CN109493849A

CN109493849A - Voice awakening method, device and electronic equipment

Info

Publication number: CN109493849A
Application number: CN201811634115.0A
Authority: CN
Inventors: 李英杰; 阮海鹏
Original assignee: Lenovo Beijing Ltd
Current assignee: Lenovo Beijing Ltd
Priority date: 2018-12-29
Filing date: 2018-12-29
Publication date: 2019-03-19

Abstract

This application provides a kind of voice awakening methods, device and electronic equipment, in the case where electronic equipment is under certain application scenarios and works, in addition to main wake-up word can be utilized, it wakes up outside speech recognition engine, the present embodiment can also meet predetermined condition in the application scenarios of electronic equipment, the the first wake-up word obtained by response, wake up speech recognition engine, realize the voice control to electronic equipment, improve the flexibility and diversity of electronic equipment voice arousal function, and main wake-up word is first inputted relative to traditional scheme, the control instruction that the first wake-up word includes is inputted again, can controlling electronic devices execute the operating process of the control instruction, the present embodiment only needs to input the first wake-up word, simplify voice control step.

Description

Voice awakening method, device and electronic equipment

Technical field

Present application relates generally to fields of communication technology, more particularly to a kind of voice awakening method, device and electricity Sub- equipment.

Background technique

Nowadays, the convenience of the electronic equipments such as mobile phone, notebook and palm PC, Hen Duo electricity are used in order to improve user In sub- equipment have speech recognition capabilities, user can export wake up word come wake up electronic equipment and it includes application, such as A sound Case, user start A speaker, and wake up the speech recognition engine (such as voice assistant) of A speaker it may be said that " your good A ", at this point, with The song of xx " play " is said at family, voice assistant can the automatic identification voice signal phonetic control command that includes, and by holding Row phonetic control command automatically controls the song that A speaker plays xx, is manually operated without user, very convenient.

Applicant have observed that in practical applications, usually requiring user and first inputting main wake-up word, wake up electronic equipment Voice assistant could continue the voice signal that input includes phonetic control command, and very troublesome, it is poor that voice wakes up flexibility.

Apply for content

In view of this, this application provides a kind of voice awakening method, device and electronic equipment, solve the prior art without By in what situations, it is necessary to main wake-up word is inputted, speech recognition engine can be waken up, causes voice wake-up flexibility poor, The technical problem of process complexity.

In order to realize above-mentioned application purpose, this application provides following technical schemes:

This application provides a kind of voice awakening methods, which comprises

It obtains first and wakes up word；

If the application scenarios of electronic equipment meet predetermined condition, respond described first and wake up word, wakes up speech recognition and draw It holds up, the speech recognition engine at least voice for identification；

Wherein, the speech recognition engine can be waken up based on main wake-up word, and the main wake-up word and described first wakes up Word is different.

Optionally, the application scenarios of the electronic equipment are unsatisfactory for predetermined condition, are not responding to described first and wake up word.

Optionally, for the different application scene of the electronic equipment, configured with corresponding different wake-up word；

The speech recognition engine is waken up directly in response to the main wake-up word if obtaining the main wake-up word.

Optionally, the first wake-up word is both to wake up word and control instruction；Or/and

The first wake-up word is operational orders multiple, that multiple first wake-up words are applied for first, and described first applies It is any application with application scenarios corresponding with the multiple first wake-up word.

Optionally, the predetermined condition includes following at least one:

An application in electronic equipment is run under scheduled application scene；

Predetermined device is in running order, predetermined device scheduled application scene for identification.

It is optionally, described to obtain the first wake-up word, comprising:

Receive the collected voice signal of sound collector；

According to the acoustic feature of the voice signal, the similarity of the voice signal with the word sequence prestored is calculated, often The corresponding wake-up word of one group of word sequence, and the word sequence is generated based on the corresponding phoneme for waking up word；

Based on obtained similarity, obtains the voice signal includes first and wake up word.

Optionally, the method also includes:

The control information inputted under the first application scenarios is obtained, and is based on deep learning algorithm, from the control information The word for meeting specified requirements is extracted, and the word extracted is determined as the corresponding wake-up word of first application scenarios；And/or

The control instruction inputted for the first application scenarios of electronic equipment is received, and the control instruction is determined as institute State the corresponding wake-up word of the first application scenarios；And/or

The default control instruction under the first application scenarios of electronic equipment is obtained, and the default control instruction is determined as The corresponding wake-up word of first application scenarios.

Optionally, described to be based on deep learning algorithm, the word for meeting specified requirements, packet are extracted from the control information It includes:

Count the corresponding word frequency of multiple words that the control information includes；

Using deep learning algorithm, the corresponding word frequency of the multiple word is adjusted；

Based on the corresponding word frequency size adjusted of the multiple word, the first biggish word of quantity word frequency is selected.

Present invention also provides a kind of voice Rouser, described device includes:

It wakes up word and obtains module, wake up word for obtaining first；

First wake-up module responds described the in the case where the application scenarios of electronic equipment meet predetermined condition One wakes up word, wakes up speech recognition engine, the speech recognition engine at least voice for identification；

Present invention also provides a kind of electronic equipment, comprising:

Sound collector；

Memory, for storing the program for realizing voice awakening method as described above；

Controller, for loading and executing the program of the memory storage, described program is used for:

It obtains first and wakes up word；

It can be seen that compared with prior art, this application provides a kind of voice awakening method, device and electronic equipment, In the case where electronic equipment is under certain application scenarios and works, in addition to main wake-up word can be utilized, speech recognition engine is waken up Outside, the present embodiment can also meet predetermined condition in the application scenarios of electronic equipment, and the first wake-up word obtained by response is called out Wake up speech recognition engine, realize the flexibility that electronic equipment voice arousal function is improved to the voice control of electronic equipment and Diversity, and main wake-up word is first inputted relative to traditional scheme, then input the control instruction that the first wake-up word includes, it can control Electronic equipment processed executes the operating process of the control instruction, and the present embodiment only needs to input the first wake-up word, simplifies voice control Step processed.

Detailed description of the invention

In order to illustrate the technical solutions in the embodiments of the present application or in the prior art more clearly, to embodiment or will show below There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this The embodiment of application for those of ordinary skill in the art without creative efforts, can also basis The attached drawing of offer obtains other attached drawings.

Fig. 1 a is a kind of existing schematic diagram of a scenario of voice awakening method；

Fig. 1 b is a kind of schematic diagram of a scenario of voice awakening method provided by the embodiments of the present application；

Fig. 2 is a kind of flow diagram of voice awakening method provided by the embodiments of the present application；

Fig. 3 is in a kind of voice awakening method provided by the embodiments of the present application, and the method flow for obtaining the first wake-up word shows It is intended to；

Fig. 4 is a kind of scene flow diagram of voice awakening method provided by the embodiments of the present application；

Fig. 5 is the flow diagram of another voice awakening method provided by the embodiments of the present application；

Fig. 6 is the flow diagram of another voice awakening method provided by the embodiments of the present application；

Fig. 7 is the corresponding wake-up of configuration different application scene in a kind of voice awakening method provided by the embodiments of the present application The process schematic of word；

To configure in another voice awakening method provided by the embodiments of the present application, different application scene is corresponding to call out Fig. 8 The process schematic of awake word；

Fig. 9 is to wake up speech recognition engine in a kind of voice awakening method provided by the embodiments of the present application and switch application The schematic diagram of scene；

Figure 10 is a kind of structural schematic diagram of voice Rouser provided by the embodiments of the present application；

Figure 11 is the structural schematic diagram of another voice Rouser provided by the embodiments of the present application；

Figure 12 is the hardware structural diagram of a kind of electronic equipment provided by the embodiments of the present application.

