CN108899028A - Voice awakening method, searching method, device and terminal - Google Patents
Voice awakening method, searching method, device and terminal Download PDFInfo
- Publication number
- CN108899028A CN108899028A CN201810587174.0A CN201810587174A CN108899028A CN 108899028 A CN108899028 A CN 108899028A CN 201810587174 A CN201810587174 A CN 201810587174A CN 108899028 A CN108899028 A CN 108899028A
- Authority
- CN
- China
- Prior art keywords
- wake
- word
- preset
- voice
- speech recognition
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 56
- 230000002618 waking effect Effects 0.000 claims abstract description 79
- 238000004458 analytical method Methods 0.000 claims description 12
- 230000008569 process Effects 0.000 claims description 12
- 238000004590 computer program Methods 0.000 claims description 11
- 230000008859 change Effects 0.000 claims description 2
- 238000005457 optimization Methods 0.000 claims 2
- 238000010586 diagram Methods 0.000 description 6
- 238000005265 energy consumption Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 230000006870 function Effects 0.000 description 4
- 230000004044 response Effects 0.000 description 4
- 238000000605 extraction Methods 0.000 description 3
- 230000003993 interaction Effects 0.000 description 3
- 238000004378 air conditioning Methods 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 235000013399 edible fruits Nutrition 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 230000037007 arousal Effects 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- NGVDGCNFYWLIFO-UHFFFAOYSA-N pyridoxal 5'-phosphate Chemical compound CC1=NC=C(COP(O)(O)=O)C(C=O)=C1O NGVDGCNFYWLIFO-UHFFFAOYSA-N 0.000 description 1
- 238000012163 sequencing technique Methods 0.000 description 1
- 230000005236 sound signal Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 230000001360 synchronised effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L15/00—Speech recognition
- G10L15/26—Speech to text systems
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L15/00—Speech recognition
- G10L15/22—Procedures used during a speech recognition process, e.g. man-machine dialogue
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L15/00—Speech recognition
- G10L15/22—Procedures used during a speech recognition process, e.g. man-machine dialogue
- G10L2015/223—Execution procedure of a spoken command
Landscapes
- Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Human Computer Interaction (AREA)
- Physics & Mathematics (AREA)
- Acoustics & Sound (AREA)
- Multimedia (AREA)
- Electric Clocks (AREA)
Abstract
The present invention relates to a kind of voice awakening methods, including step；Obtain the first voice signal；Wake-up identification is carried out to the first voice signal, obtains waking up recognition result；When wake-up recognition result is matched with preset wake-up word, speech recognition module is waken up, and speech recognition module will be sent to the matched preset wake-up word of recognition result is waken up.A kind of voice search method and a kind of voice recognition terminal are also disclosed.Wake-up identification is carried out to the first voice signal by waking up identification device 34, and after waking up speech recognition equipment 36, matched preset wake-up word is sent to speech recognition equipment 36, so that speech recognition equipment 36 can obtain corresponding to the preset path optimizing network for waking up word directly according to information progress sweep forward is waken up.Realizing that speech recognition module wakes up is synchronization gain input signal, and then obtains path optimizing network, so can be convenient realize user needed for control operate, substantially increase voice and wake up efficiency with control.
Description
Technical field
The present invention relates to technical field of voice recognition, more particularly to a kind of voice awakening method, searching method, device and
Terminal.
Background technique
With the continuous innovation of information technology, various smart machines are also quickly updating.As numerous smart machines
One of hot technology speech recognition technology, be a kind of Typical Representative of data information application.Speech recognition technology can
By given speech recognition at corresponding text information, therefore, it is widely used in various intelligent interaction devices, such as intelligently
The functions such as the voice arousal function of interactive device and voice assistant.
Traditional voice wake-up mode is identified generally by one wake-up configured with small-sized speech recognition network of setting
Module directly uses built-in speech recognition module, when user says the voice for waking up word, recognizes and prestore wake-up
The matched wake-up word of word then wakes up the corresponding function of intelligent interaction device, such as wakes up the intelligent interaction device of standby mode, with
Standby user manipulation.However, inventor in the implementation of the present invention, it is found that traditional voice wake-up mode still has wake-up
The lower problem of efficiency.
Summary of the invention
Based on this, it is necessary to for the above problem existing for traditional voice wake-up mode, provide a kind of voice wake-up side
Method, a kind of voice search method, a kind of voice Rouser, a kind of voice searching device and a kind of language identification terminal.
To achieve the above object, the embodiment of the present invention uses following technical scheme：
On the one hand, the embodiment of the present invention provides a kind of voice awakening method, includes the following steps；
Obtain the first voice signal；
Wake-up identification is carried out to first voice signal, obtains waking up recognition result；
When the wake-up recognition result is matched with preset wake-up word, speech recognition module is waken up, and will be with the wake-up
The matched preset wake-up word of recognition result is sent to the speech recognition module.
In one of the embodiments, the voice will be sent to the matched preset wake-up word of the wake-up recognition result
The process of identification module further includes：
First voice signal is sent to the speech recognition module.
The wake-up recognition result and the matched judgement of preset wake-up word in one of the embodiments, including：
The wake-up recognition result of word sequence form and the preset wake-up word of word sequence form match point
Analysis；
Or, the preset wake-up word of the wake-up recognition result of characteristic sequence form and characteristic sequence form is carried out
The matching analysis.
On the other hand, a kind of voice search method is also provided, is included the following steps：
It receives and wakes up the preset wake-up word that identification module is sent；Wherein, the preset wake-up word and the wake-up identify mould
The wake-up recognition result that block obtains matches；
Sweep forward is carried out according to the preset wake-up word, obtains corresponding to the preset path optimizing network for waking up word.
Sweep forward is carried out according to the preset wake-up word in one of the embodiments, obtains corresponding to described preset call out
After the step of path optimizing network of awake word, including：
First voice signal that receives of wake-up identification module is obtained, according to first voice signal and described excellent
Change path network and carry out sweep forward, obtains recognition result.
