A method and device for generalizing voice wake-up words

Abstract

The present invention relates to the field of speech recognition technology. More specifically, the present invention provides a method and device for generalizing speech wake-up words. By realizing the generalization of speech wake-up words, the user experience and the intelligence level of the system are significantly improved, so that users can interact with the device in a more flexible and diverse way, thereby greatly enhancing the convenience of use. At the same time, the system monitors the user's usage habits and dynamically adjusts the sensitivity of each generalized wake-up word accordingly, effectively optimizing the utilization of system resources, improving the response speed and significantly reducing the occurrence of false wake-ups. In addition, the present invention also gives the system a powerful adaptive ability, enabling it to continuously learn and adjust, automatically update and optimize the generalized wake-up words and their sensitivity according to the user's usage habits and environmental changes, so as to keep in sync with user needs for a long time and provide more personalized and accurate services.

Description

Technical Field

[0001] The present invention relates to the field of speech recognition technology, and more specifically, to a method and device for generalizing a speech wake-up word. Background Art

[0002] Voice Wake-up is a function based on speech recognition technology, which allows users to activate the device, start its voice assistant or perform other operations by issuing specific voice commands (usually a word or phrase). This technology greatly improves the convenience of human-computer interaction, allowing users to interact with the device without touching it manually.

[0003] The voice wake-up system mainly consists of two parts: the acoustic model and the language model. The acoustic model is responsible for converting the input voice signal into voice features. Its main goal is to convert continuous sound signals into discrete symbol sequences for easy computer processing. The role of the acoustic model is to extract useful features from the original voice signal and map it to phonemes or letters; the language model is used to determine whether the pre-set wake-up word appears based on the voice features extracted by the acoustic model. The wake-up word detection is usually implemented through a lightweight neural network, which can recognize specific patterns in the sound in real time to determine whether the user has said the wake-up word.

[0004] The voice systems on the market have a wake-up word by default, and the custom function can be opened to allow users to modify the wake-up word, such as changing "Xiao Ai" to "Xiao Bai" to meet the personalized needs of users. Customized wake-up word technology has been widely used in scenarios such as smart cockpits, smart homes, and smart exhibition halls. The current functional logic is to enter a fixed wake-up word. However, the wake-up habits of different users in different regions vary greatly, and fixed words are difficult to cope with diverse needs, resulting in inconvenience in wake-up. For example: In the use scenario of the smart cockpit, the user customizes the wake-up word to "Xiao Bai", and passenger A says "Hi Xiao Bai", and the wake-up fails. Passenger B says "Xiao Bai Xiao Bai", and the wake-up fails. Passenger C says "Hello Xiao Bai", and the wake-up fails. Passenger D says "Hello Xiao Bai", and the wake-up fails. Passenger E says "Call Xiao Bai", and the wake-up fails.

[0005] Therefore, how to automatically generalize the wake-up word according to the user's region after the wake-up word is customized to cover typical sentences in order to improve the hit rate of wake-up interaction is an urgent problem that needs to be solved in current voice wake-up technology. Summary of the invention

[0006] The object of the present invention is to provide a method and device for generalizing a voice wake-up word to solve the problems raised in the above background technology.

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

[0008] A method for generalizing a speech wake-up word, the method comprising the following steps:

[0009] S1. Start the input interface, enter the custom wake-up word, and confirm the input to trigger the wake-up word generalization function;

[0010] S2. Obtain the user's current location information, search and select the corresponding elimination template library in the system according to the user's current location information, match the template auxiliary words in the elimination template library with the custom wake-up word, eliminate the template auxiliary words in the custom wake-up word, and obtain a keyword;

[0011] S3. Searching for a matching generalization model in the system according to the user's current location information, and adding auxiliary words to the keyword through the generalization model to generate multiple generalized wake-up words;

[0012] S4. Generate a configuration file according to the multiple generalized wake-up words, where the configuration file includes a wake-up audio file corresponding to each generalized wake-up word and an associated sensitivity, and is sent from the server to the client, and stored and called by the client;

[0013] S5. After the user says the wake-up command, check the wake-up command to confirm whether the wake-up command is matched successfully, record the user's habits, and adjust the sensitivity of each generalized wake-up word according to the user's habits;