Specific embodiment

Present applicant have found that: currently on the market as smart home device, smart phone, iPad, notebook electricity In the electronic equipments such as brain, it will usually which voice assistant is set, and (the application can be denoted as master and be called out configured with corresponding wake-up word Awake word), to wake up the voice assistant of electronic equipment.User will use the voice control function of electronic equipment, it is necessary to first say the master Word is waken up, to wake up the voice assistant of electronic equipment, using more troublesome under certain application scenarios of electronic equipment.

For example, schematic diagram of a scenario shown in a referring to Fig.1, after putting out screen under electronic equipment is in music application scene, Electronic equipment still play music, if at this time user wish change broadcasting content, just must first say main wake-ups word (such as hello, A), carry out retriggered electronic equipment and enter working condition, wake up the voice assistant of electronic equipment, later, user, which says, " to play next It is first ", electronic equipment can be identified and be executed, and be switched to next playback of songs, and process is more troublesome.

In order to improve the above problem, applicant is proposed, is worked in the case where electronic equipment is in application scenarios, as shown in Figure 1 b The flow diagram of voice awakening method under music scenario, it is desirable to which, without first saying main wake-up word, such as " hello, A ", and user is direct Say such as " playing next " this voice signal, it will be able to directly wake up the voice assistant of electronic equipment, simultaneously, additionally it is possible to execute The phonetic control command for including in the voice content is such as switched to next playback of songs, improves electronic equipment to reach The flexibility of voice arousal function and multifarious purpose improve the impression that user uses electronic equipment voice assistant function.

Based on above-mentioned design, applicant proposed a kind of voice awakening methods, can be directed to the difference of electronic equipment in advance Application scenarios, such as music scenario, video scene, weather scene, configuring one group wake-up word, (it may include one or more Secondary wake-up word), so in practical applications, works if electronic equipment is under the application scenarios, that is, belong to the application scenarios One application is in operating status, and main wake-up word of the user in addition to being able to use electronic equipment, Ru above-mentioned, " hello, A ", to wake up electricity The voice assistant of sub- equipment；Word can also be waken up with the application scenarios associated time, to wake up electricity directly using preconfigured The voice assistant of sub- equipment, while this time can also be executed and wake up the control instruction that word includes, improve the voice of electronic equipment The flexibility and diversity of arousal function.

Based on the technical solution design that applicant above proposes, below in conjunction with the attached drawing in the embodiment of the present application, to this Technical solution in application embodiment is clearly and completely described, it is clear that described embodiment is only the application one Divide embodiment, instead of all the embodiments.Based on the embodiment in the application, those of ordinary skill in the art are not making Every other embodiment obtained, shall fall in the protection scope of this application under the premise of creative work.

It is a kind of flow diagram of voice awakening method provided by the embodiments of the present application referring to Fig. 2, this method can answer For electronic equipment, which can install speech recognition engine, i.e. voice assistant, enable a user to the electronics Equipment carries out voice control, and the application is not construed as limiting the specific product type of the electronic equipment, as shown in Fig. 2, this method can To include but is not limited to following steps:

Step S11 obtains first and wakes up word；

In the present embodiment practical application, user needs to carry out voice control to electronic equipment, so that electronic equipment is realized A certain functional application.Usually require the voice signal to electronic equipment input corresponding contents.If the voice of electronic equipment is known at this time Other engine (i.e. voice assistant) is not in working condition, possibly can not identify the voice signal, therefore, carry out speech recognition it Before, it usually needs first wake up electronic equipment speech recognition engine.

So the voice signal of user's input may include preconfigured wake-up word, for waking up the language of electronic equipment Sound identifies engine, allows it to carry out speech recognition.

It is to be appreciated that the wake-up of the speech recognition engine for electronic equipment, can respond main wake-up word and realize, that is, It says, in the case where electronic equipment is in various modes, user inputs main wake-up word, and the speech recognition that can directly wake up electronic equipment is drawn It holds up, thus, which can be waken up based on main wake-up word.

The voice awakening method that the present embodiment is proposed, what is mainly described is not that voice knowledge is waken up using main wake-up word Other engine, but be desirable for preconfigured time and wake up word, to wake up the speech recognition engine.Therefore, the of the present embodiment One wake-up word is different from main wake-up word, and the first wake-up word can be time of the different application scene configuration for electronic equipment Wake up word, the application to this first wake up word acquisition modes and its content without limitation.

Step S12 responds the first wake-up word, wakes up voice if the application scenarios of electronic equipment meet predetermined condition Identify engine.

After above description, the first wake-up word that the present embodiment obtains can be time wake-up word, and secondary wake-up word generally is intended to It asks electronic device works under its corresponding application scenarios, can be responded, the speech recognition of electronic equipment is drawn in realization The wake-up held up.

Therefore, after the present embodiment obtains the first wake-up word, the corresponding application scenarios of the first wake-up word can be determined, then test Demonstrate,prove whether application scenarios locating for electronic device works are the corresponding application scenarios of the first wake-up word, in other words, verifying electronics is set It is standby whether to work under the corresponding application scenarios of the first wake-up word, in other words, verifies whether that there are an applications and waken up in first Operation etc. under the corresponding application scenarios of word, whether the application meets predetermined condition to the application scenarios for how verifying electronic equipment Implementation method be not construed as limiting.

In this implementation practical application, it is corresponding to can be the first wake-up word for the application scenarios of electronic equipment in step S12 Application scenarios, at this point, can first determine the application scenarios of the first wake-up word association before executing step S12；Or the electronics The application scenarios of equipment are also possible to the application scenarios where electronic equipment work at present, and the application scenarios in step S12 indicate Content is different, and the content of predetermined condition can be adjusted accordingly, but verifying purpose is identical, i.e. whether verifying electronic equipment is in first It wakes up and works under the corresponding application scenarios of word.

Optionally, in conjunction with above-mentioned analysis, due in practical applications, with the development of communication technology, using (applying journey Sequence APP) function that has is more and more, user can listen to music, be amused oneself by watching other people make fools of themselves by an application, see novel, listen cross-talk, see it is new Hear etc., and during these feature operations, used operational order is often not consistent, therefore, can not be directly by these Function executes scene and is classified as a kind of application scenarios, that is to say, that an application may have plurality of application scenes, and such as above-mentioned Analysis, the present embodiment are then to configure corresponding wake-up word for different application scenarios.Based between the application scenarios and application Relationship, the preconfigured predetermined condition of the present embodiment may include following at least one, but be not limited to be set forth below Content:

An application in electronic equipment is run under the scheduled application scene, that is to say, that this using in operating status, And the application is currently at the scheduled application scene and is running.Wherein, which specifically can be the first wake-up word Corresponding application scenarios, the first wake-up word content is different, which identical can adjust.

Predetermined device in electronic equipment is in running order, which can be used to identify scheduled application scene, Therefore, the present embodiment can in this way, and whether verifying electronic equipment, which is under scheduled application scene, works, the application couple Predetermined type of device under different application scene is without limitation.

To sum up, it works if electronic equipment is under certain application scenarios, in addition to main wake-up word can be utilized, wakes up speech recognition Outside engine, the present embodiment can be after acquisition first wakes up word, and the application scenarios by verifying electronic equipment meet predetermined condition, It works as electronic equipment is under the corresponding application scenarios of the first wake-up word, can also be waken up by responding the first wake-up word Speech recognition engine realizes the flexibility that electronic equipment voice arousal function is improved to the voice control of electronic equipment and more Sample, and main wake-up word is first inputted relative to traditional scheme, then input the control instruction that the first wake-up word includes, it can control Electronic equipment executes the operating process of the control instruction, and the present embodiment only needs to input the first wake-up word, simplifies voice control Step.

