseScore is the first score, SuccessRate is the first wake-up rate, and ResponseTime is the first response time.

[0021] It can be understood that here ∈ can be used to prevent division by zero errors.

[0022] In a possible implementation of the first aspect, determining the second score based on the number of user adoptions of the first historical wake-up word and the third score, and a third weight coefficient corresponding to the third score and a fourth weight coefficient corresponding to the number of user adoptions includes: determining the second score by the following formula:

[0023] StandardScore=α×BaseScore'+β×log 10 (UsageCount+1)

[0024] Among them, α is the third weight coefficient, β is the fourth weight coefficient, which is used to balance the importance of the number of user adoptions and the third score; StandardScore is the second score, BaseScore' is the third score, and UsageCount is the number of user adoptions.

[0025] In a possible implementation of the first aspect, obtaining a third score of the first historical wake-up word based on the second wake-up rate and the second response time includes: obtaining the third score by the following formula:

[0026]

[0027] Among them, w 3 is the fifth weight coefficient corresponding to the second wake-up rate, w 4 is the sixth weight coefficient corresponding to the second response time, used to balance the importance of the second wake-up rate and the second response time; ∈ is a constant; BaseScore' is the third score, SuccessRate' is the second wake-up rate, and ResponseTime' is the second response time.

[0028] It can be understood that here ∈ can be used to prevent division by zero errors.

[0029] In a possible implementation of the first aspect above, the stored historical wake-up words include standardized historical wake-up word data, wherein the standardized historical wake-up word data at least includes initial consonants, finals, and tones determined according to the pinyin of the historical wake-up words.

[0030] It can be understood that in the application scenario of the Chinese context, the first electronic device 1500 can establish an association relationship with the text content of the historical wake-up word based on the pinyin and tone of the historical wake-up word, and store them together in the historical wake-up word library.

[0031] In some embodiments, the pinyin and the number representing its corresponding tone can be stored together, for example, "Fei Fan, Fei Fan, 12", which can represent the historical wake-up word with the text content of "Fei Fan", whose pinyin is "Fei Fan" and the tone is "first tone + second tone". This can avoid the mismatch of homophones as much as possible, improve the wake-up rate of the wake-up word for the voice interaction system, and can also help the user's newly customized first wake-up word to more quickly match the most matching first historical wake-up word in the historical wake-up word library.

[0032] In a possible implementation of the first aspect above, the first electronic device is a server; the method further includes: sending the target score to a second electronic device, where the second electronic device is a terminal device.

[0033] For example, the terminal device may be a vehicle-mounted terminal 100, so that the server 200 may send the target score to the vehicle-mounted terminal 100 through a wired connection or a wireless connection, and the vehicle-mounted terminal 100 may display the target score. The vehicle-mounted terminal 100 may visually present the target score to the user, thereby facilitating the user to intuitively understand the expected wake-up effect of the first wake-up word, and reducing the situation where the user has difficulty in waking up the voice interaction system using the first wake-up word.

[0034] In a possible implementation of the first aspect above, the first electronic device is a terminal device; and the method further includes: displaying a target score.

[0035] It can be understood that the first electronic device 1500 can be a terminal device, and the first electronic device 1500 displays the target score. The terminal device is, for example, the vehicle terminal 100 described below, and the vehicle terminal 100 obtains the user-defined first wake-up word, and can directly display the target score after determining it.

[0036] In a second aspect, the present application provides an electronic device comprising: one or more processors; one or more memories; one or more memories storing one or more programs, and when one or more programs are executed by one or more processors, the electronic device executes the custom wake-up word detection method provided by the above-mentioned first aspect and various possible implementations of the above-mentioned first aspect.

[0037] In a third aspect, the present application provides a computer-readable medium having instructions stored thereon, which, when executed on a computer, causes the computer to execute the custom wake-up word detection method provided by the first aspect and various possible implementations of the first aspect.

[0038] In a fourth aspect, the present application provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the custom wake-up word detection method provided by the above-mentioned first aspect and various possible implementations of the above-mentioned first aspect.

[0039] The beneficial effects of the second to fourth aspects mentioned above can be referred to the relevant descriptions in the first aspect mentioned above and various possible implementations of the first aspect, and will not be elaborated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 A schematic diagram of a scenario in which a custom wake-up word wakes up a voice interaction system is shown;

[0041] Figure 2A A schematic diagram showing a specific implementation process of a custom wake-up word detection method provided according to some embodiments of the present application is shown;

[0042] Figure 2B A schematic diagram of a scenario in which a custom wake-up word wake-up voice interaction system is provided according to some embodiments of the present application is shown;

[0043] Figure 2C A schematic diagram showing an implementation of a custom wake-up word detection provided according to some embodiments of the present application;

[0044] Figure 3 A schematic diagram showing a specific implementation flow of calculating a first wake-up rate and a first response time according to some embodiments of the present application is shown;

[0045] Figure 4 A schematic diagram showing a framework structure of a first electronic device provided according to some other embodiments of the present application is shown;

[0046] Figure 5 A schematic diagram showing a specific implementation process of a cyclic wake-up test provided according to some embodiments of the present application is shown;

[0047] Figure 6 A schematic diagram showing a specific implementation process of a custom wake-up word detection method provided according to some embodiments of the present application is shown;

[0048] Fig. 7A A schematic diagram showing a standardized storage principle of historical wake-up words provided according to some embodiments of the present application is shown;

[0049] Figure 7B A schematic diagram showing the matching principle of the first wake-up word provided according to some embodiments of the present application is shown;

[0050] Figure 8 A schematic structural diagram of a first electronic device provided according to some embodiments of the present application is shown. DETAILED DESCRIPTION

[0051] In order to facilitate understanding of the technical solutions provided by the embodiments of the present application, the meanings of some related field terms involved in the embodiments of the present application are explained below.

[0052] (1) Text to speech (TTS): refers to a technology that converts text into audio, for example, receiving characters or phonemes as input to generate audio.

[0053] (2) Voice wake-up: This is a technology that activates the device and triggers the voice interaction function through a specific wake-up word or phrase, which enables the device to respond to the user's voice commands.

[0054] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions provided by the embodiments of the present application will be described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0055] It can be understood that the terminal device in the embodiment of the present application can also be called a terminal, a user equipment (UE), a mobile station (MS), a mobile terminal (MT), etc. The terminal device can be a mobile phone, a smart TV, a wearable device, a tablet computer (pad), a computer with a wireless transceiver function, a virtual reality (VR) terminal device, an augmented reality (AR) terminal device, a wireless terminal in industrial control, a wireless terminal in self driving, a wireless terminal in remote medical surgery, a wireless terminal in smart grid, a wireless terminal in transportation safety, a wireless terminal in smart city, a wireless terminal in smart home, a vehicle-mounted terminal, etc.

[0056] The server in the embodiments of the present application can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers. It can also be a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0057] The following text takes the terminal device as a vehicle-mounted terminal and the server as a cloud server as an example, and combines the relevant drawings to further describe the custom wake-up word detection method provided in the embodiment of the present application in detail.

[0058] Figure 1 A schematic diagram of a scenario in which a custom wake-up word is used to wake up a voice interaction system is shown.

[0059] refer to Figure 1 The voice interaction system 101 provided in the vehicle terminal 100 may be a vehicle voice assistant, and the voice interaction system 101 may provide a voice interaction service to the user by responding to the wake-up word spoken by the user. For example, when the voice interaction system 101 detects the input audio of the user speaking the wake-up word, the voice interaction service is started.