[0014] Adjusting the sensitivity of each of the generalized wake-up words, specifically monitoring the user's habits, monitoring the frequency of the user using each generalized wake-up word, and adjusting the sensitivity value of each generalized wake-up word according to the frequency;

[0015] Adjusting the sensitivity value of each generalized wake-up word involves the following steps:

[0016] F1. Set the unit time and record the usage frequency of each generalized wake-up word in the unit time;

[0017] F2. After the device is started for the first time within this unit time, the sensitivity value of the generalized wake-up word within this unit time is adjusted accordingly according to the usage frequency of each generalized wake-up word within the previous unit time.

[0018] A further technical solution of the present application is: in the previous unit time, the sensitivity value of any generalized wake-up word with a usage frequency greater than 95% is increased by 2; in the previous unit time, the sensitivity value of any generalized wake-up word with a usage frequency less than 0.5% is reduced by 2; and any generalized wake-up word with a usage frequency of 0 in the previous unit time is deleted.

[0019] A further technical solution of the present application is: if the number of wake-up times of the device in the previous unit time is less than 50 times, the sensitivity value of the generalized wake-up word will not be adjusted in the current unit time.

[0020] A further technical solution of the present application is: the upper limit of the sensitivity value of any wake-up word is set to 90, and it will no longer increase when it reaches the maximum value; the lower limit of the sensitivity value of any wake-up word is set to 50, and it will no longer decrease when it reaches the minimum value.

[0021] A further technical solution of the present application is as follows: the step S5 of checking the wake-up instruction includes the following steps:

[0022] S51, checking whether the wake-up word part issued by the user in the current wake-up instruction is complete. If it is incomplete, automatically filling in the missing wake-up word according to the user habits recorded in the historical time and the context information of the current wake-up instruction, and obtaining the supplemented wake-up word, and recording the supplemented wake-up word as the generalized wake-up word;

[0023] S52, matching the supplemented wake-up word with the configuration file, and if the match is successful, combining the supplemented wake-up word with the context information of the current wake-up instruction to perform a wake-up operation;

[0024] S53: If the match fails, the failure is recorded and the supplemented wake-up word record is deleted.

[0025] The present application also provides a device for generalizing a voice wake-up word, including:

[0026] An input module, used to receive a custom wake-up word input by a user;

[0027] A generalization module is used to generalize a wake-up word into multiple generalized wake-up words;

[0028] A generation module, used to generate a configuration file according to multiple generalized wake-up words;

[0029] A storage module, used to store the configuration file for use in subsequent voice interactions;

[0030] A wake-up module, used for identifying a user's wake-up instruction and checking the wake-up instruction;

[0031] The adjustment module is used to adjust the sensitivity of each generalized wake-up word according to user habits.

[0032] A further technical solution of the present application is that the custom wake-up word is input by text or voice recognition.

[0033] A further technical solution of the present application: the generalization module includes an information acquisition submodule, a positioning submodule, a elimination submodule and a matching submodule;

[0034] The information acquisition submodule is used to receive the custom wake-up word input by the user;

[0035] The positioning submodule is used to obtain the user's current location information;

[0036] The elimination submodule is used to search and select the corresponding elimination template library in the system according to the user's current location information, match the template auxiliary words in the elimination template library with the custom wake-up word, eliminate the template auxiliary words in the custom wake-up word, and obtain a keyword;

[0037] The matching submodule is used to search for a matching generalization model in the system according to the user's current location information, and to add auxiliary words to the keyword through the generalization model to generate multiple generalized wake-up words;

[0038] The adjustment module includes a user habit monitoring submodule and a sensitivity value adjustment submodule.

[0039] The user habit monitoring submodule is used to monitor user habits and record the frequency of use of each generalized wake-up word per unit time;

[0040] The sensitivity value adjustment submodule is used to generate a new sensitivity value according to the usage frequency of the generalized wake-up word and automatically replace the current sensitivity value.