In conjunction with the voice awakening method that above-described embodiment describes, the wake-up word of the application can actually be voice signal, During obtaining the first wake-up word in the above-described embodiments, speech recognition technology can use, to identify the voice signal of input, Obtain first and wake up word, the application to the concrete methods of realizing of the speech recognition process without limitation.

Wherein, the process for converting voice signals into computer-readable input, the acoustics that can use voice signal are special Levies in kind is existing, if voice signal (i.e. the voice signal wave of user) identifies, obtain it includes phoneme, relative to being that text is spelled Each letter of sound, such as " next " this voice signal, the phoneme identified may include x, i, a, y, i, sh, ou, group It is xia, yi, shou that conjunction, which obtains phonetic, then thus generates corresponding character/word, under, frighten, become blind, the summer ...；Move, one, also ...；First, It is hand, receipts, thin ...；Later, the word sequence that can be composed: next, Xia Yishou, move down by ..., and then it is each by what is obtained Word sequence is matched with the word sequence for waking up word is prestored, and selects the instruction sentence of the maximum factor sequence structure of matching probability, The the first wake-up word for including for collected voice signal.

As it can be seen that in the application practical application after collecting voice signal, acoustic feature can be carried out to the voice signal It extracts, obtains multiple phonemes that the voice signal includes, later, phoneme dictionary is based on, by obtained multiple phonotactics at extremely A few word sequence calculates each word sequence and prestores the similarity between the word sequence for waking up word, similarity is selected to be greater than The corresponding content of the word sequence selected is determined as the first wake-up word by the word sequence of threshold value.

Based on this, it may include but be not limited to referring to word identification process in voice awakening method shown in Fig. 3, is waken up Following steps:

Step S21 receives the collected voice signal of sound collector；

Step S22, according to the voice signal acoustic feature, it is similar to the word sequence that prestores to calculate the voice signal Degree；

Wherein, the corresponding wake-up word of each group of word sequence, and the word sequence can be raw based on the corresponding phoneme for waking up word At concrete methods of realizing is referred to the description of foregoing embodiments corresponding portion.

Step S23 is obtained voice signal includes first and is waken up word based on obtained similarity.

In the present embodiment, obtained similarity size can be indicated, usual situation by the numerical value for the score being calculated Under, the corresponding score of word sequence is bigger, with prestore wake up word word sequence similarity it is higher, the corresponding content of the word sequence is The probability for waking up word is bigger.

As another alternative embodiment of the application, the application can also utilize speech recognition technology, and training is called out in advance Wake up word identification model, specific training method without limitation, in this way, can directly be inputted wake-up after obtaining voice signal Word identification model, output result can be the voice signal and the matching probability for prestoring wake-up word, so as to it follows that the voice Whether signal includes wake-up word, can also further learn the voice signal includes first content for waking up word, specific implementation Process the application is not detailed.

Optionally, if the application is that the corresponding wake-up word of each application scenarios that prestores is provided with weight, then, it is calculating The word sequence for including to current collected voice signal, and prestores similarity (i.e. initial between the word sequence for waking up word With probability) after, the similarity can also be calculated and prestore the product for waking up the weight of word with this, using obtained numerical value as the voice Signal wakes up the matching probability of word with prestoring, and then judges whether the matching probability is greater than threshold value, uses if it does, determining The voice signal of family input includes to wake up word, and have identified the content that the wake-up word includes.

It based on above-mentioned analysis, in practical applications, can be by after the sound collector of electronic equipment collects voice signal According to above-mentioned voice recognition mode, first is extracted from the voice signal and wakes up word, which can be by the control system of electronic equipment Dedicated control unit MCU is realized in system, and during such as above-mentioned first wakes up word and obtain, and is determining that voice signal includes the While one wake-up word, corresponding control instruction can also be obtained, i.e., after converting voice signals into text, can use this Word content generates the control instruction that electronic equipment can identify, such as binary coding or character string, by the control instruction It is sent to the processor (such as CPU) of electronic equipment, by the processor response control instruction, controlling electronic devices realizes default behaviour Make, is such as switched to next song, inquire weather tomorrow and export etc..

As it can be seen that this speech recognition and processing mode, the processor of electronic equipment is not needed to same voice signal again Speech recognition is carried out, the calculation amount of electronic equipment is reduced, reduces and calculates pressure and power consumption.It improves work efficiency.

Certainly, the above-mentioned control unit MCU of the application, can also only in the manner described above, from collected voice signal, It identifies that it comprises first to wake up word, the processor of electronic device utilizes speech recognition technology, recognition of speech signals packet The control instruction contained, that is, the control instruction for determining that the first wake-up word includes realize predetermined registration operation etc. to respond the control instruction. The application to how obtaining the first wake-up word, and it includes control instruction concrete methods of realizing without limitation.

In addition, in practical applications, the collected voice signal of electronic equipment is other than waking up word, it is also possible to include other Control instruction, in order to realize that the reliable response to the voice signal, the application can carry out semantic analysis to voice signal, with All control instructions for including to the voice signal, and the control instruction is executed, realize predetermined registration operation.It is to be appreciated that the semanteme Analytic process can realize by above-mentioned control unit MCU, and such as during identification wakes up word, obtain that the voice signal includes is each Control instruction can also be analyzed to obtain by the processor CPU of electronic equipment, and the application does not limit this.

Optionally, in conjunction with analysis above, referring to the scene flow diagram of voice awakening method shown in Fig. 4, above-mentioned In the application scenarios verification process of electronic equipment, if the application scenarios of electronic equipment are unsatisfactory for predetermined condition, that is to say, that electronics Equipment is worked under the corresponding application scenarios of the first wake-up word, and such as the first wake-up word is corresponding with music scenario, and electronics is set Standby to be currently operating under weather scene, electronic equipment can't respond the first wake-up word, cannot also utilize the first wake-up word To wake up the speech recognition engine of electronic equipment.

As it can be seen that the voice awakening method that the application proposes, solves multiple wake-up words that same rank is arranged, with main wake-up Word thunder system, the problem of using as the wake-up word for waking up speech recognition engine, be easy to happen false wake-up, as user is listening newly It hears, user has accidentally said " next ", and this wakes up word, and the speech recognition engine of electronic equipment will be waken up, and interrupts current The broadcasting of news, is switched to music scenario, greatly reduces user experience, and the voice awakening method that the application proposes, Even if user has accidentally said " next ", because the application scenarios of electronic equipment work at present are not music scenario, electronics is set Standby will not respond " next " wakes up word, cannot also wake up speech recognition engine, at all cannot be by executing the wake-up word Control instruction, interrupt news play.

Optionally, under the above situation that the present embodiment proposes, corresponding prompt information can also be exported, such as wakes up word not Matching etc. enables a user to re-enter new first according to the prompt information and wake up word, and in the manner described above again Verifying；Certainly, user can also directly input main wake-up word, and to wake up speech recognition engine, the present embodiment is to the output prompt After information, using which kind of wake-up mode, to wake up the speech recognition engine of electronic equipment without limitation.

Based on the technical concept of the voice awakening method proposed above to the application, the application can be for electronic equipment Different application scene, configures corresponding different wake-up word, which is denoted as time wake-up word, be different from above-mentioned by the present embodiment Main wake-up word can use main wake-up word in practical applications, to wake up speech recognition engine, also can use time wake-up word, To wake up the speech recognition engine, distinguishes and be that both wake up the response condition of words difference.