[0060] For example, the wake-up word proposed in the embodiment of the present application can be a word, a vocabulary, a phrase or a sentence, and there is no specific restriction on the number of words and content of the wake-up word.

[0061] As mentioned above, the above wake-up word can be customized by the user (hereinafter referred to as the user-defined wake-up word as the "first wake-up word") to enhance the user experience.

[0062] In order to solve the above-mentioned problem that it is difficult to wake up the voice interaction system using a custom wake-up word, an embodiment of the present application provides a custom wake-up word detection method, which is applied to a first electronic device, including: obtaining a first wake-up word customized by a user, and determining a first score according to a first parameter of the first wake-up word, wherein the first parameter includes a first wake-up rate of the first wake-up word and a first response time of the first wake-up word; determining a first historical wake-up word among the stored historical wake-up words that meets a matching condition with the first wake-up word; determining a target score for the first wake-up word according to the first score, the second score of the first historical wake-up word, and the matching degree between the first wake-up word and the first historical wake-up word, wherein the second score of the first historical wake-up word is related to the number of times the first historical wake-up word is adopted by users, the second wake-up rate of the first historical wake-up word, and the second response time of the first historical wake-up word.

[0063] Among them, the first wake-up rate is the wake-up rate of the voice interaction system by the speech synthesized by the first wake-up word (for example, the first audio data), and the first response time is the response time (for example, the average response time obtained by test) required for the speech synthesized by the first wake-up word (for example, the first audio data) to wake up the voice interaction system. Furthermore, based on the above-mentioned first wake-up rate and first response time, the first score representing the test wake-up effect of the synthesized speech of the first wake-up word can be determined.

[0064] Furthermore, based on the first score, the second score representing the actual wake-up effect of the speech synthesized by the first historical wake-up word that meets the matching conditions with the first wake-up word on the voice interaction system, and the degree of match between the two, the target score representing the expected wake-up effect of the first wake-up word in the actual wake-up application can be accurately estimated. In this way, the expected wake-up effect of the first wake-up word can be accurately estimated, and the method based on the target score can allow the user to intuitively perceive the wake-up effect. Furthermore, the user can intuitively modify, replace, delete or adopt the first wake-up word according to the target score, thereby improving the wake-up efficiency of the wake-up word set by the user and reducing the number of cases where the user fails to wake up the voice interaction system.

[0065] In some embodiments, based on the mapping relationship that the second wake-up rate corresponding to the first historical wake-up word is proportional to the wake-up effect, and the mapping relationship that the second response time corresponding to the first historical wake-up word is inversely proportional to the wake-up effect, a third score representing the test wake-up effect of the synthesized speech of the first historical wake-up word can be determined. Furthermore, the second score can be determined according to the number of user adoptions and the third score.

[0066] Combine the following Figure 2A , Figure 2B and Figure 2C The specific implementation process and principle of the custom wake-up word detection provided in the embodiment of the present application are described in detail.

[0067] Figure 2A A schematic diagram of a specific implementation process of a custom wake-up word detection method provided according to some embodiments of the present application is shown. Figure 2B A schematic diagram of a scenario in which a custom wake-up word wakes up a voice interaction system according to some embodiments of the present application is shown.

[0068] Here, Figure 2A The execution subject of the specific implementation process shown may be the first electronic device 1500 ( Figure 2A ), the first electronic device 1500 may be, for example, Figure 2B The vehicle-mounted terminal 100 or the server 200 shown will not be described in detail here.

[0069] refer to Figure 2A , the specific implementation process includes the following steps:

[0070] S201, obtaining a first wake-up word defined by a user, and determining a first score according to a first parameter of the first wake-up word, wherein the first parameter includes a first wake-up rate of the first wake-up word and a first response time of the first wake-up word.

[0071] Exemplarily, the first score can be used to characterize the wake-up effect of the speech synthesized by the first wake-up word on the voice interaction system. For example, the first electronic device 1500 can determine the synthesized speech by obtaining the first wake-up word, and test the first wake-up rate and first response time of the speech synthesized by the first wake-up word to the voice interaction system 101. Since the first score is positively correlated with the first wake-up rate and the first score is negatively correlated with the first response time, the first electronic device 1500 can determine the first score by the association between the first wake-up rate and the first response time and the first score.

[0072] In some embodiments, the first electronic device 1500 may provide a speech synthesis engine, and use the speech synthesis engine to synthesize the first wake-up word into the first audio data (for example, the speech synthesized by the first wake-up word), and may test whether the first audio data can wake up the voice interaction system, and how long the response time is required to wake up the voice interaction system, thereby determining the first wake-up rate and the first response time corresponding to the speech synthesized by the first wake-up word. The specific test method may be combined with the following Figure 3 , Figure 4 ,and Figure 5 A detailed description is given and no further elaboration is given here.

[0073] In some embodiments of the present application, the first wake-up word can be Chinese, English, or the language of any country, and is not specifically limited here.

[0074] In some embodiments of the present application, the first electronic device 1500 may be a terminal device, for example Figure 1 or Figure 2B The vehicle-mounted terminal 100 shown can obtain the user-defined first wake-up word by providing an input control. For example, the first electronic device 1500 can be provided with input software that can provide a text input control or a voice input control.

[0075] In some other embodiments of the present application, the first electronic device 1500 may also be a server (eg Figure 2B The server 200 shown in the figure) can obtain the first wake-up word sent by the second electronic device 1600 in a wireless or wired manner. The second electronic device 1600 can be a terminal device (such as the vehicle-mounted terminal 100 mentioned above).

[0076] For example, refer to Figure 2BIn the scenario where the user customizes the wake-up word, the vehicle terminal 100 detects the user's setting operation on the wake-up word, can obtain the customized wake-up word set by the user, and send the first wake-up word to the server 200. Exemplarily, the server 200 can be a cloud server, and the server 200 can provide cloud services. For example, the server 200 may include a software platform that uses application virtualization technology (application virtualization), and the software platform can provide an independent virtualization environment to encapsulate multiple types of commonly used software. Furthermore, the server 200 may be encapsulated with a speech synthesis engine (TTS engine), and can use the first wake-up word in text format obtained as input to generate corresponding first audio data, and can also evaluate whether the first audio data can effectively wake up the voice interaction system 101 mentioned above.

[0077] In some embodiments of the present application, the server 200 can intuitively quantify the wake-up effectiveness of the first audio data in the form of a score, and send the target score that represents the wake-up effectiveness of the first wake-up word to the vehicle terminal 100, which displays the target score. This allows the user to intuitively understand whether the first wake-up word customized by the user can effectively wake up the voice interaction system 101, which is convenient for the user to determine the first wake-up word that is easy to wake up the voice interaction system 101, and effectively reduces the situation where the user-defined wake-up word is difficult to wake up the voice interaction system 101 in actual use.

[0078] S202: Determine a first historical wake-up word among the stored historical wake-up words that satisfies a matching condition with the first wake-up word.

[0079] Exemplarily, the above-mentioned historical wake-up words are wake-up words that have been adopted by the user in the past. The first electronic device 1500 can record the wake-up words that the user has used to wake up the voice interaction system, so that after obtaining the first wake-up word, it can refer to the wake-up effects of the historical wake-up words in actual applications and estimate the actual wake-up effect of the first wake-up word.