[0041] Compared with the prior art, the technical solution provided by the present invention has the following beneficial effects:

[0042] The present invention significantly improves the user experience and the intelligence level of the system by realizing the generalization of voice wake-up words, allowing users to interact with the device in a more flexible and diverse way, thereby greatly enhancing the convenience of use. At the same time, the system monitors the user's usage habits and dynamically adjusts the sensitivity of each generalized wake-up word accordingly, effectively optimizing the utilization of system resources, improving the response speed and significantly reducing the occurrence of false wake-ups. In addition, the present invention also gives the system a powerful adaptive ability, enabling it to continuously learn and adjust, automatically update and optimize the generalized wake-up words and their sensitivity according to the user's usage habits and environmental changes, thereby maintaining synchronization with user needs for a long time and providing more personalized and accurate services. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 A flow chart of a method for generalizing a voice wake-up word provided by the present invention;

[0044] Figure 2 A modular structure diagram of a device for generalizing voice wake-up words provided by the present invention. DETAILED DESCRIPTION

[0045] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. The present invention is further described below in combination with the embodiments.

[0046] See also Figure 1 , an embodiment of the present application provides a method for generalizing a voice wake-up word, the method comprising the following steps:

[0047] S1. Start the input interface, enter the custom wake-up word, and confirm the input to trigger the wake-up word generalization function;

[0048] S2. Obtain the user's current location information, search and select the corresponding elimination template library in the system according to the user's current location information, match the template auxiliary words in the elimination template library with the custom wake-up word, eliminate the template auxiliary words in the custom wake-up word, and obtain a keyword;

[0049] S3. Searching for a matching generalization model in the system according to the user's current location information, and adding auxiliary words to the keyword through the generalization model to generate multiple generalized wake-up words;

[0050] S4. Generate a configuration file according to the multiple generalized wake-up words, where the configuration file includes a wake-up audio file corresponding to each generalized wake-up word and an associated sensitivity, and is sent from the server to the client, and stored and called by the client;

[0051] S5. After the user says the wake-up command, check the wake-up command to confirm whether the wake-up command is matched successfully, record the user's habits, and adjust the sensitivity of each generalized wake-up word according to the user's habits;

[0052] During use, the user only needs to enter an initial custom wake-up word, which can be entered manually or by voice input. The system can automatically generate multiple related generalized wake-up words, which greatly simplifies the user's setting steps and improves usage efficiency. At the same time, the diversified wake-up words increase the flexibility of user interaction with the device and improve the user experience. The system can dynamically select the corresponding elimination template library and generalization model according to the user's current location information and environment, and generate generalized wake-up words related to the user's context, which reflects a high degree of intelligent processing capabilities. In addition, by recording and analyzing the user's usage habits, the system can continuously adjust the sensitivity of the generalized wake-up words to provide more personalized and accurate services; by monitoring the user's usage frequency, the system can optimize the sensitivity setting of the wake-up words, reduce false wake-ups and unnecessary resource consumption, thereby improving the overall operation efficiency of the system. At the same time, the generated configuration file is sent from the server to the client for unified management and rapid updates;

[0053] This method gives the system powerful adaptive capabilities, enabling it to continuously learn and adjust, automatically updating and optimizing generalized wake-up words and their sensitivity based on user usage habits and environmental changes. This ensures that the system can stay synchronized with user needs for a long time and continue to provide high-quality services.

[0054] Furthermore, in step S5, the sensitivity of each generalized wake-up word is adjusted, specifically by monitoring the user's habits, monitoring the frequency of the user using each generalized wake-up word, and adjusting the sensitivity value of each generalized wake-up word according to the frequency;

[0055] Specifically, the specific steps for adjusting the sensitivity value of each generalized wake-up word include setting a unit time and recording the frequency of each generalized wake-up word being awakened by the user within the unit time, and automatically adjusting the sensitivity value of the wake-up word within the current unit time according to the usage frequency of the previous unit time, which brings many benefits. This dynamic adjustment strategy can optimize the sensitivity setting in real time to make it more in line with the actual usage of the user, thereby improving the response speed and accuracy of the system. At the same time, by effectively managing resources, the system can allocate more resources to wake-up words with high usage frequency to ensure their rapid response and reduce the waste of resources for wake-up words with low usage frequency. This sensitivity adjustment strategy based on usage frequency also simplifies user operations, improves the user experience, and enables the system to have long-term optimization and adaptive capabilities, so that it can always keep pace with user needs and provide more personalized and accurate services.