Sweep forward is carried out according to the preset wake-up word in one of the embodiments, obtains corresponding to described preset call out
After the step of path optimizing network of awake word, further include：
The second voice signal is obtained, sweep forward is carried out according to second voice signal and the path optimizing network,
Obtain the recognition result.
Sweep forward is carried out according to the preset wake-up word in one of the embodiments, obtains corresponding to described preset call out
After the step of path optimizing network of awake word, further include：
Obtain the second voice signal；
Sweep forward is carried out according to first voice signal, the path optimizing network and second voice signal,
Obtain the recognition result.
Another aspect also provides a kind of voice Rouser, including：
First signal acquisition module, for obtaining the first voice signal；
Wake-up module obtains waking up recognition result for carrying out wake-up identification to first voice signal；It is called out described
When recognition result of waking up is matched with preset wake-up word, speech recognition module is waken up, and will be matched pre- with the wake-up recognition result
It sets wake-up word and is sent to the speech recognition module.
In another aspect, a kind of voice searching device is also provided, including：
Receiving module, for receiving the preset wake-up word for waking up identification module and sending；Wherein, the preset wake-up word and institute
The wake-up recognition result that wake-up identification module obtains is stated to match；
Search module obtains corresponding to the preset wake-up word for carrying out sweep forward according to the preset wake-up word
Path optimizing network.
In another aspect, also providing a kind of computer readable storage medium, it is stored thereon with computer program, feature exists
In being realized when the computer program is executed by processor the step of the voice awakening method and/or the voice searched
The step of Suo Fangfa.
In another aspect, also providing a kind of voice recognition terminal, including voice signal receiver, wake-up identification device and voice
Identification device, the voice signal receiver is electrically connected the wake-up identification device and the speech recognition equipment, described
It wakes up identification device and is electrically connected the speech recognition equipment；
After the voice signal receiver receives the first voice signal, it is sent to the wake-up identification device；It is described to call out
Identification device of waking up carries out wake-up identification to first voice signal, obtains waking up recognition result, and identify and tie in the wake-up
When fruit matches with preset wake-up word, the speech recognition module is waken up, and send and the wake-up to the speech recognition module
The matched preset wake-up word of recognition result；
The speech recognition equipment carries out sweep forward according to the preset wake-up word, obtains corresponding to the preset wake-up word
Path optimizing network.
A technical solution in above-mentioned technical proposal has the advantages that：
Wake-up identification is carried out to the first voice signal by waking up identification module, and after waking up speech recognition module, it will
Preset wake-up word is sent to speech recognition module so that speech recognition module can directly be carried out according to preset wake-up word before to searching
Rope obtains path optimizing network.It inputs special wake-up word sound in advance without user, then inputs the instruction for needing equipment to execute
Voice, realizing that speech recognition module wakes up is synchronization gain input signal, and then obtains search result namely path optimizing net
Network, and then control needed for search result facilitates realization user can be directly based upon and operated.It reduces equipment energy consumption simultaneously, improves language
The response speed of sound control substantially increases the efficiency that voice wakes up with controls.
Detailed description of the invention
Fig. 1 is the applied environment figure of voice awakening method in one embodiment；
Fig. 2 is the flow diagram of voice awakening method in one embodiment；
Fig. 3 is the flow diagram of voice search method in one embodiment；
Fig. 4 is the structural block diagram of voice Rouser in one embodiment；
Fig. 5 is the structural block diagram of voice searching device in one embodiment；
Fig. 6 is the structural block diagram of voice recognition terminal in one embodiment；
Fig. 7 is the application exemplary diagram of voice recognition terminal in another embodiment.
Specific embodiment
It is with reference to the accompanying drawings and embodiments, right in order to which the objects, technical solutions and advantages of the application are more clearly understood
The application is further elaborated.It should be appreciated that specific embodiment described herein is only to explain the application, and do not have to
In restriction the application.
Voice awakening method provided by the present application and voice search method, can be applied to application environment as shown in Figure 1
In.Wherein, terminal 102 passes through data network or wireless network connection to external communication network, or works offline.In terminal 102
It can be set and waken up identification module 11 and speech recognition module 13, after wake-up identification module 11 gets the first voice signal,
Such as the first voice signal that user speech input or other terminals send over, it wakes up identification module 11 and first voice is believed
Number wake-up identification is carried out, obtains waking up recognition result.It is identified when waking up the wake-up recognition result that identification module 11 obtains with wake-up
When being pre-configured with any one matching in multiple preset wake-up words in module 11, speech recognition module 13 is waken up, and will be aforementioned
The preset wake-up word being matched to is sent to speech recognition module 13.In turn, speech recognition module 13 can be obtained according to when waking up
Preset wake-up word carry out sweep forward, obtain corresponding to the preset path optimizing network for waking up word.Wherein, terminal 102 can with but
It is whole to be not limited to various smart phones, personal computer, laptop, tablet computer, intelligent appliance equipment and vehicle intelligent
End equipment.
In one embodiment, as shown in Fig. 2, providing a kind of voice awakening method, it is applied in Fig. 1 in this way
For terminal 102, it is illustrated from the angle for waking up identification module 11.It is appreciated that above-mentioned speech recognition module 13 can be with
It is arranged in a terminal 102 respectively as two hardware modules or discrete is arranged at two not with identification module 11 is waken up
In same terminal 102, it can also be separately positioned in server and terminal 102, be not construed as limiting in this specification.Wake up identification mould
Block 11 and speech recognition module 13 are also possible to two software function modules.Waking up identification module 11 can also be with speech recognition mould
13 integrated setting of block, such as in the chip circuit module of the same voice recognition processing, it is conventional by embedded development etc.
Means, setting wake up identification module 11, so as to by waking up the triggering of identification module 11 starting or waking up entire speech recognition
Module 13.It should be noted that waking up identification module 11 can be also used for conventional 102 wake operation of terminal, for example, wake up to
The terminal 102 of machine state does not do expansion explanation in the embodiment of this specification.
The awakening method of above-mentioned speech recognition can specifically include following steps S12 to S16：
S12 obtains the first voice signal.