[0080] Exemplarily, the above matching condition may be that the matching degree between the first wake-up word and the first historical wake-up word is greater than a preset threshold, or that the historical wake-up word with the greatest matching degree with the first wake-up word is the first historical wake-up word.

[0081] Among them, the matching degree can be used to characterize the similarity of the role played by the content between the first wake-up word and the first historical wake-up word in the voice wake-up process. In some embodiments, the matching degree can be the similarity of the pronunciation between the first wake-up word and the first historical wake-up word. For example, the more similar the initial consonants and finals are and the closer the pronunciation is, the greater the matching degree between the words. In other embodiments, the matching degree can be the similarity of the words used in the first wake-up word and the first historical wake-up word. For example, the number of words between the words is the same and the pronunciation of each word in the words is similar, the greater the matching degree.

[0082] In some embodiments, the above matching conditions may include the following:

[0083] (1) determining a first historical wake-up word among the stored historical wake-up words, whose matching degree with the first wake-up word is greater than a preset matching degree threshold; or,

[0084] (2) Determine a first historical wake-up word that has the highest matching degree with the first wake-up word among the stored historical wake-up words.

[0085] In some embodiments of the present application, the first electronic device 1500 may provide a storage environment for storing the above-mentioned historical wake-up words, and store the historical wake-up words in the storage environment.

[0086] In some other embodiments of the present application, in order to determine the first historical wake-up word with the greatest matching degree with the first wake-up word, the first electronic device 1500 can traverse each historical wake-up word in the matching historical wake-up word library to determine the matching degree between the first wake-up word and each historical wake-up word, thereby determining the first historical wake-up word with the greatest matching degree. The specific matching method can be combined with the following Fig. 7A and Figure 7B A detailed description is given and no further elaboration is given here.

[0087] S203, determining the target score of the first wake-up word according to the first score, the second score of the first historical wake-up word, and the degree of match between the first wake-up word and the first historical wake-up word, wherein the second score of the first historical wake-up word is related to the number of times the first historical wake-up word is adopted by users, the second wake-up rate of the first historical wake-up word, and the second response time of the first historical wake-up word.

[0088] It can be understood that by comprehensively considering the first score, the second score of the first historical wake-up word, and the matching degree between the first wake-up word and the first historical wake-up word, the expected wake-up effect of the first wake-up word can be determined more accurately and quantified as the target score.

[0089] In some embodiments, the first electronic device 1500 can determine the test wake-up effect of the voice synthesized by the first historical wake-up word according to the wake-up rate and response time of the first historical wake-up word. Furthermore, by comprehensively considering the test wake-up effect of the voice synthesized by the first historical wake-up word and the adoption of the first historical wake-up word, the actual wake-up effect of the voice synthesized by the first historical wake-up word on the voice interaction system can be determined to determine the second score.

[0090] In some embodiments of the present application, the wake-up rate, response time and number of user adoptions of the above-mentioned first historical wake-up word are associated with the first historical wake-up word and stored in the historical wake-up word library, so that when the first electronic device 1500 matches the first historical wake-up word, the wake-up rate and response time of the first historical wake-up word can be associated and retrieved to determine the third score representing the corresponding test effect of the first historical wake-up word, and the second score can be determined in combination with the number of user adoptions.

[0091] In some embodiments of the present application, the third score corresponding to the wake-up rate and response time of the above-mentioned first historical wake-up word can be stored in the historical wake-up word library together with the number of user adoptions, so that when the first electronic device 1500 matches the first historical wake-up word, it can associate and retrieve the third score of the first historical wake-up word, and determine the second score in combination with the number of user adoptions, thereby eliminating the time for calculating the third score.

[0092] In some other embodiments of the present application, each time the above-mentioned first historical wake-up word is adopted, the number of user adoptions is updated, and the second score associated with the first historical wake-up word is updated, and the updated second score is stored in the historical wake-up word library, so that when the first electronic device 1500 matches the first historical wake-up word, the second score can be associated and retrieved, eliminating the time for calculating the second score and the third score.

[0093] It can be understood that through the above steps S201 to S203, the first electronic device 1500 can estimate the test wake-up effect of the user-defined first wake-up word, and then based on the first score corresponding to the test wake-up effect of the first wake-up word, the second score corresponding to the actual wake-up effect of the first historical wake-up word that best matches the first wake-up word, and the degree of match between the first wake-up word and the first historical wake-up word, the expected wake-up effect of the first wake-up word can be accurately evaluated based on the actual wake-up effect of the first historical wake-up word, and the target score representing the expected wake-up effect can be obtained.

[0094] To understand how to implement custom wake-up word detection, Figure 2C A schematic diagram of an implementation of custom wake-up word detection provided according to some embodiments of the present application is shown.

[0095] refer to Figure 2CIn an embodiment of the present application, the expected wake-up effect of the first wake-up word can be comprehensively determined from three aspects. On the first aspect, the test wake-up effect of the speech synthesized by the first wake-up word on the voice interaction system can be determined; on the second aspect, the first historical wake-up word that is most similar to the first wake-up word can be determined from the historical wake-up word library, and the actual wake-up effect of the speech synthesized by the first historical wake-up word on the voice interaction system can be determined; on the third aspect, the degree of match between the above-mentioned first historical wake-up word and the first wake-up word can be determined. Furthermore, based on the test wake-up effect of the speech synthesized by the first wake-up word, the actual wake-up effect of the speech synthesized by the first historical wake-up word that best matches the first wake-up word on the voice interaction system, and the degree of match between the two, the expected wake-up effect of the first wake-up word in the actual wake-up application can be accurately estimated.

[0096] Continue to refer Figure 2C Combined with the reference above Figure 2B In some embodiments of the present application, when the first electronic device is the server 200 mentioned above, the expected wake-up effect of the first wake-up word in the actual wake-up application can be quantified as a target score and sent by the first electronic device 1500 to the second electronic device 1600 (for example Figure 2B The target score is displayed to the user in the vehicle terminal 100 shown, so that the user can intuitively understand the expected wake-up effect of the first wake-up word, and then the user can modify, replace, delete or adopt the first wake-up word based on the target score, effectively reducing the situation where the user-defined wake-up word is difficult to wake up the voice interaction system.

[0097] Combine the following Figure 3 , the specific implementation process of calculating the first wake-up rate and the first response time provided in the embodiment of the present application is described in detail.

[0098] Figure 3 A schematic diagram of a specific implementation process for calculating a first wake-up rate and a first response time according to some embodiments of the present application is shown. The execution subject of the following specific implementation process may be the first electronic device 1500 mentioned below, and the first electronic device 1500 may be, for example, the vehicle-mounted terminal 100 or the server 200 mentioned above, which will not be described in detail here.

[0099] The specific implementation process includes the following steps:

[0100] S301, generating first audio data according to a first wake-up word.

[0101] Here, the first wake-up word can be a machine-readable string, and the string can be associated with the text content corresponding to the first wake-up word, so that the machine can determine the text content corresponding to the first wake-up word by reading the machine-readable string. Furthermore, the first electronic device 1500 can analyze the text corresponding to the content of the first wake-up word, identify text features such as words, sentence structures, punctuation marks, and convert them into the basic unit of speech, "phonemes". And based on the pronunciation of each generated phoneme, determine the intonation, stress and rhythm information of the text of the first wake-up word, generate appropriate rhythm, and obtain the first audio data corresponding to the speech of the first wake-up word.