[0056] Furthermore, adjusting the sensitivity value of each generalized wake-up word includes the following steps:

[0057] F1. Set the unit time and record the usage frequency of each generalized wake-up word in the unit time;

[0058] F2. After the device is started for the first time within this unit time, the sensitivity value of the generalized wake-up word within this unit time is adjusted accordingly according to the usage frequency of each generalized wake-up word within the previous unit time.

[0059] Furthermore, the sensitivity value of any generalized wake-up word with a usage frequency greater than 95% in the previous unit time is increased by 2; the sensitivity value of any generalized wake-up word with a usage frequency less than 0.5% in the previous unit time is reduced by 2; and any generalized wake-up word with a usage frequency of 0 in the previous unit time is deleted.

[0060] Furthermore, if the number of wake-up times of the device in the previous unit time is less than 50 times, the sensitivity value of the generalized wake-up word will not be adjusted in this unit time;

[0061] Furthermore, the upper limit of the sensitivity value of any wake-up word is set to 90, and it will no longer increase when it reaches the maximum value; the lower limit of the sensitivity value of any wake-up word is set to 50, and it will no longer decrease when it reaches the minimum value.

[0062] Specifically, by setting a unit time refinement sensitivity adjustment strategy, including increasing or decreasing the sensitivity values ​​for generalized wake-up words with a usage frequency higher than 95% and lower than 0.5%, respectively, deleting wake-up words with a usage frequency of 0, and not triggering the adjustment function when the device wakes up less frequently, and setting the upper and lower limits of the wake-up word sensitivity value, this series of measures can not only dynamically optimize the sensitivity of the wake-up word according to the actual frequency of use of the user, improve the response speed and accuracy of the system, but also effectively manage resources and avoid resource waste. By deleting unused wake-up words, the redundancy of the system is reduced and the operating efficiency is improved. At the same time, setting the upper and lower limits of the sensitivity value prevents excessive adjustment and ensures the stability of the system and user experience. In addition, when the device wakes up less frequently, the sensitivity adjustment is not triggered, which avoids unnecessary operations and further improves the intelligence of the system and user experience.

[0063] Further, checking the wake-up instruction in step S5 includes the following steps:

[0064] S51, checking whether the wake-up word part issued by the user in the current wake-up instruction is complete. If it is incomplete, automatically filling in the missing wake-up word according to the user habits recorded in the historical time and the context information of the current wake-up instruction, and obtaining the supplemented wake-up word, and recording the supplemented wake-up word as the generalized wake-up word;

[0065] S52, matching the supplemented wake-up word with the configuration file, and if the match is successful, combining the supplemented wake-up word with the context information of the current wake-up instruction to perform a wake-up operation;

[0066] S53. If the match fails, the failure is recorded and the supplemented wake-up word record is deleted;

[0067] Specifically, by checking the integrity of the wake-up word, the system can ensure that the instructions issued by the user are valid and understandable, thereby improving the accuracy and efficiency of the interaction. Secondly, when the wake-up word is incomplete, the system can automatically fill in the missing wake-up word based on the user's historical habits and the contextual information of the current wake-up instruction, which not only simplifies the user's operation, but also improves the user experience. In addition, the supplemented wake-up word is matched with the configuration file, and the wake-up operation is performed when the match is successful, ensuring that the system can accurately respond to the user's instructions. Finally, if the match fails, the system will record the failure and delete the corresponding supplemented wake-up word, which helps the system to continuously learn and optimize, and improve future matching accuracy and user satisfaction. This series of refined steps makes the system's wake-up instruction checking process more intelligent and humane, providing users with a more convenient and efficient interactive experience.

[0068] See also Figure 2 This embodiment provides a device for generalizing a voice wake-up word, including:

[0069] An input module, used to receive a custom wake-up word input by a user;

[0070] A generalization module is used to generalize a wake-up word into multiple generalized wake-up words;

[0071] A generation module, used to generate a configuration file according to multiple generalized wake-up words;

[0072] A storage module, used to store the configuration file for use in subsequent voice interactions;

[0073] A wake-up module, used for identifying a user's wake-up instruction and checking the wake-up instruction;

[0074] The adjustment module is used to adjust the sensitivity of each generalized wake-up word according to user habits.