It is appreciated that terminal 102, in standby mode or normal course of operation, speech recognition module is typically at pass
Closed state or sleep state, at this point it is possible to receive first at any time by the lower operation for waking up identification module 11 of power consumption
Voice signal.First voice signal is, for example, one section of voice that user says, and is also possible to the audio that other terminals send over
Signal.
S14 carries out wake-up identification to the first voice signal, obtains waking up recognition result.
It is provided with small-sized speech recognition network, such as common WFST network it is appreciated that waking up in identification module 11, it can
To carry out quickly identification output to preset multiple preset wake-up words, so as to from the first voice signal of input,
Quickly recognize wherein whether containing any one setting preset wake-up word.Waking up recognition result can be corresponding to the first voice
The word sequence of signal is also possible to the corresponding characteristic sequence of the first voice signal, such as MFCC feature extraction obtains characteristic sequence
(such as feature vector).
Specifically, after waking up the first voice signal of acquisition of identification module 11, to the first voice signal in the knowledge constructed in advance
It is scanned in other network, completes the wake-up to the first voice signal and identify, obtain knowing corresponding to the wake-up of the first voice signal
Other result.
S16 wakes up speech recognition module, and will identify with waking up when wake-up recognition result is matched with preset wake-up word
As a result matched preset wake-up word is sent to speech recognition module 13.
It is appreciated that can refer to that the wake-up of word sequence form is known when waking up recognition result and the preset matching for waking up word
In other result, exist when matching with the preset wake-up word of any one word sequence form；It may also mean that characteristic sequence form
It wakes up in recognition result, exists and be higher than given threshold with the same or similar degree of the preset wake-up word of any one characteristic sequence form
When；It can also refer to wake up in recognition result and exist and the preset consistent frequency content of characteristic frequency or ability for waking up word
Other in domain can characterize the form of above-mentioned matching meaning.The preset word that wakes up according to the specific type of terminal 102 and can answer
It is arranged with scene etc. multiple.
Specifically, waking up identification module 11 can identify that obtained wake-up identifies to wake up based on the first voice signal
As a result, carrying out the matching analysis with preset wake-up word.When wake-up recognition result is matched with any one preset wake-up word, identification is waken up
The triggering of module 11 starting wakes up speech recognition module 13, and the preset wake-up word being matched to is sent to speech recognition mould
Block 13.
In this way, the preset wake-up word that speech recognition module 13 can be directly based upon starting or obtain when waking up, passes through inside
Identification network carry out identification search.
It is appreciated that wake-up identification module 11 will be after when waking up recognition result and not matching with any one preset wake-ups word
It is continuous in etc. the pending operating status for waking up identification so that speech recognition module 13 remains off state or sleep state.
Wake-up identification is carried out to the first voice signal by waking up identification module, and after waking up speech recognition module, it will
Preset wake-up word is sent to speech recognition module so that speech recognition module can directly be carried out according to preset wake-up word before to searching
Rope obtains path optimizing network.It inputs special wake-up word sound in advance without user, then inputs the instruction for needing equipment to execute
Voice, realizing that speech recognition module wakes up is synchronization gain input signal, and then obtains search result namely path optimizing net
Network, and then control needed for search result facilitates realization user can be directly based upon and operated.It reduces equipment energy consumption simultaneously, improves language
The response speed of sound control substantially increases the efficiency that voice wakes up with controls.
Recognition result and the matched judgement of preset wake-up word are waken up in one of the embodiments, may include：By word order
The preset wake-up progress the matching analysis for waking up recognition result and word sequence form of column form.Alternatively, by aligned phoneme sequence form
The preset wake-up word for waking up recognition result and aligned phoneme sequence form carries out the matching analysis.
It is appreciated that above-mentioned wake-up recognition result and preset wake-up word matching judgment, can directly pass through word sequence form
The matching analysis is carried out, whether has any one preset wake-up word to match with the resulting wake-up recognition result of determination, matching operation
Simply, judging efficiency is higher.
Recognition result is waken up during acquisition, acoustics spy can be carried out to the first voice signal by waking up identification module 11
Sign is extracted, and the acoustic feature information for corresponding to first voice signal, such as the characteristic sequence that MFCC feature extraction obtains are obtained.Cause
This, can also be by match point the wake-up recognition result of characteristic sequence form and the preset wake-up word of characteristic sequence form
Analysis determines and wakes up whether recognition result has any one preset wake-ups word to match, realize above-mentioned wake-up recognition result in advance
It sets and wakes up the matched judgement of word.Matching is convenient and accuracy is higher.Above-mentioned matching judgment can be respectively especially by this field
Corresponding routine techniques is realized, such as the skills such as deep neural network, similarity calculation, mixed Gauss model or PLP feature extraction
Art is realized.
Recognition result and the matched judgement of preset wake-up word are waken up in one of the embodiments, can also be feature
The preset wake-up word progress the matching analysis for waking up recognition result and characteristic frequency form of frequency form.It is appreciated that each
The voice of user's input can have the characteristic frequency of the corresponding sound pronunciation, therefore be called out based on what the first voice signal obtained
Awake recognition result, is also possible to corresponding characteristic frequency.To can also by the wake-up recognition result of characteristic frequency form with
The preset wake-up word of characteristic frequency form carries out the matching analysis, determines and wakes up whether recognition result has and any one preset wake-up
Word matches, and matching speed is very fast and simple.It should be noted that above-mentioned characteristic frequency can be by conventional in the art
Frequency abstraction mode obtains.
In one of the embodiments, speech recognition module will be sent to the matched preset wake-up word of recognition result is waken up
13 process, it is further comprising the steps of：The first voice signal is sent to speech recognition module 13.
It is appreciated that identification module 11 is waken up when waking up speech recognition module 13, the first language that can also will be received
Sound signal is sent to speech recognition module 13, so that before speech recognition module 13 is waken up, the voice signal of input.Such as
This, speech recognition module 13 can upon awakening, have not been obtained user continue input voice before, according to the first voice signal into
Row speech recognition obtains corresponding recognition result, so as to execute various relevant operations, such as directly according to recognition result
Execute corresponding terminal control；Alternatively, the identification of the voice signal of for example, subsequent input provides the state starting point of decoding search,
Conducive to the accuracy for improving speech recognition.