[0102] S302, input the first audio data into a preset wake-up model to perform a preset number of cyclic wake-up tests to obtain a first wake-up rate and a first response time, wherein the first wake-up rate is determined based on the number of successful wake-up times and a preset number of times of the preset number of cyclic wake-up tests, and the first response time is determined based on the average response time of the response times of the preset number of cyclic wake-up tests.

[0103] Here, the first electronic device 1500 can perform a preset number of cyclic wake-up tests on the voice interaction system 101 based on the synthesized first audio data, and record the average response time tested in the cyclic wake-up test as the first response time, and record the probability of successful wake-up in the cyclic wake-up test as the first wake-up rate, thereby accurately determining the wake-up effect of the first audio data corresponding to the first wake-up word in the test.

[0104] For example, the first electronic device 1500 can perform a cyclic wake-up test on the voice interaction system with a preset number of times as the total number of tests, and record a more accurate first wake-up rate and first response time through multiple tests. Thus, the first electronic device 1500 can determine the first wake-up rate and the first response time based on the number of successful wake-ups, the total number of tests, and the wake-up response duration obtained by each test. For example, the ratio of the number of successful wake-ups determined by the cyclic wake-up test to the preset number of times (such as the total number of tests) can be the first wake-up rate. The wake-up response durations obtained by each test in the cyclic wake-up test are combined and accumulated to obtain the total response duration, and the ratio of the total response duration to the preset number of times (such as the total number of tests) can be the first response time, that is, the average value obtained by dividing the total response duration by the preset number of times can be used as the first response time.

[0105] In some embodiments, reference Figure 4 The first electronic device 1500 may include a speech synthesis engine 201a and a preset wake-up model 201b to facilitate the implementation of the above Figure 3 The specific steps required to implement the process.

[0106] The first electronic device 1500 can convert text into speech based on the speech synthesis engine 201a. For example, the speech synthesis engine 201a can analyze the input text, identify text features such as words, sentence structures, punctuation marks, etc., convert the text into the basic unit of speech "phonemes", determine the pronunciation of each phoneme, and then determine the tone, stress and rhythm information of the text according to the text features, generate appropriate rhythm, and make the synthesized speech more natural and fluent. Finally, a synthesis algorithm (such as waveform synthesis, splicing synthesis or synthesis based on deep learning) is used to generate a speech waveform to obtain audio data corresponding to the input text.

[0107] The first electronic device 1500 can recognize voice based on the preset wake-up model 201b, and match the recognized voice with a pre-stored specific wake-up word. When the match is successful, the preset wake-up model 201b can simulate a preset number of cyclic wake-up calls to the voice interaction system 101, and record the number of times the voice interaction system 101 is awakened and the total response time, and then determine the first wake-up rate according to the ratio of the recorded wake-up number to the preset number of times, determine the average response time according to the ratio of the total response time to the preset number of times, and record the average response time as the first response time.

[0108] Therefore, combined with the first wake-up rate recorded above, the first electronic device 1500 can accurately determine the wake-up effect of the first audio data corresponding to the first wake-up word in the test.

[0109] To facilitate understanding of the specific implementation of the cyclic wake-up test mentioned in step S302 above, Figure 5 The specific implementation process of a cyclic wake-up test shown in some embodiments of the present application is described in detail.

[0110] Figure 5 A schematic diagram of a specific implementation process of a cyclic wake-up test provided according to some embodiments of the present application is shown.

[0111] refer to Figure 5 The execution subject of the following specific implementation process may be the first electronic device 1500, which will not be elaborated here.

[0112] The specific implementation process includes the following steps:

[0113] S501, using the first audio data to test and wake up the voice interaction system.

[0114] Exemplarily, the first electronic device 1500 can input the first audio data into the preset wake-up model 201b, and implement a test process of simulating a cyclic wake-up of the voice interaction system 101 based on the preset wake-up model 201b. For example, the preset wake-up model 201b can match the first wake-up word according to the text features identified by the first audio data. If the matching degree is greater than the preset wake-up threshold, a successful wake-up is recorded; if the matching degree is less than or equal to the preset wake-up threshold, the wake-up fails.

[0115] S502, record the number of successful wake-up times and the response time.

[0116] Exemplarily, the first electronic device 1500 may record the number of successful wake-up attempts and the wake-up response time based on a preset wake-up model 201 b .

[0117] S503, determine whether 20 tests are completed. If yes, the test is completed and step S504 can be executed; if no, the test needs to be continued and the above step S501 is returned to execute.

[0118] Exemplarily, the first electronic device 1500 can record the current total number of tests in each test based on the preset wake-up model 201b, and then determine whether the preset number of tests has been completed. For example, when 20 cycles of wake-up tests are preset, the test process can be terminated when the 20th wake-up test is completed.

[0119] S504, stop the test and record the total number of tests.

[0120] Exemplarily, the first electronic device 1500 can stop testing after completing 20 cycles of wake-up tests, and record the total number of tests, and use the wake-up response time, number of successful wake-ups and the total number of tests obtained in each test to record the accurate first wake-up rate and first wake-up time based on the preset wake-up model 201b.

[0121] In some embodiments, the first electronic device 1500 may obtain a vector S including a first wake-up rate and a first response time returned by the preset wake-up model 201b: S=(SuccessRate, ResponseTime), wherein SuccessRate is the first wake-up rate, and ResponseTime is the first response time.

[0122] It can be understood that through the above specific implementation process S501 to S504, the embodiment of the present application can use a preset number of times as the total test times, thereby controlling the number of cyclic executions of the cyclic wake-up test, so that the first electronic device 1500 can obtain a more accurate first wake-up rate and first wake-up time.

[0123] In other embodiments of the present application, the first electronic device 1500 can determine a scoring function based on the mapping relationship between the first wake-up rate and the first response time and the first score, such as the formula (1) shown below, so as to quickly obtain the first score of the first wake-up word. In addition, the first electronic device 1500 can quickly determine the second score representing the actual wake-up effect of each historical wake-up word based on the third score of each historical wake-up word, the number of user adoptions and the second score, such as the formula (2) shown below, wherein the first electronic device 1500 can also calculate the second wake-up rate and the second response time of the historical wake-up word according to the above scoring function, such as the formula (3) shown below, so as to quickly determine the third score of each historical wake-up word. Thereby improving the efficiency of calculating the second score. Furthermore, the first electronic device 1500 can quickly determine the corresponding target score according to the second score, the matching degree and the mapping relationship between the first score and the target score, such as the formula (4) shown below.

[0124] In addition, the first electronic device 1500 can also store each of the above historical wake-up words in association with the obtained second score, and obtain a standardized score table containing historical wake-up words and second scores corresponding to the historical wake-up words. When the first electronic device 1500 determines the first historical wake-up word that best matches the first wake-up word, it can search the standardized score table to quickly determine the second score associated with the first historical wake-up word, eliminating the calculation time of instantly calculating the second score of the first historical wake-up word, and effectively improving the efficiency of determining the target score.

[0125] Combine the following Figure 6 , Fig. 7A and Figure 7B , the specific implementation process of obtaining the target score based on formulas (1) to (4) and the standardized score table proposed in the embodiment of the present application is described in detail.

[0126] Figure 6 A schematic diagram of a specific implementation process of a custom wake-up word detection method provided according to some other embodiments of the present application is shown.