[0075] Specifically, the input module can be a text input box, allowing the user to manually input a custom wake-up word. In actual application scenarios, the voice recognition technology can also be used to allow the user to input a custom wake-up word by voice. After the input is completed, the generalization module is triggered;

[0076] The generalization module includes information acquisition submodule, positioning submodule, elimination submodule and matching submodule;

[0077] After the user inputs a custom wake-up word, the input module transmits the custom wake-up word to the generalization module; the information acquisition submodule is used to receive the custom wake-up word input by the user;

[0078] The positioning submodule obtains the user's current location information: the user's country, city and other location information can be obtained through GPS positioning, user registration information and other methods. The location accuracy requirement depends on the requirements of the template library and generalization model. It is recommended to use the country as the model evaluation dimension;

[0079] The elimination submodule searches and selects the corresponding elimination template library in the system according to the user's current location information, matches the template auxiliary words in the elimination template library with the custom wake-up word, eliminates the template auxiliary words in the custom wake-up word, and obtains a keyword. For example, if the user is located in China, the "China" elimination template library in the system will be searched. For the custom wake-up word "Hello Xiaobai", according to the template in the elimination template library, "Hello" belongs to the template auxiliary word, and the keyword "Xiaobai" is obtained after elimination. The content of the elimination template library can be manually input or obtained by big data learning. It is recommended to input it manually. The elimination template library can be configured on the client or on the server. It is recommended to configure it on the client.

[0080] The matching submodule is used to search for a matching generalization model in the system according to the user's current location information, and to add auxiliary words to the keywords through the generalization model to generate multiple generalized wake-up words. For example, if the user is located in China, the generalization model corresponding to "China" in the system will be searched, and auxiliary words such as "hi" and "hello" will be added to the keyword "Xiaobai" according to the generalization model, thereby generating multiple generalized wake-up words such as "hi Xiaobai", "hello Xiaobai", "Xiaobai Xiaobai", etc. The generalization model can be manually input or learned from big data. Manual input is recommended. The generalization model can be configured on the client side or on the server side. It is recommended to configure it on the client side. The generalization module passes the obtained multiple generalized wake-up words to the generation module.

[0081] When the generation module is working, it can generate a configuration file based on the generalized wake-up word. The configuration file contains the audio file and the associated sensitivity of each wake-up word. The sensitivity can be manually configured. For example, a total of 3 wake-up words are generalized. The user-entered wake-up word "Hello Xiaobai" has a sensitivity of 80, the generated wake-up word 1 "Hi Xiaobai" has a sensitivity of 70, and the generated wake-up word 2 "Xiaobai Xiaobai" has a sensitivity of 70. The generation model is configured on the server side. After the server generates the configuration file, it is sent to the client. The generation module passes the generated configuration file to the storage module to store the configuration file for use in subsequent voice interactions.

[0082] At this time, the wake-up module receives the wake-up instruction. When the user says a wake-up word, it checks whether the wake-up word part issued by the user in the current wake-up instruction is complete. If it is incomplete, the missing wake-up word is automatically filled in according to the user habits recorded in the historical time and the context information of the current wake-up instruction, and the supplemented wake-up word is obtained, and the supplemented wake-up word is recorded as a generalized wake-up word; the supplemented wake-up word is matched with the configuration file. If the match is successful, the supplemented wake-up word is combined with the context information of the current wake-up instruction to perform the wake-up operation; if the match fails, the failure is recorded, and the supplemented wake-up word record is deleted. The wake-up module matches the audio file in the configuration file to confirm whether the wake-up instruction is matched successfully;

[0083] At the same time, the adjustment module is triggered, which collects user habits and adjusts the sensitivity of each wake-up word.

[0084] Furthermore, the adjustment module includes a user habit monitoring submodule and a sensitivity value adjustment submodule.