In another embodiment, as shown in figure 3, additionally providing a kind of voice search method, it is applied to Fig. 1 in this way
In terminal 102 for, and be illustrated from the angle of speech recognition module 13, include the following steps S20 to S22：
S20 is received and is waken up the preset wake-up word that identification module is sent；Wherein, preset wake-up word is obtained with identification module is waken up
To wake-up recognition result match.
S22 carries out sweep forward according to preset wake-up word, obtains corresponding to the preset path optimizing network for waking up word.
It is appreciated that speech recognition is that the voice signal of input is converted into the decoding search process of word content, generally
Process is input in the decoding search network constructed in advance to the voice signal of input, constantly progress sweep forward, is therefrom looked for
To a maximum probability searching route and export the corresponding word sequence of the searching route, namely obtain recognition result output.
Speech recognition module 13 can receive wake up identification module 11 send over, with wake-up recognition result phase
After the preset wake-up word matched, according to obtained preset wake-up word, in the speech recognition network (such as WFST network) built in advance
Middle carry out sweep forward, such as according to matched preset wake-up word, the sweep forward initialized in speech recognition network,
Obtain each searching route relevant to preset wake-up word namely path optimizing network.
By receiving the preset wake-up word for waking up identification module 13 and sending over, as the initial defeated of speech recognition module 13
Enter, so as to directly obtain corresponding path optimizing network based on preset wake-up word, or according to obtaining path optimizing network
The corresponding preset search result for waking up word of the path output of middle maximum probability to execute corresponding control operation, such as is transmitted across
The preset wake-up word come is " opening blower ", then speech recognition module 13 can complete corresponding control instruction output, and realization is beaten
Blow in machine control operation.In this way, on the one hand user, which may be implemented, only need to input a voice signal, it can obtain corresponding
Path optimizing network, or required identification output and control are completed, it wakes up and is increased dramatically with the efficiency of control is identified；Separately
On the one hand, by the way that the preset preposition search for waking up word, obtained path optimizing network can input again for subsequent user
Voice provides the state starting point of sweep forward namely the voice signal of subsequent input, can be in speech recognition network, with preset
State where waking up word is that start node carries out sweep forward, obtains exporting with the maximally related search result of preset wake-up word,
Effectively promote recognition accuracy.
In one of the embodiments, after above-mentioned step S22, it can also comprise the steps of：It obtains and wakes up identification mould
The first voice signal that block 11 receives carries out sweep forward according to the first voice signal and path optimizing network, is identified
As a result.
It is appreciated that speech recognition module 13, which can also receive, wakes up the first voice letter that identification module 11 passes over
Number, in path optimizing network, sweep forward is carried out, obtains the recognition result corresponding to the first voice signal, such as word sequence
Recognition result output or word sequence corresponding control instruction output so that terminal 102 is completed to control operation accordingly.Such as
This, is sent to speech recognition module 13 by the first voice signal and preset wake-up word, may not need and user is waited to know in voice
Voice again after other module 13 is waken up inputs, and realizes that user inputs a voice to terminal 102, terminal 102 can be complete
Operation is waken up and controlled accordingly at corresponding；Or speech recognition module 13 wake up after the voice signal that inputs, mention
For the state starting point of sweep forward, the recognition efficiency of subsequent input speech signal is improved.
In one of the embodiments, after above-mentioned steps S22, it can also comprise the steps of：The second voice signal is obtained,
Sweep forward is carried out according to the second voice signal and path optimizing network, obtains recognition result.
It is appreciated that the second voice signal is the voice signal of input after waking up speech recognition module 13.Specifically,
Speech recognition module 13 can also carry out sweep forward in path optimizing network, obtain according to the second voice signal received
To corresponding second voice signal recognition result.For example, speech recognition module 13 can be pre- according to receiving in WFST network
It sets and wakes up the preposition search of word progress, obtain the relevant path optimizing network of preset wake-up word, and then the second voice signal is inputted
Path optimizing network above-mentioned identified, available recognition result relevant to preset wake-up word.So, it is possible to reduce language
The searching route for needing to undergo in sound identification process promotes speech recognition accuracy simultaneously, promotes the processing speed of speech recognition,
Promote voice wake-up and control efficiency.
In one of the embodiments, after above-mentioned step S22, it can also comprise the steps of：Obtain the second voice letter
Number；Sweep forward is carried out according to the first voice signal, path optimizing network and the second voice signal, obtains recognition result.
It is appreciated that speech recognition module 13 can also receive the first voice signal for waking up identification module 11 and sending
After preset wake-up word, the second voice signal of input is received.Speech recognition module 13 in turn can be according to the first voice signal
With the second voice signal, sweep forward is carried out in path optimizing network, obtains maximally related with preset wake-up word, namely is most connect
The recognition result of 102 current application scene of nearly terminal.Speech recognition module 13 is so as to exporting recognition result, such as exports
Word sequence, alternatively, corresponding control instruction can be exported according to recognition result, so that terminal 102 completes corresponding control operation.
Pass through above-mentioned search step, it is possible to reduce after speech recognition module 13 is waken up, the voice signal of input
In speech recognition process, the searching route for needing to undergo promotes phonetic search accuracy rate simultaneously, promotes the place of speech recognition process
Manage speed.
It in one embodiment, is that " me please be help to open KuGoo sound to the first voice signal that terminal 102 inputs with user
For the voice of pleasure ", after waking up acquisition the first voice signal input of identification module 11, waking up identification module 11 may search for obtaining
The word sequence of " opening " or " da kai " etc. or the wake-up recognition result of aligned phoneme sequence form, and then by the wake-up recognition result
The matching analysis is carried out with each preset wake-up word.It is matched to " opening " and preset wake-up word " opening " phase waken up in recognition result
When matching, the speech recognition module 13 of triggering starting closed state, or dormant speech recognition module 13 is waken up, such as
Trigger signal is sent to speech recognition module 13, so that terminal 102 starts speech recognition module 13 or restores speech recognition module
13 working power, speech recognition module 13 enter working condition.