[0127] It can be understood that the executor of this specific implementation process is the first electronic device 1500, which will not be elaborated here.

[0128] refer to Figure 6 , the specific implementation process includes the following steps:

[0129] S601, obtaining a user-defined wake-up word.

[0130] It can be understood that the specific implementation of the above step S601 can refer to the specific implementation process of the above step S201, which will not be repeated here.

[0131] S602, synthesized audio.

[0132] It can be understood that the specific implementation of step S602 can refer to the specific implementation process of step S301 above, and will not be repeated here.

[0133] S603, cyclically testing the wake-up effect, and recording the first wake-up rate and the first response time.

[0134] It can be understood that the specific implementation of step S603 can refer to the specific implementation process of step S302 above, which will not be repeated here.

[0135] S604: Determine a first score based on the first awakening rate and the first response time, and a first weight coefficient corresponding to the first awakening rate and a second weight coefficient corresponding to the first response time.

[0136] The first weight coefficient and the second weight coefficient are used to characterize the importance of the first wake-up rate and the first response time relative to the first score.

[0137] Exemplarily, the first electronic device 1500 may obtain the first score based on the first weight coefficient, the second weight coefficient, the first wake-up rate, and the inverse of the first response time.

[0138] For example, the first score can be obtained by the following formula (1):

[0139]

[0140] Among them, w 1 is the first weight coefficient, w 2 is the second weight coefficient, used to balance the importance of the first wake-up rate and the first response time; ∈ is a constant; BaseScore is the first score, SuccessRate is the first wake-up rate, and ResponseTime is the first response time.

[0141] It can be understood that here ∈ can be used to prevent division by zero errors.

[0142] S605: Obtain the wake-up word adopted by the user.

[0143] Exemplarily, the first electronic device 1500 may record the acquired wake-up word adopted by the user, and at least include the wake-up word text adopted by the user. In some embodiments, the wake-up word adopted by the user may be standardized, and the standardized wake-up word is stored in a historical wake-up word library.

[0144] S606, standardize the wake-up word adopted by the user.

[0145] Exemplarily, the stored historical wake-up words include standardized historical wake-up word data, wherein the standardized historical wake-up word data at least includes initial consonants, finals, and tones determined according to the pinyin of the historical wake-up words.

[0146] It is understood that the first electronic device 1500 can store the wake-up words historically adopted by the user, for example, as a historical wake-up word library. In order to facilitate matching the first wake-up word with the stored historical wake-up words, each historical wake-up word stored in the historical wake-up word library can be standardized so that the first wake-up word can quickly match the most matching historical wake-up word.

[0147] For ease of understanding, the following Fig. 7A and Figure 7B , the standardized historical wake-up words proposed in the embodiments of the present application are described in detail.

[0148] Fig. 7A A schematic diagram showing a standardized storage principle of historical wake-up words provided according to some embodiments of the present application is shown. Figure 7B A schematic diagram showing the matching principle of the first wake-up word provided according to some embodiments of the present application is shown.

[0149] Here, in the application scenario of the Chinese context, the first electronic device 1500 can establish an association relationship with the text content of the historical wake-up word based on the pinyin and tone of the historical wake-up word, and store them together in the historical wake-up word library. Fig. 7A , the historical wake-up word "Fei Fan", its pinyin can be expressed as a combination of initial consonants and finals, such as "Fei fan", and its tone can be "first tone + second tone". In some embodiments, the pinyin and the numbers representing their corresponding tones can be stored together, for example, "Fei Fan, Fei fan, 12", which can represent the historical wake-up word with the text content "Fei Fan", its pinyin is "Fei fan", and the tone is "first tone + second tone". In this way, the incorrect matching of homophones can be avoided as much as possible, the wake-up rate of the wake-up word for the voice interaction system can be improved, and it can also help the user's newly customized first wake-up word to match the most matching first historical wake-up word in the historical wake-up word library more quickly.

[0150] For example, refer to Figure 7BWhen the first electronic device 1500 obtains the first wake-up word "Fei fan" newly customized by the user, the first electronic device 1500 can apply the pinyin of the first wake-up word "Fei fan" in the historical wake-up word library for matching, or apply the tone "first tone + second tone" of the first wake-up word for matching. Through multi-level matching calculations, the first historical wake-up word "Fei fan" closest to the first wake-up word "Fei fan" can be obtained. Therefore, based on the number of times the first historical wake-up word has been adopted and the wake-up effect under test, as well as the matching degree between the first wake-up word and the first historical wake-up word, it can be accurately estimated whether the newly customized first wake-up word is easy to wake up the voice interaction system in actual wake-up applications.

[0151] In some other embodiments, the above-mentioned tone may also be level and oblique tone information. The "level and oblique tone" referred to here refers to the four tones in Middle Chinese, collectively known as "level, shang, qu and ru". In addition to the level tone, the other three tones vary, so the three tones of "shang, qu and ru" are collectively referred to as oblique tones. Therefore, when the first electronic device 1500 recognizes that the acquired first wake-up word is ancient Chinese (for example, a line of poetry), the historical wake-up words stored in the "pinyin + level and oblique tone" format can be used to quickly match the best matching first historical wake-up word.

[0152] Furthermore, in some embodiments of the present application, the first electronic device 1500 can preliminarily screen out relatively similar historical wake-up words based on the "pinyin + tone" method, and then perform precise matching based on the "pinyin + tone" method to obtain the most matching first historical wake-up word.

[0153] In addition, in application scenarios in non-Chinese contexts, such as English scenarios, the first electronic device 1500 can also standardize the storage of the first wake-up word based on the international phonetic symbols of the first wake-up word and the associated characters used to represent tone information such as intonation and transposition, without restricting the language and country of the first wake-up word and the historical wake-up words.

[0154] In some embodiments, before calculating the matching degree between the first wake-up word and the historical wake-up word, the first electronic device 1500 may first filter out relatively similar historical wake-up words, thereby reducing the amount of matching degree calculation.

[0155] In some embodiments, the first electronic device 1500 can determine the number of characters corresponding to the first wake-up word, and match relatively similar historical wake-up words in the historical wake-up word library based on the number of characters. Assuming that the first wake-up word is "good morning", the corresponding historical wake-up words of three characters can be screened out. In some embodiments, the number of characters can be determined by adding a preset separator between the pinyin corresponding to each character of the standardized first wake-up word, for example, "zao&shang&hao", where the separator is "&", and two separators are three characters. Then, wake-up words with the same number of separators can be relatively similar wake-up words, so as to preliminarily screen out relatively close historical wake-up words before calculating the matching degree, thereby reducing the amount of calculation.

[0156] In some embodiments, the first electronic device 1500 can match a relatively close historical wake-up word based on the length of the pinyin. For example, the length of the pinyin "zhuang" corresponding to "zhuang" in the word Zhuangzi is 6 characters, and the length of the pinyin "zi" corresponding to "zi" is 2 characters, so the similarity between the two characters is low. Therefore, the length of the pinyin character corresponding to each word in the first wake-up word can be compared with each word in the historical wake-up word library to determine the difference in pinyin length, and the characters with a pinyin length difference less than a preset difference threshold are used as relatively close historical wake-up words, so as to preliminarily screen out relatively close historical wake-up words before calculating the matching degree, thereby reducing the amount of calculation.

[0157] In some embodiments, the preset difference threshold may be 10%.