[0085] The user habit monitoring submodule is used to monitor user habits and record the frequency of use of each generalized wake-up word per unit time;

[0086] The sensitivity value adjustment submodule is used to generalize the frequency of use of the wake-up word, generate a new sensitivity value, and automatically replace the current sensitivity value. The adjustment module consists of two parts: user habit monitoring and sensitivity adjustment.

[0087] Specifically, record the frequency of each wake-up word being woken up by the user on a weekly basis;

[0088] After the device is started each week, the wake-up word sensitivity value is adjusted based on the wake-up frequency of the previous week. If the total number of device wake-ups in the previous week is less than 50 times, no adjustment will be triggered this week.

[0089] The sensitivity value adjustment rules are as follows: when the weekly usage frequency is greater than 95%, the wake-up word sensitivity value is increased by 2; when the weekly usage frequency is less than 0.5%, the wake-up word sensitivity value is reduced by 2; in other cases, the wake-up word sensitivity value does not change;

[0090] The upper limit of the sensitivity value of each wake-up word is set to 90, and it will no longer increase when it reaches the extreme value; the lower limit of the sensitivity value of each wake-up word is set to 50, and it will no longer decrease when it reaches the extreme value.

[0091] Specifically, the working principle of the sensitivity value adjustment submodule is as follows: assuming that there are three wake-up words, namely "Hello Xiaobai", "Hi Xiaobai", and "Xiaobai Xiaobai", and the sensitivity value of each wake-up word is preset to 80, 70, and 70;

[0092] In the first week, the usage frequencies were 99%, 0.4%, and 0.6%, respectively, so the adjustment results were 82, 68, and 70, respectively; in the second week, the usage frequencies were 99%, 0.3%, and 0.7%, respectively, so the adjustment results were 84, 66, and 70, respectively; and so on.

[0093] Specifically, after the user customizes the wake-up word, a wake-up word configuration file can be generated according to the user input, and the generalization action can be performed synchronously in the background. After the generalization is successful, the configuration file is replaced without affecting the user's use.

[0094] Specifically, the number of generalized templates for each region should be controlled within 20, and the number of wake-up words should be controlled to reduce the false wake-up rate.

[0095] Specifically, after the wake-up words are generalized, the client can add a user confirmation process to prompt the user that multiple wake-up words have been automatically generated and ask whether to enable them, thereby enhancing user perception.

[0096] Specifically, after the wake-up words are generalized, the client can join the user selection process, and the user can manually remove unwanted wake-up words and manually intervene in the wake-up effect.

[0097] The present invention significantly improves the user experience and the intelligence level of the system by realizing the generalization of voice wake-up words, allowing users to interact with the device in a more flexible and diverse way, thereby greatly enhancing the convenience of use. At the same time, the system monitors the user's usage habits and dynamically adjusts the sensitivity of each generalized wake-up word accordingly, effectively optimizing the utilization of system resources, improving the response speed and significantly reducing the occurrence of false wake-ups. In addition, the present invention also gives the system a powerful adaptive ability, enabling it to continuously learn and adjust, automatically update and optimize the generalized wake-up words and their sensitivity according to the user's usage habits and environmental changes, thereby maintaining synchronization with user needs for a long time and providing more personalized and accurate services.

[0098] The present invention and its embodiments are described schematically above, and the description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. Therefore, if a person skilled in the art is inspired by it and designs a structural method and an embodiment similar to the technical solution without creativity without departing from the purpose of the invention, they shall all fall within the protection scope of the present invention.

[0099] In addition, it should be understood that although this specification is described according to implementation methods, not every implementation method only includes independent technical solutions. This narrative method of the specification is only for the sake of clarity. Those skilled in the art should take the specification as a whole. The technical solutions in each embodiment can also be appropriately combined to form other implementation methods that can be understood by those skilled in the art.