Identification module 11 is waken up by the first voice signal and the preset wake-up word being matched to, is sent to the language of working condition
Sound identification module 13, initial input when speech recognition module 13 is to obtain wake-up, and according to the first voice signal and matching
The preset wake-up word arrived carries out sweep forward, and then obtains recognition result.Such as output " me please be help to open KuGoo music " word order
The recognition result of column such as shows user, or output to open KuGoo music to execute corresponding operation based on recognition result
Enabled instruction, to start KuGoo music.
It in another embodiment, is " me please be help to open " to the first voice signal that terminal 102 inputs with user
For voice, wake up identification module 11 obtain the first voice signal input after, wake up identification module 11 by with above-mentioned embodiment party
The wakeup process of formula wakes up speech recognition module 13, and transmits the first voice signal and matched pre- to speech recognition module 13
It sets and wakes up word " opening ".Then available recognition result " me please be help to open " of speech recognition module 13.Speech recognition module 13 can
To receive after the second voice signal of rear input, it is based on recognition result above-mentioned, the second voice signal is identified,
Obtain final recognition result.Such as second voice signal be " air-conditioning ", then speech recognition module 13 " can please help me to beat
Open " scan for exporting in relevant multiple paths, quickly obtain the final recognition result of " me please be help to open air-conditioning ", without
After traversing all paths in entire speech recognition network, recognition result is obtained.
In above-mentioned search process, speech recognition module 13 can be completed by following optional implementation to the second voice
The identification of signal：Second voice signal " can open " corresponding search condition in the WFST network of speech recognition module 13
For start node, carry out sweep forward, obtain recognition result relevant to " opening ", as " opening " blower, " opening " rain brush or
" opening " address list etc..In other words, the second voice signal can be input to and " opening " degree of correlation by speech recognition module 13
In higher path optimizing network, sweep forward is carried out, the recognition result obtained closest to 102 application scenarios of terminal exports, and keeps away
The identification for exempting from the second voice signal traverses all searching routes in entire WFST network, reduces to wake up and imitate with the whole of identification
Rate and accuracy rate.
In another embodiment, after speech recognition module 13 is waken up, if within the set duration, not receiving input
The second voice signal, then set duration terminate when, automatically into closed state or the sleep state of low-power consumption.Shape above-mentioned
State automatically switches, and can be realized by way of conventional delay switching states various in this field, such as delay setting duration
Afterwards, jump level state or by the master controller of terminal send trigger signal so that main controller controls speech recognition
The modes such as 13 switchover operation state of module.Automatically switch operating status by delay, makes speech recognition module 13 ineffective
Voice signal input when, be automatically closed or sleep, reduce terminal 102 power consumption.
The preset wake-up word in the various embodiments described above can be the preset word for waking up word itself in one of the embodiments,
Sequence and the preset number for waking up word corresponding distribution in setting.
It is appreciated that each preset wake-up word can be word when preset wake-up word is arranged into wake-up identification module 11
The word or short sentence of sequence form, and distribute corresponding number.The obtained wake-up recognition result of identification module 11 is being waken up, and is being appointed
When one preset wake-up word matches, for example, waking up identification module when being matched to multiple preset wake-up words in wake-up recognition result
11 can send each matched preset wake-up word and its number to speech recognition module 13, may thereby determine that each preset
Wake up the sequencing that word occurs in waking up recognition result.Number above-mentioned, e.g. digital number, text is numbered or other
The number of form, as long as each preset sequence for waking up word can be distinguished.
Specifically, speech recognition module 13, which can receive, wakes up the preset of the transmission of identification module 11 in the various embodiments described above
Word and its corresponding number are waken up, so that speech recognition module 13 can carry out preceding to searching according to preset wake-ups word and its number
Rope guarantees in the searching route exported that confusion will not occur for each preset appearance order for waking up word.In this way, in speech recognition
After the carry out sweep forward of 13 pairs of module the first voice signals inputted and/or the second voice signal, preset wake-up can be maintained
The appearance sequence of word, quickly obtains the output of accurate recognition result, avoids before and after recognition result asking for the not identifications error such as corresponding
Topic.
Referring to Fig. 4, in one embodiment, also providing a kind of Rouser 100 of speech recognition, including the first signal
Obtain module 12 and wake-up module 14.First signal acquisition module 12 is for obtaining the first voice signal.Wake-up module 14 is used for
Wake-up identification is carried out to the first voice signal, obtains waking up recognition result；When wake-up recognition result is matched with preset wake-up word,
Speech recognition module is waken up, and speech recognition module will be sent to the matched preset wake-up word of recognition result is waken up.
When starting by wake-up module 14 or wake up speech recognition module 13, matched preset wake-up word is sent to voice
Identification module 13 obtains path optimizing so that speech recognition module 13 directly can carry out sweep forward according to preset wake-up word
Network.It inputs special wake-up word sound in advance without user, then inputs the instruction voice for needing equipment to execute, realize that voice is known
It is synchronization gain input signal that other module 13, which wakes up, and then obtains search result namely path optimizing network, and then can be direct
Control needed for realizing user is facilitated to operate based on search result.It reduces equipment energy consumption simultaneously, improves the response speed of voice control
Degree substantially increases the efficiency that voice wakes up with controls.
The Rouser 100 of above-mentioned speech recognition in one of the embodiments, can also be real by wake-up module 14
In existing the various embodiments described above the step of voice awakening method.
Referring to Fig. 5, in one embodiment, a kind of speech recognition equipment 200 is also provided, including receiving module 22 and search
Rope module 24.Receiving module 22, which is used to receive, wakes up the preset wake-up word that identification module is sent.Wherein, the preset wake-up
The wake-up recognition result that word is obtained with the wake-up identification module matches.Search module 24 is used for according to the preset wake-up word
Sweep forward is carried out, obtains corresponding to the preset path optimizing network for waking up word.