[0158] In some embodiments, the above-mentioned pinyin length difference can be determined by the following method: the length difference between the pinyin corresponding to the character of the first wake-up word and the pinyin corresponding to the character of the historical wake-up word, and the ratio of the pinyin length of the character in the first wake-up word.

[0159] In some embodiments, the first electronic device 1500 may match the pinyin of each character of the first wake-up word and the position order of the character in the first wake-up word with the pinyin of the character in the corresponding position in the historical wake-up word in the historical wake-up word library, and filter out at least one historical wake-up word with exactly the same position and pinyin as a relatively similar historical wake-up word. Assuming that the first wake-up word is "Hello, little friend", the wake-up word with the first character "you" in all historical wake-up words can be determined based on "you" as a relatively similar historical wake-up word. Similarly, historical wake-up words with the second character "good", the third character "small", the fourth character "friend" and the fifth character "friend" can be determined as relatively close historical wake-up words, so as to preliminarily filter out relatively close historical wake-up words before calculating the matching degree, thereby reducing the amount of calculation.

[0160] It can be understood that the "you" in "Hello, little friend" and "You are a big partner" are both the first words, so the two have similar word orders and have relatively close wake-up effects in terms of content. However, the "you" in "Hello, little friend" and "Little friend is you" is in different positions among the five words. These two phrases have different word orders and word structures, and the wake-up effects in terms of content are quite different. Therefore, when screening the historical wake-up words corresponding to each word in the first wake-up word, it is necessary to screen based on the specific position of each word in the first wake-up word to avoid affecting the subsequent matching calculation, so as to accurately screen out relatively close historical wake-up words.

[0161] It can be understood that the first electronic device 1500 can perform a preliminary screening by "selecting historical wake-up words with the same number of characters", and then "select characters with a pinyin length difference less than a preset difference threshold" for a secondary screening, and then "select historical wake-up words with the same single character and corresponding position according to the word order of the first wake-up word" for a third screening, so as to screen out relatively close historical wake-up words before calculating the matching degree through the above triple screening method, thereby reducing the amount of calculation. This effectively improves the efficiency of the first electronic device 1500 in determining the first historical wake-up word.

[0162] S607: Obtain a third score of the first historical wake-up word based on the second wake-up rate and the second response time.

[0163] Exemplarily, the first electronic device 1500 can calculate the corresponding standard score based on the standardized historical wake-up word data, that is, determine the corresponding second wake-up rate and second wake-up time according to the standardized historical wake-up word data, and then determine the third score corresponding to the standardized historical wake-up word data.

[0164] It can be understood that the third score is proportional to the second wake-up rate, and the third score is inversely proportional to the second response time.

[0165] Therefore, based on the association relationship expressed by the above-mentioned formula (1), the second wake-up rate can be substituted into SuccessRate, and the second wake-up time can be substituted into ResponseTime to obtain the third score.

[0166] S608: Determine a second score based on the number of times the first historical wake-up word is adopted by users and the third score, a third weight coefficient corresponding to the third score, and a fourth weight coefficient corresponding to the number of times the first historical wake-up word is adopted by users.

[0167] The third weight coefficient and the fourth weight coefficient are used to characterize the number of times users adopt the solution and the importance of the third score relative to the second score.

[0168] Exemplarily, the first electronic device 1500 may obtain the second score according to the following formula (2):

[0169] StandardScore=α×BaseScore'+β×log 10 (UsageCount+1) Formula (2)

[0170] Among them, α is the third weight coefficient, β is the fourth weight coefficient, which is used to balance the importance of the number of user adoptions and the third score; StandardScore is the second score, BaseScore' is the third score, and UsageCount is the number of user adoptions.

[0171] Furthermore, the first electronic device 1500 can obtain the third score BaseScore' through the following formula (3):

[0172]

[0173] Among them, w 3 is the fifth weight coefficient corresponding to the second wake-up rate, w 4 is the sixth weight coefficient corresponding to the second response time, used to balance the importance of the second wake-up rate and the second response time; ∈ is a constant; BaseScore' is the third score, SuccessRate' is the second wake-up rate, and ResponseTime' is the second response time.

[0174] It can be understood that here ∈ can be used to prevent division by zero errors.

[0175] In some embodiments, the first electronic device 1500 may store each historical wake-up word in association with its corresponding second score to form a standardized score table.

[0176] S609, obtaining a matching degree.

[0177] Exemplarily, the first electronic device 1500 may calculate the matching degree between the first wake-up word and each historical wake-up word, and then determine the first historical wake-up word that meets the matching degree condition. For example, the first electronic device 1500 may determine the first historical wake-up word with the greatest matching degree, and determine the second score corresponding to the first historical wake-up word based on the standardized score table.

[0178] In some embodiments, the first electronic device 1500 can determine the matching degree of the first wake-up word and the historical wake-up word by an edit distance algorithm. The edit distance algorithm may include Levenshtein distance, which can obtain the minimum number of editing operations required to convert one string into another, and the allowed editing operations include: replacing one character with another character (substitutions), inserting a character (insertions), or deleting a character (deletions). The best matching first historical wake-up word obtained in this way is the most similar historical wake-up word matched.

[0179] S610, obtaining a target score.

[0180] Exemplarily, the first electronic device 1500 may add a first natural logarithm to the first score to obtain a target score, wherein the first natural logarithm is a true number obtained by multiplying the matching degree and the second score.

[0181] For example, the first electronic device 1500 may obtain the target score using the following formula (4):

[0182] Score new =BaseScore+ln(StandardScore*M) Formula (4)

[0183] Among them, Score new is the target score, BaseScore is the first score, StandardScore is the second score, and M is the matching degree.

[0184] In some embodiments of the present application, the first electronic device 1500 may be a server 200, and the first electronic device 1500 sends the target score to the second electronic device 1600, which is a terminal device, such as the vehicle-mounted terminal 100 mentioned above; the second electronic device 1600 displays the target score.

[0185] In this regard, the second electronic device 1600 (for example, the vehicle-mounted terminal 100 mentioned above) can obtain the user-defined first wake-up word, and then the second electronic device 1600 sends the user-defined first wake-up word to the first electronic device 1500 (for example, the server 200 mentioned above), so that the first electronic device 1500 executes the above steps S201, S202 and S203.

[0186] For example, the server 200 may send the target score to the vehicle terminal 100 via a wired connection or a wireless connection, so that the vehicle terminal 100 displays the target score. The vehicle terminal 100 can present the target score to the user visually, thereby facilitating the user to intuitively understand the expected wake-up effect of the first wake-up word, and reducing the situation where the user has difficulty in waking up the voice interaction system using the first wake-up word.

[0187] In other embodiments of the present application, the first electronic device 1500 may be a terminal device, and the first electronic device 1500 displays the target score. The terminal device is, for example, the vehicle terminal 100 mentioned above, and the vehicle terminal 100 obtains the user-defined first wake-up word, and the target score may be determined and then directly displayed.

[0188] According to the custom wake-up word detection method provided in the embodiments of the present application, the present application also provides a computer program product, which includes: a computer program code, when the computer program code runs on a computer, enables the computer to implement the steps performed by the first electronic device 1500 in any one of the above embodiments.

[0189] According to the custom wake-up word detection method provided in the embodiments of the present application, the present application also provides a computer-readable medium, which stores a program code. When the program code runs on a computer, the computer implements the steps performed by the first electronic device 1500 in any one of the above embodiments.