Specifically, the flow diagram of voice awakening method referring to Figure 5 will be direct if obtaining main wake-up word The main wake-up word is responded, the speech recognition engine of electronic equipment is waken up；It, can be according to foregoing embodiments if obtaining time wake-up word The verification mode of description, whether verifying electronic equipment, which is under the corresponding application scenarios of this time wake-up word, works, if so, just meeting It responds this time and wakes up word, wake up the speech recognition engine of electronic equipment.

Wherein, after determining the corresponding target application scene of secondary wake-up word currently obtained, directly in verifying electronic equipment The application of operating status is currently at whether under the target application scene, i.e., it is current with the presence or absence of in target application scene The application of lower operation specifically, the application can first determine the target application for being currently at operating status in electronic equipment, then is tested Demonstrate,prove the target application and whether be under target application scene and run, specifically can by detecting the output information of the target application, Determine whether the target application is in target application scene, however, it is not limited to this.

And for the application with a kind of application scenarios, answer corresponding with the secondary wake-up word currently obtained can be directly verified With whether in operating status, to determine whether the application scenarios of electronic equipment can meet condition etc..The application is to how real Whether the verification method of predetermined condition is now met without limitation to the application scenarios of electronic equipment.

It should be understood that in above-mentioned verification process, if there is currently no applications in operating status in electronic equipment, or Person's application in operating status is not in target application scene, and the present embodiment can consider the application scenarios of electronic equipment not Meet predetermined condition, is not responding to the secondary wake-up word obtained.

It can be seen that main wake-up word is able to use as shown in figure 5, work in the case where electronic equipment is in a certain application scenarios, Or corresponding with the application scenarios wake-up word is realized to wake up the speech recognition engine of electronic equipment to electronic equipment voice Control can only use main wake-up word to wake up the scheme of speech recognition engine, improve the spirit of voice wake-up compared with the existing technology Activity and diversity simplify electronic equipment and are in the step of voice to work under a certain application scenarios wakes up, and pass through application Whether scene meets the verification step of predetermined condition, ensure that the reliability of voice arousal function.

It should be noted that the implementation method of the application wake-up word different to different application scene configuration is not construed as limiting, Several implementations will be enumerated below, but be not limited to implementation enumerated herein.

Optionally, the class that can be applied according to these for the application only run under a kind of application scenarios, the present embodiment Type is divided, and will belong to same type of application corresponding to a kind of application scenarios, and the content of the application scenarios can be based on The type of application determines, such as music application, the weather forecast application of simple version, can determine that music scenario is answered with the music With correspondence, weather scene is corresponding with the weather forecast application.In this case, the corresponding wake-up word of music/weather scene is obtained Afterwards, directly music/the weather forecast application can be verified, to decide whether to respond the wake-up word.

Certainly, in conjunction with the description of above-described embodiment corresponding portion, many applications on the market are at different at present It is run under application scenarios, it, can not be simply by application type, the division of Lai Shixian application scenarios for this kind of application.Such as certain The application such as La Ya, certain dragonfly, can be used to play music, play literary works, play cross-talk, play news etc., it is evident that it plays This few class application content, tend not to be divided into an application scenarios, so, the present embodiment can not be according to the application of application Type directly determines this using locating application scenarios, specifically when whether verifying application scenarios meet predetermined condition, often not It is whether in operating status that the application with the application scenarios can directly be verified, wake up answering for the corresponding application scenarios of word because having It is likely to be at operating status with current, but it is not to run under the corresponding application scenarios of wake-up word, it is seen then that this straight Connecing application of the verifying with application scenarios, whether verification mode in operating status is inaccurate.

It can be seen that the application in electronic equipment may correspond to one or more application scene, in difference The operational order of the application of application scenarios, user's input is often different, and such as music scenario, the operational order of input be can be Next, upper one first, pause, play etc., and in scene of listening to storytelling, the operational order of input tended not to next, on One is first, and input is usually next chapter, above chapter etc., and therefore, in order to guarantee the reliability of voice wake-up, the application is to be directed to Using each application scenarios that can be run, it is pre-configured with and generates corresponding one or more wake-up words, rather than for every One application, configuration generate corresponding one or more wake-up words.Based on above-mentioned analysis, it referring to Fig. 6, is mentioned for the embodiment of the present application The flow diagram of another voice awakening method out, this method can be applied to electronic equipment, can be above-mentioned Fig. 1 pairs A kind of concrete methods of realizing for the embodiment answered, but it is not limited to this implementation method of the present embodiment description, such as Fig. 6 institute Show, this method may include but be not limited to following steps:

Step S31 obtains first and wakes up word；

Step S32 obtains currently running first application of electronic equipment；

Step S33 verifies whether the locating application scenarios of the first application are application scenarios corresponding with the first wake-up word, If not, entering step S34；If so, executing step S35；

Optionally, in conjunction with analysis above, in practical applications, after executing step S31, also it is available with this One wakes up at least one application of the corresponding application scenarios of word, verifies at least one application with the presence or absence of in the applied field The application run under scape, if so, executing step S35；If not, entering step S34.

Wherein, at least one application corresponding with the first wake-up word may be in application corresponding with the first wake-up word It is run under scene, therefore, which can share the first wake-up word, any application at least one application at this It is run under the application scenarios, electronic equipment can respond the first wake-up word, wake up the speech recognition engine of electronic equipment.

Optionally, the present embodiment can be according to the output information of application, to determine that this using locating application scenarios, that is, leads to The type of the current output information of detection application, such as music, weather forecast, joke, literary works, movie and television play are crossed, to determine this Using being currently operating under what application scenarios, but it is not limited to this implementation.

Step S34 is not responding to the first wake-up word；

In this case, electronic equipment may not generate any reaction, or may export the prompt for waking up word mistake Information reminds user to re-enter and wakes up word etc., and the application is not construed as limiting.

Step S35, response first wake up word, wake up speech recognition engine；

Step S36 executes the phonetic control command that the first wake-up word includes, controls the first application execution predetermined registration operation.

In conjunction with above-mentioned analysis, first wakes up word either waking up word, and can be control instruction, that is to say, that this reality It applies the first of example and wakes up word with multiple functions, in the case where meeting certain condition, can both be used to wake up electronic equipment Speech recognition engine, and because it is generated for the application configuration under application scenarios, it can be used to carry out voice to the application Control.

If electronic equipment is under music scenario, the first wake-up word of acquisition is the wake-up word suitable for the music scenario, Such as " next ", after being verified according to manner described above, electronic equipment is able to respond " next " and wakes up word, this When, it can not only directly wake up the speech recognition engine of electronic equipment, additionally it is possible to execute this voice control of broadcasting " next " Instruction, the music player of controlling electronic devices are switched to next music, simplify wake-up speech recognition engine, switch To the operating procedure of next playback of songs, user's body-sensing is improved.

Optionally, word is waken up for the first of the application, quantity can be multiple, that is to say, that answer suitable for same Can have with the wake-up word under scene it is multiple, this it is multiple first wake-up words can be any application belonged under the application scenarios Operational order, i.e. the application can use the operational order respectively applied under an application scenarios, to generate suitable for the applied field The first of scape wakes up word, and the specific generating process of the first wake-up word is without limitation.