By receiving module 22 and search module 24, user, which on the one hand may be implemented, only need to input a voice signal, i.e.,
Available corresponding path optimizing network, or complete required identification and export and control, wake up and identify the efficiency of control
It is increased dramatically；It on the other hand, can also be by the way that the preset preposition search for waking up word, obtained path optimizing network can
To provide the state recognition starting point of sweep forward namely the voice signal of subsequent input for the voice that subsequent user inputs again,
Sweep forward identification can be carried out as start node using the state where preset wake-up word, obtained in W speech recognition FST network
It is exported to the maximally related search result of preset wake-up word, effectively promotion recognition accuracy.
Above-mentioned speech recognition equipment 200 in one of the embodiments, can also pass through receiving module 22 and search mould
Block 24 realizes the step of voice search method in the various embodiments described above.
In one embodiment, a kind of computer readable storage medium is also provided, computer program is stored thereon with, it is described
When computer program is executed by processor, the step of awakening method of above-mentioned speech recognition may be implemented：Obtain the first voice
Signal；Wake-up identification is carried out to the first voice signal, obtains waking up recognition result；Waking up recognition result and preset wake-up word
Timing wakes up speech recognition module, and sends speech recognition module for preset wake-up word；Wherein, preset wake-up word and wake-up
Recognition result matching.Alternatively, the step of above-mentioned audio recognition method may be implemented：It receives and wakes up the preset of identification module transmission
Wake up word；Wherein, preset wake-up word matches with the wake-up recognition result that identification module obtains is waken up；According to preset wake-up word into
Row speech recognition, obtains recognition result.Or the step of realizing the awakening method of above-mentioned speech recognition and above-mentioned voice
The step of recognition methods.
Computer readable storage medium above-mentioned can also be realized in the various embodiments described above in one of the embodiments,
The step of the step of awakening method of speech recognition and/or above-mentioned audio recognition method.
It will appreciated by the skilled person that realize all or part of the process in above-described embodiment method, it can be with
Relevant hardware is instructed to complete by computer program, the computer program can be stored in non-volatile computer can
It reads in storage medium, the computer program is when being executed, it may include such as the process of the embodiment of above-mentioned each method.Wherein, originally
Any reference used in each embodiment provided by applying to memory, storage, database or other media, can wrap
Include non-volatile and/or volatile memory.Nonvolatile memory may include read-only memory (ROM), programming ROM
(PROM), electrically programmable ROM (EPROM), electrically erasable ROM (EEPROM) or flash memory.Volatile memory may include
Random access memory (RAM) or external cache.By way of illustration and not limitation, RAM is available in many forms,
Such as static state RAM (SRAM), dynamic ram (DRAM), synchronous dram (SDRAM), double data rate sdram (DDRSDRAM), enhancing
Type SDRAM (ESDRAM), synchronization link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM
(RDRAM), direct memory bus dynamic ram (DRDRAM) and memory bus dynamic ram (RDRAM) etc..
Fig. 6 and Fig. 7 are please referred to, in one embodiment, also provides a kind of voice recognition terminal 300, including voice signal
Receiver 32 wakes up identification device 34 and speech recognition equipment 36.Voice signal receiver 32 is electrically connected wake-up identification dress
Set 34 and speech recognition equipment 36.It wakes up identification device 34 and is electrically connected speech recognition equipment 36.As shown in fig. 7, user is to voice
Identification terminal 300 issues the first voice, after voice signal receiver 32 receives the first voice signal, is sent to wake-up identification dress
Set 34.It wakes up identification device 34 and wake-up identification is carried out to the first voice signal, obtain waking up recognition result, and waking up identification knot
When fruit matches with preset wake-up word, speech recognition module 36 is waken up, and send to speech recognition module 36 and wake up recognition result
Matched preset wake-up word.Speech recognition equipment 36 carries out sweep forward according to preset wake-up word, obtains corresponding to preset wake-up word
Path optimizing network.
It is appreciated that voice signal receiver 32 can be various conventional sound receivers in this field.Wake up identification
Device 34 can be but not limited to the wake-up identification module circuit using dsp processor as master devices.Speech recognition equipment 36 can
To be but not limited to speech recognition module circuit of the AP processor as master devices.Waking up identification device 34 can be with speech recognition
Device 36 is arranged independently of each other, and waking up identification device 34 can also be embedded into speech recognition equipment 36, in other words, two moulds
Block can also be with integrated setting, to improve integrated level.Voice recognition terminal 300 can also include the place connected by system bus
Manage device, memory, network interface, display screen and input unit.Wherein, the processor of voice recognition terminal 300 is based on providing
Calculation and control ability.The memory of voice recognition terminal 300 includes non-volatile memory medium, built-in storage.This is non-volatile
Storage medium is stored with operating system and computer program.The built-in storage be non-volatile memory medium in operating system and
The operation of computer program provides environment.The voice wakes up to be used to pass through net with external terminal with the network interface of identification equipment
Network connection communication.The display screen of voice recognition terminal 300 can be liquid crystal display, LED display or electric ink and show
Screen, the input unit of voice recognition terminal 300 can be the touch layer covered on display screen, be also possible to computer equipment shell
Key, trace ball or the Trackpad of upper setting can also be external keyboard, Trackpad or mouse etc..
Voice recognition terminal 300 can also realize that the voice in the various embodiments described above wakes up in one of the embodiments,
The step of the step of method and/or above-mentioned voice search method.
Wake-up identification is carried out to the first voice signal by waking up identification device 34, and is waking up speech recognition equipment 36
Afterwards, matched preset wake-up word is sent to speech recognition equipment 36, so that speech recognition equipment 36 can be directly according to wake-up
Information carries out sweep forward, obtains corresponding to the preset path optimizing network for waking up word.Input special wake-up in advance without user
Word sound, then the instruction voice for needing equipment to execute is inputted, realizing that speech recognition module wakes up is synchronization gain input signal, into
And search result namely path optimizing network are obtained, and then control needed for search result facilitates realization user can be directly based upon
System operation.It reduces equipment energy consumption simultaneously, improves the response speed of voice control, substantially increase the effect that voice wakes up with controls
Rate.