[0190] Figure 8 A schematic structural diagram of a first electronic device 1500 provided according to some embodiments of the present application is shown.

[0191] It can be understood that the first electronic device 1500 can be the vehicle-mounted terminal 100 or the server 200 mentioned above.

[0192] like Figure 8 As shown, the first electronic device 1500 includes one or more processors 1501, a system memory 1502, a non-volatile memory (NVM) 1503, a communication interface 1504, an input / output (I / O) device 1505, and a system control logic 1506 for coupling the processor 1501, the system memory 1502, the non-volatile memory 1503, the communication interface 1504 and the input / output (I / O) device 1505. Among them:

[0193] The processor 1501 may include one or more processing units, for example, a data processing unit or processing circuit that may include a central processing unit (CPU), a graphics processing unit (GPU), a digital signal processor (DSP), a microprocessor (MCU), an artificial intelligence (AI) processor or a programmable logic device (field programmable gate array, FPGA), a neural network processor (neural-network processing unit, NPU), etc. may include one or more single-core or multi-core processors. In some embodiments, the processor 1501 can be used to execute instructions to implement the above-mentioned custom wake-up word detection method.

[0194] The system memory 1502 is a volatile memory, such as a random-access memory (RAM), a double data rate synchronous dynamic random access memory (DDR SDRAM), etc. The system memory is used to temporarily store data and / or instructions. For example, in some embodiments, the system memory 1502 can be used to store instructions, and can also be used to store original data objects and changed data objects.

[0195] The non-volatile memory 1503 may include one or more tangible, non-transitory computer-readable media for storing data and / or instructions. In some embodiments, the non-volatile memory 1503 may include any suitable non-volatile memory such as a flash memory and / or any suitable non-volatile storage device, such as a hard disk drive (HDD), a compact disc (CD), a digital versatile disc (DVD), a solid-state drive (SSD), etc. In some embodiments, the non-volatile memory 1503 may also be a removable storage medium, such as a secure digital (SD) memory card, etc. In other embodiments, the non-volatile memory 1503 may be used to store instructions, and may also be used to store original data objects and changed data objects.

[0196] In some embodiments, the system memory 1502 and the non-volatile memory 1503 may respectively include: a temporary copy and a permanent copy of the instruction 1507. The instruction 1507 may include: when executed by at least one of the processors 1501, the first electronic device 1500 implements the custom wake-up word detection method provided in various embodiments of the present application.

[0197] The communication interface 1504 may include a transceiver for providing a wired or wireless communication interface for the first electronic device 1500, so as to communicate with any other suitable device through one or more networks. In some embodiments, the communication interface 1504 may be integrated into other components of the first electronic device 1500, for example, the communication interface 1504 may be integrated into the processor 1501. In some embodiments, the first electronic device 1500 may communicate with other devices through the communication interface 1504, for example, the first electronic device 1500 may establish a communication connection with other devices through the communication interface 1504, so as to send data change requests to other devices, obtain original data objects, and send changed data objects through the communication connection.

[0198] When the first electronic device 1500 is a vehicle-mounted terminal 100, the first electronic device 1500 can establish a communication connection with the server 200 through the communication interface 1504 to send a data change request to the server 200, obtain the original data object and send the changed data object through the communication connection.

[0199] When the first electronic device 1500 is a server 200, the first electronic device 1500 can establish a communication connection with the vehicle terminal 100 through the communication interface 1504 to send a data change request to the vehicle terminal 100, obtain the original data object and send the changed data object through the communication connection.

[0200] The input / output (I / O) device 1505 may include input devices such as a keyboard, a mouse, etc., and output devices such as a display, etc. A user may interact with the first electronic device 1500 via the input / output (I / O) device 1505. For example, a business person may input / select content for data change via the input / output (I / O) device 1505.

[0201] The system control logic 1506 may include any suitable interface controller to provide any suitable interface with other modules of the first electronic device 1500. For example, in some embodiments, the system control logic 1506 may include one or more memory controllers to provide interfaces connected to the system memory 1502 and the non-volatile memory 1503.

[0202] In some embodiments, at least one of the processors 1501 may be packaged together with the logic of one or more controllers for the system control logic 1506 to form a system in package (SiP). In other embodiments, at least one of the processors 1501 may also be integrated with the logic of one or more controllers for the system control logic 1506 on the same chip to form a system-on-chip (SoC).

[0203] Understandably, Figure 8 The structure of the first electronic device 1500 shown is only an example. In other embodiments, the first electronic device 1500 may include more or fewer components than shown in the figure, or combine some components, or split some components, or arrange the components differently. The components shown in the figure may be implemented in hardware, software, or a combination of software and hardware.

[0204] The various embodiments of the mechanism disclosed in the present application can be implemented in hardware, software, firmware or a combination of these implementation methods. The embodiments of the present application can be implemented as a computer module or module code executed on a programmable system, and the programmable system includes at least one processor, a storage system (including volatile and non-volatile memory and / or storage elements), at least one input device and at least one output device.

[0205] The module code may be applied to input instructions to perform the functions described herein and generate output information. The output information may be applied to one or more output devices in a known manner. For purposes of this application, a processing system includes any system having a processor such as, for example, a digital signal processor (DSP), a microcontroller, an application specific integrated circuit (ASIC), or a microprocessor.

[0206] Module code can be implemented with high-level modular language or object-oriented programming language to communicate with the processing system. When necessary, module code can also be implemented with assembly language or machine language. In fact, the mechanism described in this application is not limited to the scope of any specific programming language. In either case, the language can be a compiled language or an interpreted language.

[0207] In some cases, the disclosed embodiments may be implemented in hardware, firmware, software, or any combination thereof. The disclosed embodiments may also be implemented as instructions carried or stored on one or more temporary or non-temporary machine-readable (e.g., computer-readable) storage media, which may be read and executed by one or more processors. For example, instructions may be distributed over a network or through other computer-readable media. Therefore, a machine-readable medium may include any mechanism for storing or transmitting information in a machine (e.g., computer) readable form, including, but not limited to, floppy disks, optical disks, optical disks, read-only memories (CD-ROMs), magneto-optical disks, read-only memories (ROM), random access memories (RAM), erasable programmable read-only memories (EPROM), electrically erasable programmable read-only memories (EEPROM), magnetic or optical cards, flash memory, or a tangible machine-readable memory for transmitting information (e.g., carrier waves, infrared signals, digital signals, etc.) using the Internet in electrical, optical, acoustic, or other forms of propagation signals. Accordingly, machine-readable media include any type of machine-readable media suitable for storing or transmitting electronic instructions or information in a form readable by a machine (eg, a computer).

[0208] In the accompanying drawings, some structural or method features may be shown in a specific arrangement and / or order. However, it should be understood that such a specific arrangement and / or order may not be required. Instead, in some embodiments, these features may be arranged in a manner and / or order different from that shown in the illustrative drawings. In addition, the inclusion of structural or method features in a particular figure does not mean that such features are required in all embodiments, and in some embodiments, these features may not be included or may be combined with other features.