It is waken up in conjunction with the description of above-described embodiment corresponding portion in acquisition first as another alternative embodiment of the application After word, whether the application scenarios about verifying electronic equipment meet the implementation method of predetermined condition, it is not limited to verify electronics The mode run under application scenarios is corresponded in the first wake-up word with the presence or absence of an application in equipment, electronic equipment can also be verified Predetermined device it is whether in running order, which can be used for identifying scheduled application scene, i.e. electronic equipment is in It works under the corresponding application scenarios of the first wake-up word obtained, exclusive significant device can be with area by the predetermined device Application scenarios locating for other electronic equipment work at present, the application to the concrete type of the predetermined device without limitation.

As it can be seen that the speech recognition engine of electronic equipment exits in the case where certain predetermined device of electronic equipment is in running order Working condition, user, which can input, wakes up word with the predetermined device corresponding first, reawakes the speech recognition engine, makes it Into working condition, and the predetermined device or its corresponding application, additionally it is possible to execute the voice control that the first wake-up word includes Instruction, realizes corresponding function, not have to input main wake-up word, to wake up speech recognition engine, then input the first wake-up The phonetic control command that word includes, simplifies operating procedure, improves the flexibility of voice wake-up.

Optionally, in the various embodiments described above, for different application scene it is corresponding it is different wake up words, can use with Lower several ways generate to configure, but are not limited to several implementations enumerated herein, it is to be appreciated that following methods The wake-ups word that configuration generates can be the secondary wake-up word for being different from main wake-up word, hereafter no longer intensity explanation.

Mode one: according to system default rule, dividing application scenarios, and obtain the common instructions under each application scenarios, For the wake-up word suitable for the application scenarios.

That is, the application scenarios that the present embodiment can be realized according to electronic equipment, it will be each in electronic equipment Using corresponding application scenarios are allocated to, i.e., what determining electronic equipment included respectively passes through system later using affiliated application scenarios It counts and respectively applies common control instruction under each application scenarios, be determined as the corresponding wake-up word of the application scenarios, but not office It is limited to this implementation.

As it can be seen that this implementation one provided in this embodiment, for same type of electronic equipment, what is configured is applicable in It is essentially identical in the quantity and content of the wake-up word of different application scene, there is certain versatility, but user may be unable to satisfy Individual demand.

Mode two: the universal wake word of difference mode one, it is corresponding that the present embodiment can be customized by the user each application scenarios Wake-up word.Such as in the wake-up word configuration interface of electronic equipment, each application scenarios that can have with output electronic equipment, Yong Huxuan A certain application scenarios are selected, into after waking up word editing mode, can directly select from alternative multiple wake-up words and be suitable for being somebody's turn to do The wake-up word of application scenarios can be edited with oneself and wake up word, meet the individual demand of user.

Based on this, the present embodiment can receive at least one wake-up of any application scenarios input for electronic equipment Word, and establish the corresponding relationship between at least one wake-up word and the application scenarios.

In conjunction with the description for waking up word to first above, it should be appreciated that at least one of the present embodiment input wakes up word can be with It is user's common control instruction under the application scenarios.As it can be seen that user can set according to oneself habit or hobby, setting electronics Under standby each application scenarios, the secondary wake-up word of speech recognition engine not only meets users ' individualized requirement, and user is to this The secondary wake-up word memory that mode generates is stronger, more convenient to use, improves voice and wakes up efficiency.

Mode three: the present embodiment can use artificial intelligence technology, to configure the different application scene for generating electronic equipment Corresponding wake-up word, specifically, word configuration method is waken up referring to shown in fig. 6 time, in the case where obtaining different scenes, user's input After controlling information, deep learning algorithm can use, from the corresponding control information of each application scenarios, extract and meet specified item The word of part is determined as the corresponding wake-up word of the application scenarios to get to different application scene corresponding time and wakes up word.

Wherein, the wake-up word configuration method proposed for mode three, can be using in flow diagram as shown in Figure 8 Processing step, however, it is not limited to this, as shown in figure 8, configuration wakes up in another kind voice awakening method provided in this embodiment The process of word can wrap:

Step S41 obtains the control information inputted under the first application scenarios；

Wherein, the first application scenarios can be any application scenarios that electronic equipment can work, and may be usually based on this The content using current output information in electronic equipment determines that the application is not construed as limiting the type of the first application scenarios.

In the present embodiment practical application, which may include: under the first application scenarios, and user is directed to and is in The control instruction of any application input of first application scenarios operation, under music scenario, user is directed to any voice applications The control instruction inputted during music is played, such as " upper one is first ", " next ", " broadcasting ", " pause ", " increase/reduction sound Amount " etc., the control instruction of input generally includes each operational order that the application has, content of the application to the control information And its acquisition modes are without limitation.

It is to be appreciated that the control information that the present embodiment obtains, can be and carry out semantic point to the voice signal of user's input After analysis, the set of obtained control instruction, rather than the set of voice signal that user directly inputs.

Step S42 counts the corresponding word frequency of multiple words that the control information includes；

It optionally, in practical applications, can be by setting word granularity minimum value and maximum value, using dividing granular manner to unite Count word frequency.Specifically, determining the word granularity for carrying out word frequency statistics in the section being made of the minimum value and maximum value word granularity Length after, the content of text of the control information of acquisition can be segmented according to the length of the word granularity, then count participle The word frequency of obtained each word.This present embodiment to the specific value of the length of the word granularity without limitation.

Certainly, the application also can use existing dictionary, and the text for including to control information segments, to realize word Frequency counts, and how the application is to realizing to the statistical method of the word frequency of control information each word for including without limitation.

Step S43 is adjusted the corresponding word frequency of this multiple word using deep learning algorithm；

Step S44 is based on the corresponding word frequency size adjusted of multiple words, selects the first quantity word frequency biggish Word is determined as the first application scenarios corresponding time and wakes up word；

Optionally, in practical applications, after obtaining control information each word for including and its word frequency, can using a kind of or A variety of deep learning algorithms, are adjusted the word frequency of each word, the location information of the word in the text are such as combined, to beginning of the sentence Corresponding weight is arranged in word in word, sentence, the distribution of sentence tail word, later, by the result of product of the weight and the word frequency of equivalent, as The new word frequency of the word, later, the word frequency obtained to all word granularities are ranked up, and select maximum N number of (the i.e. first number of word frequency Amount) word is as the corresponding time wake-up word of current application scene.

As it can be seen that corresponding quantity for waking up word of each application scenarios that the present embodiment obtains can be N, which be can be Integer greater than 0, the i.e. quantity suitable for the secondary wake-up word of an application scenarios can be one or more.Such as above-mentioned analysis, if It is multiple for belonging to the number of applications of the application scenarios, then, the N number of wake-up word that this multiple application can be shared.

For example: for weather scene, in the manner described above, determine secondary wake-up word may include " tomorrow ", " after It ", and the weather scene can correspond to the weather forecast software in electronic equipment；Music scenario is corresponding time wake up word can be with Including " upper one is first ", " next ", " pause ", " increase/reduction volume " etc., can be set in electronic equipment just like certain Q music Software, certain degree music software, one or more audio softwares such as cruel x music software, in conjunction with the description of voice awakening method above, If any one corresponding time of input music scenario wakes up word, no matter which audio software in electronic equipment be playing music or It is run in music operation interface, this time can be responded and wake up word, realize the voice control to the music software.

It is to be appreciated that the application does not limit the type and its working principle of the deep learning algorithm for realizing word frequency adjustment It is fixed.

Step S45 creates the corresponding relationship for secondary the first application scenarios of wake-up word selected, and carries out to the corresponding relationship Storage.