Each technical characteristic of embodiment described above can be combined arbitrarily, for simplicity of description, not to above-mentioned reality
It applies all possible combination of each technical characteristic in example to be all described, as long as however, the combination of these technical characteristics is not deposited
In contradiction, all should be considered as described in this specification.
The embodiments described above only express several embodiments of the present invention, and the description thereof is more specific and detailed, but simultaneously
It cannot therefore be construed as limiting the scope of the patent.It should be pointed out that coming for those of ordinary skill in the art
It says, without departing from the inventive concept of the premise, various modifications and improvements can be made, these belong to protection of the invention
Range.Therefore, the scope of protection of the patent of the invention shall be subject to the appended claims.
Claims (11)
1. a kind of voice awakening method, which is characterized in that include the following steps；
Obtain the first voice signal；
Wake-up identification is carried out to first voice signal, obtains waking up recognition result；
When the wake-up recognition result is matched with preset wake-up word, speech recognition module is waken up, and will identify with the wake-up
As a result matched preset wake-up word is sent to the speech recognition module.
2. voice awakening method according to claim 1, which is characterized in that will be matched pre- with the wake-up recognition result
The process that word is sent to the speech recognition module that wakes up is set, further includes：
First voice signal is sent to the speech recognition module.
3. voice awakening method according to claim 1, which is characterized in that the wake-up recognition result and preset wake-up word
Matched judgement, including：
The preset wake-up word of the wake-up recognition result of word sequence form and word sequence form is subjected to the matching analysis；
Or, the wake-up recognition result of characteristic sequence form is matched with the preset wake-up word of characteristic sequence form
Analysis.
4. a kind of voice search method, which is characterized in that include the following steps：
It receives and wakes up the preset wake-up word that identification module is sent；Wherein, the preset wake-up word is obtained with the wake-up identification module
To wake-up recognition result match；
Sweep forward is carried out according to the preset wake-up word, obtains corresponding to the preset path optimizing network for waking up word.
5. voice search method according to claim 4, which is characterized in that searching before being carried out according to the preset wake-up word
Rope, after obtaining the step of corresponding to the preset path optimizing network for waking up word, including：
The first voice signal that the wake-up identification module receives is obtained, according to first voice signal and the optimization road
Diameter network carries out sweep forward, obtains recognition result.
6. voice search method according to claim 4, which is characterized in that searching before being carried out according to the preset wake-up word
Rope further includes after the step of obtaining corresponding to the preset path optimizing network for waking up word：
The second voice signal is obtained, sweep forward is carried out according to second voice signal and the path optimizing network, is obtained
The recognition result.
7. voice search method according to claim 5, which is characterized in that searching before being carried out according to the preset wake-up word
Rope further includes after the step of obtaining corresponding to the preset path optimizing network for waking up word：
Obtain the second voice signal；
Sweep forward is carried out according to first voice signal, the path optimizing network and second voice signal, is obtained
The recognition result.
8. a kind of voice Rouser, which is characterized in that including：
First signal acquisition module, for obtaining the first voice signal；
Wake-up module obtains waking up recognition result for carrying out wake-up identification to first voice signal；Know in the wake-up
When other result is matched with preset wake-up word, wake up speech recognition module, and will with the wake-up recognition result is matched preset calls out
Awake word is sent to the speech recognition module.
9. a kind of voice searching device, which is characterized in that including：
Receiving module, for receiving the preset wake-up word for waking up identification module and sending；Wherein, the preset wake-up word is called out with described
The wake-up recognition result that awake identification module obtains matches；
Search module obtains corresponding to the preset optimization for waking up word for carrying out sweep forward according to the preset wake-up word
Path network.
10. a kind of computer readable storage medium, is stored thereon with computer program, which is characterized in that the computer program
The step of voice awakening method described in any one of claims 1 to 3 is realized when being executed by processor and/or claim 4
The step of to voice search method described in any one of 7.
11. a kind of voice recognition terminal, which is characterized in that including voice signal receiver, wake up identification device and speech recognition
Device, the voice signal receiver are electrically connected the wake-up identification device and the speech recognition equipment, the wake-up
Identification device is electrically connected the speech recognition equipment；
After the voice signal receiver receives the first voice signal, it is sent to the wake-up identification device；The wake-up is known
Other device carries out wake-up identification to first voice signal, obtains waking up recognition result, and in the wake-ups recognition result and
When preset wake-up word matches, the speech recognition module is waken up, and send to the speech recognition module and identify with the wake-up
As a result matched preset wake-up word；
The speech recognition equipment carries out sweep forward according to the preset wake-up word, obtains corresponding to the excellent of the preset wake-up word