[0209] It should be noted that the units / modules mentioned in the various device embodiments of the present application are all logical units / modules. Physically, a logical unit / module can be a physical unit / module, or a part of a physical unit / module, or can be implemented as a combination of multiple physical units / modules. The physical implementation method of these logical units / modules themselves is not the most important. The combination of functions implemented by these logical units / modules is the key to solving the technical problems proposed by the present application. In addition, in order to highlight the innovative part of the present application, the above-mentioned device embodiments of the present application do not introduce units / modules that are not closely related to solving the technical problems proposed by the present application, which does not mean that there are no other units / modules in the above-mentioned device embodiments.

[0210] The various embodiments of the mechanism disclosed in the present application can be implemented in hardware, software, firmware or a combination of these implementation methods. The embodiments of the present application can be implemented as a computer program or program code executed on a programmable system, which includes at least one processor, a storage system (including volatile and non-volatile memory and / or storage elements), at least one input device and at least one output device.

[0211] Program code can be applied to input instructions to perform the functions described in this application and generate output information. The output information can be applied to one or more output devices in a known manner. For the purposes of this application, a processing system includes any system having a processor such as, for example, a digital signal processor (DSP), a microcontroller, an application specific integrated circuit (ASIC), or a microprocessor.

[0212] Program code can be implemented with high-level programming language or object-oriented programming language to communicate with the processing system. When necessary, program code can also be implemented with assembly language or machine language. In fact, the mechanism described in this application is not limited to the scope of any specific programming language. In either case, the language can be a compiled language or an interpreted language.

[0213] It should be noted that the units / modules mentioned in the various device embodiments of the present application are all logical units / modules. Physically, a logical unit / module can be a physical unit / module, or a part of a physical unit / module, or can be implemented as a combination of multiple physical units / modules. The physical implementation method of these logical units / modules themselves is not the most important. The combination of functions implemented by these logical units / modules is the key to solving the technical problems proposed by the present application. In addition, in order to highlight the innovative part of the present application, the above-mentioned device embodiments of the present application do not introduce units / modules that are not closely related to solving the technical problems proposed by the present application, which does not mean that there are no other units / modules in the above-mentioned device embodiments.

[0214] It should be noted that in the examples and description of this patent, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "including one" do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0215] References to "some embodiments" or "embodiments" in the specification mean that the specific features, structures, or characteristics described in conjunction with the embodiments are included in at least one exemplary implementation or technology disclosed according to the embodiments of the present application. The appearance of the phrase "in some embodiments" in various places in the specification does not necessarily all refer to the same embodiment.

[0216] In addition, the language used in this specification has been primarily selected for readability and instructional purposes and may not be selected to describe or limit the disclosed subject matter. Therefore, the present application embodiment disclosure is intended to illustrate rather than limit the scope of the concepts discussed herein.

Claims

1. A custom wake-up word detection method, characterized in that: Applied to a first electronic device, the method includes: Obtaining a first wake-up word defined by a user, and determining a first score according to a first parameter of the first wake-up word, wherein the first parameter includes a first wake-up rate of the first wake-up word and a first response time of the first wake-up word; Determine a first historical wake-up word among the stored historical wake-up words that satisfies a matching condition with the first wake-up word; A target score for the first wake-up word is determined based on the first score, the second score of the first historical wake-up word, and the degree of match between the first wake-up word and the first historical wake-up word, wherein the second score of the first historical wake-up word is related to the number of times the first historical wake-up word is adopted by users, the second wake-up rate of the first historical wake-up word, and the second response time of the first historical wake-up word.

2. The method according to claim 1, characterized in that The method for obtaining the first parameter of the first wake-up word includes: Generate first audio data according to the first wake-up word; The first audio data is input into a preset wake-up model to perform a preset number of cyclic wake-up tests to obtain the first wake-up rate and the first response time, wherein the first wake-up rate is determined based on the number of successful wake-up times of the preset number of cyclic wake-up tests and the preset number of times, and the first response time is determined based on the average response time of the response time of the preset number of cyclic wake-up tests.

3. The method according to claim 1, characterized in that Determining a first score according to a first parameter of the first wake-up word includes: A first score is determined based on the first awakening rate and the first response time, and a first weight coefficient corresponding to the first awakening rate and a second weight coefficient corresponding to the first response time.

4. The method according to claim 1, characterized in that: The determining a target score of the first wake-up word according to the first score, the second score of the first historical wake-up word, and the matching degree between the first wake-up word and the first historical wake-up word includes: The target score is obtained by the following formula: Score new =BaseScore+ln(StandardScore*M) Among them, Score new is the target score, BaseScore is the first score, StandardScore is the second score, and M is the matching degree.

5. The method according to claim 1, characterized in that The method for obtaining the second score of the first historical wake-up word includes: Obtaining a third score of the first historical wake-up word based on the second wake-up rate and the second response time; The second score is determined based on the number of times the first historical wake-up word is adopted by users and the third score, a third weight coefficient corresponding to the third score, and a fourth weight coefficient corresponding to the number of times the first historical wake-up word is adopted by users.

6. The method according to claim 3, characterized in that: The determining a first score based on the first wake-up rate and the first response time, and a first weight coefficient corresponding to the first wake-up rate and a second weight coefficient corresponding to the first response time, includes: The first score is obtained by the following formula: Among them, w1 is the first weight coefficient, w2 is the second weight coefficient, which is used to balance the importance of the first wake-up rate and the first response time relative to the first score; ∈ is a constant; BaseScore is the first score, SuccessRate is the first wake-up rate, and ResponseTime is the first response time.

7. The method according to claim 5, characterized in that The determining the second score based on the number of user adoptions of the first historical wake-up word and the third score, a third weight coefficient corresponding to the third score, and a fourth weight coefficient corresponding to the number of user adoptions includes: The second score is determined by the following formula: StandardScore=α×BaseScore'+β×log 10 (UsageCount+1) Among them, α is the third weight coefficient, β is the fourth weight coefficient, which is used to balance the importance of the number of user adoptions and the third score; StandardScore is the second score, BaseScore' is the third score, and UsageCount is the number of user adoptions.

8. The method according to claim 5, characterized in that The obtaining a third score of the first historical wake-up word based on the second wake-up rate and the second response time includes: The third score is obtained by the following formula: Among them, w3 is the fifth weight coefficient corresponding to the second wake-up rate, w4 is the sixth weight coefficient corresponding to the second response time, which is used to balance the importance of the second wake-up rate and the second response time; ∈ is a constant; BaseScore' is the third score, SuccessRate' is the second wake-up rate, and ResponseTime' is the second response time.

9. The method according to claim 1, characterized in that: The stored historical wake-up words include standardized historical wake-up word data, wherein the standardized historical wake-up word data at least includes initial consonants, finals, and tones determined according to the pinyin of the historical wake-up words.

10. The method according to claim 1, characterized in that The first electronic device is a server; The method further comprises: The target score is sent to a second electronic device, where the second electronic device is a terminal device.

11. The method according to claim 1, characterized in that: The first electronic device is a terminal device; The method further comprises: The target score is displayed.

12. An electronic device, characterized in that: include: one or more processors; One or more memories; the one or more memories store one or more programs, and when the one or more programs are executed by the one or more processors, the electronic device executes the custom wake-up word detection method described in any one of claims 1 to 11.

13. A computer readable medium, characterized in that The computer-readable medium stores instructions, which, when executed on a machine, enable the machine to execute the custom wake-up word detection method described in any one of claims 1 to 11.

14. A computer program product, characterized in that It includes a computer program / instruction, which, when executed by a processor, implements the custom wake-up word detection method described in any one of claims 1 to 11.

Images