The present embodiment determines the secondary wake-up under each application scenarios that electronic equipment has according to manner described above After word, word can be waken up to this time and stored, while can also stored between determining secondary wake-up word and the application scenarios Corresponding relationship can be by the secondary wake-up word of the first wake-up word and storage of acquisition so that subsequent actual speech again wakes up in application It is compared, determines the corresponding application scenarios of the first wake-up word, and then verify applied field locating for electronic equipment work at present Whether scape is the corresponding application scenarios of the first wake-up word, that is, whether verifying currently meets the condition that response first wakes up word, Specific verification process is referred to the description of above-described embodiment corresponding portion.

It can be seen that the present embodiment is also possible to control for the secondary wake-up word that application scenarios generate either waking up word Therefore instruction can use the history control information of user's input, analysis obtains the secondary wake-up word suitable for the application scenarios. And the content of the secondary wake-up word of the example above is combined to learn, the secondary wake-up word that the present embodiment determines, which can be, belongs to corresponding applied field The operational order of the application of scape, that is to say, that the present embodiment can be from the operational order that respectively application has in electronic equipment, really Fixed this applies the secondary wake-up word of affiliated application scenarios.

Wherein, it can be embodied in using the operational order having, what the operation interface under the corresponding scene of the application had Operation button or electronic equipment are directed to the mechanically actuated button of application, so, this implements the operation that available each application has The corresponding operational order of button can choose the biggish behaviour of operating frequency by counting to multiple operational orders of acquisition Make instruction and be used as time wake-up word, is i.e. the selection biggish operational order of word frequency (it can be used as a word) is determined as time wake-up word.

Further, for according to one or more modes above, the obtained secondary wake-up suitable for a certain application scenarios In the case that the quantity of word is multiple, this multiple wake-up words that the application can also be carry out classification setting, that is, generate Multiple times of corresponding same application scene wake up the priority of word, wake up word identification to realize according to the priority.

Optionally, it is corresponding for any application scenarios it is multiple time wake up word, the present embodiment can according to secondary wake-ups word and The degree of correlation size of the application scenarios, each wake-up word weight is arranged, the degree of correlation is bigger, and set weight is bigger.Certainly, The application can also wake up the weight of word for customized adjustment each time, such as set customized weight for waking up word of user to most Greatly, then the highest priority etc. of this kind of wake-up word, the application are called out how to be arranged same application scene corresponding multiple times The weight method of awake word is not construed as limiting.

Wherein, the calculation method about the secondary degree of correlation for waking up word and application scenarios can use similarity algorithm to realize, Such as cosine similarity algorithm, the application do not limit this.In practical applications, three time wake-up is determined in the manner described above During word, each word for extracting from control information, it will usually have corresponding score, therefore, be waken up for this kind of time Word, the present embodiment can be by the score sizes, to indicate the corresponding degree of correlation waken up between word and application scenarios, usual score It is bigger, it can indicate that this time wakes up word and the degree of correlation of application scenarios is bigger, set weight is often also bigger, but not office It is limited to the acquisition methods of this degree of correlation and weight, and it is respective to wake up word the application multiple times corresponding to each application scenarios Weighted value is also without limitation.

It should be understood that in practical applications, the secondary priority for waking up word is higher, this time wakes up the identified efficiency of word and get over Height, so, the application would generally wake up word use priority higher time, realize the wake-up to speech recognition engine.

And, it is generally the case that in order to wake up speech recognition engine, realize the voice control to electronic equipment, user's input Wake-up word be usually one, such as main a wake-ups word or secondary wake-up word, usually will not in a voice signal include it is multiple Secondary wake-up word.It but can simultaneously include that a main wake-up word and one time wake up word, such case in a voice signal Under, secondary wake-up word is equivalent to phonetic control command, and electronic equipment can first respond main wake-up word, speech recognition engine is waken up, then It responds this time and wakes up word, controlling electronic devices executes corresponding operating, such as realizes the switching of the application scenarios of electronic equipment.

In the another alternative embodiment of the application, the application corresponding to application scenarios time can wake up word setting it is effective when It is long, once the time that electronic equipment is in continuous work under the application scenarios is more than the effective time, even if user inputs Secondary wake-up word under the application scenarios, electronic equipment also no longer respond this time and wake up word.As it can be seen that in this case, Bu Nengzai Word is waken up to wake up speech recognition engine using this time, is needed user to input main wake-up word, is waken up the speech recognition of electronic equipment Engine.

Based on this, on the basis of the various embodiments described above, the application can also be switched to any applied field to electronic equipment Scape work carries out timing, obtains the operating time of electronic equipment continuous work under the application scenarios, and verify the operating time Whether it is more than preset duration (i.e. above-mentioned effective time), if it does, electronic equipment obtains time wake-up word again, does not need to carry out again Whether application scenarios meet the verifying of predetermined condition, will directly be not responding to this time and wake up word, reduce the calculating pressure of electronic equipment Power.

Optionally, if being multiple for the secondary wake-up word quantity of same application scene, and this multiple times is pre-set and is called out The priority of awake word, and the score size of word is waken up with each time to indicate priority height, specific implementation such as above-described embodiment phase The description of part is answered, long more than in the case where preset duration at work, it is corresponding that the present embodiment can also adjust the application scenarios This multiple time score for waking up words, such as adjust this multiple times weight coefficients for waking up words, waken up to adjust this multiple times The priority of word, the adjustment process can be in the realization of the backstage of electronic equipment, and specific implementation process is without limitation.

Further, the application can also wake up word for each application scenarios corresponding time, and setting enables probability, and the starting Probability can be adjusted dynamically, and the enabling probability of corresponding wake-up word of the application scenarios as where electronic equipment work at present is most Greatly, in this way, after acquisition first wakes up word, it can directly verify whether the corresponding application scenarios of the first wake-up word are that electronics is set Application scenarios locating for standby work at present are avoided in the case where being no to the verification result, accidentally respond the first wake-up word.

Based on this, in the case where above-mentioned determining operating time is more than preset duration, the present embodiment can also reduce electronics The starting probability of corresponding wake-up word of application scenarios locating for equipment work at present, or even the starting probability is adjusted to 0, i.e., The access right that these times are waken up with word is set, these times that limitation electronic equipment response obtains wake up word, so that electronic equipment Need to respond main wake-up word, the arousal function of Lai Shixian speech recognition engine reduces the control program beginning that response time wakes up word It is in running order eventually, to the occupancy of resource, reduce power consumption.

In addition, in the case where electronic equipment carries out application scenarios switching, such as A in the operation under A application scenarios The operation interface of application, user can input the B with B application scenarios application enabled instruction (at this time and input A application It is switched to the switching command run under B application scenarios), in this way, electronic equipment responds the enabled instruction, it can be from A application switching To B application run, and B application can enter B application scenarios under run, complete electronic equipment application scenarios switching after, this The enabling probability that embodiment can wake up word for corresponding to the two application scenarios time carries out dynamic adjustment, i.e. reduction A is answered The enabling probability that word is waken up with scene corresponding time increases the B application scenarios corresponding time enabling probability for waking up word, so that subsequent After obtaining the first wake-up word again, can quickly recognize the first wake-up word is the secondary wake-up word under which application scenarios, is improved Working efficiency.

Description based on the various embodiments described above to voice awakening method, in order to more clearly illustrate the technical side of this method Case will be illustrated by taking specifically used scene as an example below, but be not limited to the usage scenario illustrated herein.