Claims

1. A method for generalizing a voice wake-up word, characterized in that: The method comprises the following steps: S1. Start the input interface, enter the custom wake-up word, and confirm the input to trigger the wake-up word generalization function; S2. Obtain the user's current location information, search and select the corresponding elimination template library in the system according to the user's current location information, match the template auxiliary words in the elimination template library with the custom wake-up word, eliminate the template auxiliary words in the custom wake-up word, and obtain a keyword; S3. Searching for a matching generalization model in the system according to the user's current location information, and adding auxiliary words to the keyword through the generalization model to generate multiple generalized wake-up words; S4. Generate a configuration file according to the multiple generalized wake-up words, where the configuration file includes a wake-up audio file corresponding to each generalized wake-up word and an associated sensitivity, and is sent from the server to the client, and stored and called by the client; S5. After the user says the wake-up command, check the wake-up command to confirm whether the wake-up command is matched successfully, record the user's habits, and adjust the sensitivity of each generalized wake-up word according to the user's habits; Adjusting the sensitivity of each of the generalized wake-up words, specifically monitoring the user's habits, monitoring the frequency of the user using each generalized wake-up word, and adjusting the sensitivity value of each generalized wake-up word according to the frequency; Adjusting the sensitivity value of each generalized wake-up word involves the following steps: F1. Set the unit time and record the usage frequency of each generalized wake-up word in the unit time; F2: After the device is started for the first time within this unit time, the sensitivity value of the generalized wake-up word within this unit time is adjusted accordingly according to the usage frequency of each generalized wake-up word within the previous unit time; The step S5 of checking the wake-up instruction comprises the following steps: S51, checking whether the wake-up word part issued by the user in the current wake-up instruction is complete. If it is incomplete, automatically filling in the missing wake-up word according to the user habits recorded in the historical time and the context information of the current wake-up instruction, and obtaining the supplemented wake-up word, and recording the supplemented wake-up word as the generalized wake-up word; S52, matching the supplemented wake-up word with the configuration file, and if the match is successful, combining the supplemented wake-up word with the context information of the current wake-up instruction to perform a wake-up operation; S53: If the match fails, the failure is recorded and the supplemented wake-up word record is deleted.

2. The method for generalizing a speech wake-up word according to claim 1, characterized in that: In the previous unit time, the sensitivity value of any generalized wake-up word with a usage frequency greater than 95% increases by 2; in the previous unit time, the sensitivity value of any generalized wake-up word with a usage frequency less than 0.5% decreases by 2; Delete any generalized wake-up words with a usage frequency of 0 in the previous unit time.

3. The method for generalizing a speech wake-up word according to claim 1, characterized in that: If the device wakes up less than 50 times in the previous unit time, the sensitivity value of the generalized wake-up word will not be adjusted in this unit time.

4. The method for generalizing a speech wake-up word according to claim 2, characterized in that: The upper limit of the sensitivity value of any wake-up word is set to 90, and it will not increase when it reaches the maximum value; the lower limit of the sensitivity value of any wake-up word is set to 50, and it will not decrease when it reaches the minimum value.

5. A device for generalizing a voice wake-up word, characterized in that: include: An input module, used to receive a custom wake-up word input by a user; A generalization module is used to generalize a wake-up word into multiple generalized wake-up words; A generation module, used to generate a configuration file according to multiple generalized wake-up words; A storage module, used to store the configuration file for use in subsequent voice interactions; A wake-up module, used for identifying a user's wake-up instruction and checking the wake-up instruction; An adjustment module is used to adjust the sensitivity of each generalized wake-up word according to user habits; The generalization module includes an information acquisition submodule, a positioning submodule, a elimination submodule and a matching submodule; The information acquisition submodule is used to receive the custom wake-up word input by the user; The positioning submodule is used to obtain the user's current location information; The elimination submodule is used to search and select the corresponding elimination template library in the system according to the user's current location information, match the template auxiliary words in the elimination template library with the custom wake-up word, eliminate the template auxiliary words in the custom wake-up word, and obtain a keyword; The matching submodule is used to search for a matching generalization model in the system according to the user's current location information, and to add auxiliary words to the keyword through the generalization model to generate multiple generalized wake-up words; The adjustment module includes a user habit monitoring submodule and a sensitivity value adjustment submodule. The user habit monitoring submodule is used to monitor user habits and record the frequency of use of each generalized wake-up word per unit time; The sensitivity value adjustment submodule is used to generate a new sensitivity value according to the usage frequency of the generalized wake-up word and automatically replace the current sensitivity value.

6. The device for generalizing a speech wake-up word according to claim 5, characterized in that: The custom wake-up word is text input or voice recognition input.

Images