The flow diagram of voice awakening method under music scenario shown in b referring to Fig.1, user input to electronic equipment Main wake-up word, " hello, A ", and after waking up the speech recognition engine of electronic equipment, input " playing music " this voice signal will Start the music software operation of electronic equipment, and music is played by the music software automatically, so that the music software of electronic equipment It works under music scenario.During music, electronic equipment is in static condition for a period of time, often screen locking, If the screen of schematic diagram among Fig. 1 b, electronic equipment is closed, this can not influence the music of electronic equipment, but its voice is known Other engine may have dropped out working condition, if user wishes to carry out voice control to music software, need to reawake electronics and set Standby speech recognition engine.

At this point, if user wishes to play next song, since " next " has been preset as electronic equipment sound Secondary wake-up word under happy scene, so, user can directly say " playing next ", electronic equipment determine therefrom recognize " under One is first " this time wakes up word, and electronic equipment can call out currently still in working under music pattern directly in response to " next " The speech recognition engine of awake electronic equipment, and next control instruction is executed by the speech recognition engine or processor (CPU), it controls Music software processed plays next song.

Wherein, if user's input " changing next, the XXX for playing week XX is sung ", since electronic equipment at this time is still in music Under scene, other than the speech recognition engine that can wake up electronic equipment in the manner described above, additionally it is possible to other control instructions are obtained, And execute, the XXX song of week XX is such as searched, and the song found is played when being switched to next.

Similarly, if user wishes to carry out music software other operations, and the instruction of the operation has been set in advance as this Secondary wake-up word under music scenario, so that it may be handled according to the above-mentioned operating method for being switched to next song, the application No longer it is described in detail one by one.

In addition, work in the case where electronic equipment is in music scenario, if the voice signal of user's input does not include wake-up word, or Person is the wake-up word for including is not the corresponding wake-up word of music scenario, and electronic equipment will not be responded to be identified from voice signal Word is waken up, so as to avoid the corresponding mistake response waken up between word of the different application scene in same rank, improves electricity The voice of sub- equipment wakes up reliability.

In addition, if user needs electronic equipment to be switched to other application scenarios work, usually it is also required to first to wake up electronics and sets Standby speech recognition engine, then the control instruction of switching application scenarios is responded, wake-up of the process to speech recognition engine can be with It is realized using main wake-up word, also can use application scenarios locating for work at present corresponding time and wake up word realization, the application couple This is without limitation.

Referring to Fig. 9, such as during electronic equipment plays music, user's input " you are good A, says a joke to me " is utilized " your good A " this main wake-up word, can wake up the speech recognition engine of electronic equipment, and from " saying a joke to me ", this voice is believed In number, available application scenarios switching command, such as obtaining control voice applications (can be above-mentioned music software, is also possible to Other audio softwares) and enter joke operation interface, the control instruction of joke is played, the control instruction is responded, can control electricity Sub- equipment is switched to joke scene from music scenario, and in this case, user inputs music scenario corresponding time again and wakes up word, nothing Method wakes up speech recognition engine again, needs to input joke scene corresponding time and wakes up word, can wake up speech recognition engine, protect The reliability for having demonstrate,proved voice arousal function, the maloperation bring for avoiding voice control are inconvenient.

Voice wakeup process based on upper segment description, same application may have plurality of application scenes, different application field The corresponding wake-up word of scape is typically different, and thus, same application may correspond to multiple groups and wake up word, each group of wake-up word corresponding one Kind application scenarios.Such as above-mentioned analysis, the present embodiment can be current to determine by the output information of the currently running application of detection Operation using locating application scenarios, that is, determine electronic equipment work at present locating for application scenarios.Based on this, in above-mentioned act In example, after obtaining " next " this wake-up word, the output information of the currently running music software of electronic equipment can detecte Whether it is music information, if so, responding this time wakes up word, wakes up speech recognition engine, and play next song.

Referring to Fig.1 0, it is a kind of structural schematic diagram of voice Rouser provided by the embodiments of the present application, which can be with Applied to electronic equipment, the apparatus may include but be not limited to consisting of structure:

It wakes up word and obtains module 11, wake up word for obtaining first；

Optionally, as shown in figure 11, wake-up word acquisition module 11 may include:

Voice signal receiving unit 111, for receiving the collected voice signal of sound collector；

Similarity calculated 112 calculates the voice signal and pre- for the acoustic feature according to the voice signal The similarity for the word sequence deposited, the corresponding wake-up word of each group of word sequence, and the word sequence is based on the corresponding sound for waking up word Element generates；

Word obtaining unit 113 is waken up, for the voice signal includes first being obtained and being waken up based on obtained similarity Word.

First wake-up module 12, in the case where the application scenarios of electronic equipment meet predetermined condition, described in response First wakes up word, wakes up speech recognition engine, the speech recognition engine at least voice for identification；

Wherein, the speech recognition engine can be waken up based on main wake-up word, and the main wake-up word and described first wakes up Word is different, which can be the secondary wake-up word for application scenarios configuration, that is to say, that not for electronic equipment Same application scenarios can configure corresponding different wake-up words, and the description as described in main wake-up word and time wake-up word is referred to above The record of embodiment of the method corresponding portion.

In the application, above-mentioned first wake-up word is both to wake up word and control instruction；Or/and first wake up word be it is multiple, Multiple first wake-up words can be the operational order of the first application, which is to have to wake up word pair with the multiple first Any application for the application scenarios answered, the operation which can be the first application and have under the application scenarios refer to Enable, the application to first wake-up word include content without limitation.

In practical applications, above-mentioned first wake-up module 12 can be also used for being unsatisfactory in the application scenarios of electronic equipment pre- In the case where fixed condition, it is not responding to the first wake-up word.

Wherein, which may include following at least one:

The verification process for whether meeting predetermined condition about the application scenarios of electronic equipment is referred to above method implementation The description of example corresponding portion.

Optionally, as shown in figure 11, which can also include:

Second wake-up module 13, for directly in response to the main wake-up word, waking up institute in the case where obtaining main wake-up word State speech recognition engine.On the basis of the various embodiments described above, in order to obtain the corresponding wake-up word of each application scenarios, the dress Setting to include:

First configuration module for obtaining the control information inputted under the first application scenarios, and is based on deep learning algorithm, The word for meeting specified requirements is extracted from the control information, and the word extracted is determined as first application scenarios and is corresponded to Wake-up word；And/or

Second configuration module, for receiving the control instruction for being directed to the first application scenarios of electronic equipment and inputting, and by institute It states control instruction and is determined as the corresponding wake-up word of first application scenarios；And/or

Third configuration module, the default control instruction under the first application scenarios for obtaining electronic equipment, and will be described Default control instruction is determined as the corresponding wake-up word of first application scenarios.

Optionally, above-mentioned first configuration module may include:

Word frequency statistics unit, the corresponding word frequency of multiple words for including for counting the control information；

Adjustment unit is adjusted the corresponding word frequency of the multiple word for utilizing deep learning algorithm；

Selecting unit selects the first quantity for being based on the corresponding word frequency size adjusted of the multiple word The biggish word of word frequency；

Determination unit, for the word selected to be determined as the corresponding wake-up word of the first application scenarios.

About above-mentioned several processes for configuring the corresponding different wake-up words of different application scene, it is referred to above-described embodiment The description of corresponding portion.The embodiment of the present application also provides a kind of storage mediums, are stored thereon with program, and the program is by processor Or controller executes, and realizes each step of above-mentioned voice awakening method.

The embodiment of the present application also provides a kind of controllers to realize above-mentioned for loading and executing the program in memory Each step of voice awakening method.

Optionally, which can be with the special equipment that realization voice wakes up in electronic equipment, such as micro-control unit (Microcontroller Unit, MCU), i.e., analyze collected voice signal by micro-control unit MCU, such as ties It closes acoustic model to be analyzed, identifies whether the voice signal includes wake-up word, can also further generate corresponding control and refer to It enables, the processor CPU being sent in electronic equipment executes control instruction, realizes predetermined registration operation.

Alternatively, micro-control unit MCU is only used to whether recognition of speech signals includes wake-up word, know about from voice signal Process that is other and generating control instruction, can be realized, the application is not limited this by the processor CPU in electronic equipment.Ginseng It is the hardware structural diagram of a kind of electronic equipment provided by the embodiments of the present application, which may include sound according to Figure 12 Sound collector 21, memory 22 and controller 23, in which:

Sound collector 21 can be used to carry out sound collection, obtain voice signal, and the voice signal is sent to control Device 23 processed carries out voice recognition processing.The application to the quantity and type of the sound collector 21 without limitation.

Optionally, in order to improve the reliability of sound collection, reduce speech recognition workload, the application can use multiple Sound collector 21 is realized, such as the microphone array that multiple microphones are constituted carries out sound collection, implements process sheet Embodiment is not detailed.

Memory 21 can be used to store the program for realizing above-mentioned voice awakening method；

In the present embodiment practical application, which can also be used to store preconfigured wake-up word, i.e., this is called out The corresponding relationship of awake word and each application scenarios, the corresponding application program of speech recognition engine can also be used to store as needed The intermediate data etc. generated in voice wakeup process.

Optionally, which can be the non-volatile memory in computer-readable medium, random access memory (RAM) and/or the forms such as Nonvolatile memory, such as read-only memory (ROM) or flash memory (flash RAM), memory includes extremely A few storage chip.

Controller 22 can be used for loading and executing the program of the memory storage, which is mainly used for:

It obtains first and wakes up word；

Each step that program realizes voice wake-up is executed about controller 22, it is corresponding to be referred to above method embodiment Partial description.

Optionally, the controller 22 of the present embodiment can be microcontroller, and in this case, which may be used also To include processor, such as CPU, for realizing the voice control to electronic equipment.

In addition, the hardware about electronic equipment forms, it is not limited to which several parts that foregoing embodiments are enumerated can also wrap Such as display, player output equipment are included, can be determined according to the product type of electronic equipment, the application is no longer detailed one by one It states.

To sum up, electronic equipment provided in this embodiment can use main wake-up word, wake up speech recognition engine at any time, hold The phonetic control command that row subsequent acquisition arrives；And work in the case where electronic equipment is in certain application scenarios, the application can also be utilized Corresponding wake-up word of scene, wakes up speech recognition engine, it is not limited to which main wake-up word improves the flexibility of voice wake-up And diversity.

It is and since secondary wake-up word only can just be responded when electronic equipment is under corresponding application scenarios and works, i.e., secondary to call out Word of waking up is lower than the main rank for waking up word, so as to avoid setting and the main multiple wake-up words for waking up the same rank of word, it is easy to make Occur at the case where false wake-up, improves the reliability of voice wake-up.

There are the different application scenarios of embodiment disclosed in the present application different first to wake up word (that is, secondary wake-up word). The first wake-up word is both to wake up word and control instruction, that is, the first wake-up word under each application scenarios is exactly the application The meaning indicated by the operational controls of application program under scene, for example, music scenario, at least one sound corresponding to music scenario Happy application program (Netease's music), the music application have multiple manipulation controls (for example, the icon of triangle identifies meaning What is shown is to play, and is pause indicated by the icon mark of two vertical bars).It is each the first wake-up word under each application scenarios It is exactly control instruction corresponding to the operational controls applied under the application scenarios.For example, music scenario, corresponding to music scenario At least one music application (Netease's music), triangle icon mark if by user triggering if generate play instruction with It controls one of song and carries out output broadcasting).So the first of the music scenario wakes up word exactly " broadcasting ", " pause " etc. Deng.

Finally, it should be noted that about in the various embodiments described above, such as first, second or the like relational terms are only Only it is used to an operation, unit or module and another is operated, unit or module distinguish, and not necessarily requires or secretly Show that there are any actual relationship or orders between these units, operation or module.Moreover, term " includes ", " packet Containing " or any other variant thereof is intended to cover non-exclusive inclusion, so that including the process, method of a series of elements Or system not only includes those elements, but also including other elements that are not explicitly listed, or it is this for further including Process, method or the intrinsic element of system.In the absence of more restrictions, being limited by sentence "including a ..." Element, it is not excluded that include the element process, method or system in there is also other identical elements.

Each embodiment in this specification is described in a progressive manner, the highlights of each of the examples are with other The difference of embodiment, the same or similar parts in each embodiment may refer to each other.For device disclosed in embodiment, For electronic equipment, since it is corresponding with method disclosed in embodiment, so being described relatively simple, related place is referring to method Part illustrates.

The foregoing description of the disclosed embodiments makes professional and technical personnel in the field can be realized or use the application. Various modifications to these embodiments will be readily apparent to those skilled in the art, as defined herein General Principle can be realized in other embodiments without departing from the spirit or scope of the application.Therefore, the application It is not intended to be limited to the embodiments shown herein, and is to fit to and the principles and novel features disclosed herein phase one The widest scope of cause.

Claims

1. a kind of voice awakening method, which comprises

It obtains first and wakes up word；

If the application scenarios of electronic equipment meet predetermined condition, responds described first and wake up word, wake up speech recognition engine, institute State speech recognition engine at least voice for identification；

Wherein, the speech recognition engine can be waken up based on main wake-up word, and the main wake-up word and described first wakes up word not Together.

2. being not responding to described according to the method described in claim 1, the application scenarios of the electronic equipment are unsatisfactory for predetermined condition First wakes up word.

3. being configured with corresponding difference according to the method described in claim 1, being directed to the different application scene of the electronic equipment Wake-up word；

4. according to the method described in claim 1, the first wake-up word is both to wake up word and control instruction；Or/and

It is described first wake up word be it is multiple, it is multiple first wake up words be first application operational order, it is described first application be tool There is any application of application scenarios corresponding with the multiple first wake-up word.

5. according to the method described in claim 1, the predetermined condition includes following at least one:

6. according to the method described in claim 1, described obtain the first wake-up word, comprising:

Receive the collected voice signal of sound collector；

According to the acoustic feature of the voice signal, the similarity of the voice signal with the word sequence prestored is calculated, each group Word sequence corresponds to a wake-up word, and the word sequence is generated based on the corresponding phoneme for waking up word；

7. method described in any one according to claim 1~6, the method also includes:

The control information inputted under the first application scenarios is obtained, and is based on deep learning algorithm, is extracted from the control information Meet the word of specified requirements, and the word extracted is determined as the corresponding wake-up word of first application scenarios；And/or

The control instruction inputted for the first application scenarios of electronic equipment is received, and the control instruction is determined as described the The corresponding wake-up word of one application scenarios；And/or

The default control instruction under the first application scenarios of electronic equipment is obtained, and described in the default control instruction is determined as The corresponding wake-up word of first application scenarios.

8. extracting satisfaction from the control information according to the method described in claim 7, described be based on deep learning algorithm and referring to The word of fixed condition, comprising:

9. a kind of voice Rouser, described device include:

It wakes up word and obtains module, wake up word for obtaining first；

First wake-up module, in the case where the application scenarios of electronic equipment meet predetermined condition, response described first to be called out It wakes up word, wakes up speech recognition engine, the speech recognition engine at least voice for identification；

10. a kind of electronic equipment, comprising:

Sound collector；

Memory, for storing the program for realizing voice awakening method as described in any one of claims 1 to 8；

It obtains first and wakes up word；