Change path network.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810587174.0A CN108899028A (en) | 2018-06-08 | 2018-06-08 | Voice awakening method, searching method, device and terminal |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810587174.0A CN108899028A (en) | 2018-06-08 | 2018-06-08 | Voice awakening method, searching method, device and terminal |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108899028A true CN108899028A (en) | 2018-11-27 |
Family
ID=64344285
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810587174.0A Pending CN108899028A (en) | 2018-06-08 | 2018-06-08 | Voice awakening method, searching method, device and terminal |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108899028A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109448720A (en) * | 2018-12-18 | 2019-03-08 | 维拓智能科技（深圳）有限公司 | Convenience service self-aided terminal and its voice awakening method |
| CN109545211A (en) * | 2018-12-07 | 2019-03-29 | 苏州思必驰信息科技有限公司 | Voice interactive method and system |
| CN109559743A (en) * | 2018-12-05 | 2019-04-02 | 嘉兴行适安车联网信息科技有限公司 | Vehicle-mounted immediate communication tool information sharing method based on android system |
| CN110689887A (en) * | 2019-09-24 | 2020-01-14 | Oppo广东移动通信有限公司 | Audio verification method and device, storage medium and electronic equipment |
| CN110989963A (en) * | 2019-11-22 | 2020-04-10 | 北京梧桐车联科技有限责任公司 | Awakening word recommendation method and device and storage medium |
Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104282301A (en) * | 2013-07-09 | 2015-01-14 | 安徽科大讯飞信息科技股份有限公司 | Voice command processing method and system |
| CN104538030A (en) * | 2014-12-11 | 2015-04-22 | 科大讯飞股份有限公司 | Control system and method for controlling household appliances through voice |
| US20150179166A1 (en) * | 2013-12-24 | 2015-06-25 | Kabushiki Kaisha Toshiba | Decoder, decoding method, and computer program product |
| CN104866274A (en) * | 2014-12-01 | 2015-08-26 | 联想(北京)有限公司 | Information processing method and electronic equipment |
| CN105575395A (en) * | 2014-10-14 | 2016-05-11 | 中兴通讯股份有限公司 | Voice wake-up method and apparatus, terminal, and processing method thereof |
| CN105654943A (en) * | 2015-10-26 | 2016-06-08 | 乐视致新电子科技（天津）有限公司 | Voice wakeup method, apparatus and system thereof |
| CN105869637A (en) * | 2016-05-26 | 2016-08-17 | 百度在线网络技术（北京）有限公司 | Voice wake-up method and device |
| CN106297777A (en) * | 2016-08-11 | 2017-01-04 | 广州视源电子科技股份有限公司 | A kind of method and apparatus waking up voice service up |
| US20170263242A1 (en) * | 2016-03-14 | 2017-09-14 | Kabushiki Kaisha Toshiba | Information processing device, information processing method, computer program product, and recognition system |
| CN107369439A (en) * | 2017-07-31 | 2017-11-21 | 北京捷通华声科技股份有限公司 | A kind of voice awakening method and device |
| CN107450879A (en) * | 2016-05-30 | 2017-12-08 | 中兴通讯股份有限公司 | Terminal operation method and device |
| CN107622652A (en) * | 2016-07-15 | 2018-01-23 | 青岛海尔智能技术研发有限公司 | The sound control method and appliance control system of appliance system |
| CN107886944A (en) * | 2017-11-16 | 2018-04-06 | 出门问问信息科技有限公司 | A kind of audio recognition method, device, equipment and storage medium |
-
2018
- 2018-06-08 CN CN201810587174.0A patent/CN108899028A/en active Pending
Patent Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104282301A (en) * | 2013-07-09 | 2015-01-14 | 安徽科大讯飞信息科技股份有限公司 | Voice command processing method and system |
| US20150179166A1 (en) * | 2013-12-24 | 2015-06-25 | Kabushiki Kaisha Toshiba | Decoder, decoding method, and computer program product |
| CN105575395A (en) * | 2014-10-14 | 2016-05-11 | 中兴通讯股份有限公司 | Voice wake-up method and apparatus, terminal, and processing method thereof |
| CN104866274A (en) * | 2014-12-01 | 2015-08-26 | 联想(北京)有限公司 | Information processing method and electronic equipment |
| CN104538030A (en) * | 2014-12-11 | 2015-04-22 | 科大讯飞股份有限公司 | Control system and method for controlling household appliances through voice |
| CN105654943A (en) * | 2015-10-26 | 2016-06-08 | 乐视致新电子科技（天津）有限公司 | Voice wakeup method, apparatus and system thereof |
| US20170263242A1 (en) * | 2016-03-14 | 2017-09-14 | Kabushiki Kaisha Toshiba | Information processing device, information processing method, computer program product, and recognition system |
| CN105869637A (en) * | 2016-05-26 | 2016-08-17 | 百度在线网络技术（北京）有限公司 | Voice wake-up method and device |
| CN107450879A (en) * | 2016-05-30 | 2017-12-08 | 中兴通讯股份有限公司 | Terminal operation method and device |
| CN107622652A (en) * | 2016-07-15 | 2018-01-23 | 青岛海尔智能技术研发有限公司 | The sound control method and appliance control system of appliance system |
| CN106297777A (en) * | 2016-08-11 | 2017-01-04 | 广州视源电子科技股份有限公司 | A kind of method and apparatus waking up voice service up |
| CN107369439A (en) * | 2017-07-31 | 2017-11-21 | 北京捷通华声科技股份有限公司 | A kind of voice awakening method and device |
| CN107886944A (en) * | 2017-11-16 | 2018-04-06 | 出门问问信息科技有限公司 | A kind of audio recognition method, device, equipment and storage medium |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109559743A (en) * | 2018-12-05 | 2019-04-02 | 嘉兴行适安车联网信息科技有限公司 | Vehicle-mounted immediate communication tool information sharing method based on android system |
| CN109545211A (en) * | 2018-12-07 | 2019-03-29 | 苏州思必驰信息科技有限公司 | Voice interactive method and system |
| CN109448720A (en) * | 2018-12-18 | 2019-03-08 | 维拓智能科技（深圳）有限公司 | Convenience service self-aided terminal and its voice awakening method |
| CN110689887A (en) * | 2019-09-24 | 2020-01-14 | Oppo广东移动通信有限公司 | Audio verification method and device, storage medium and electronic equipment |
| CN110689887B (en) * | 2019-09-24 | 2022-04-22 | Oppo广东移动通信有限公司 | Audio verification method and device, storage medium and electronic equipment |
| CN110989963A (en) * | 2019-11-22 | 2020-04-10 | 北京梧桐车联科技有限责任公司 | Awakening word recommendation method and device and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108899028A (en) | Voice awakening method, searching method, device and terminal | |
| CN111223497B (en) | Nearby wake-up method and device for terminal, computing equipment and storage medium | |
| CN106653021A (en) | Voice wake-up control method and device and terminal | |
| CN102543071B (en) | Voice recognition system and method used for mobile equipment | |
| CN110890093B (en) | Intelligent equipment awakening method and device based on artificial intelligence | |
| CN108520743A (en) | Sound control method, smart machine and the computer-readable medium of smart machine | |
| CN110534099A (en) | Voice wakes up processing method, device, storage medium and electronic equipment | |
| CN108735209A (en) | Wake up word binding method, smart machine and storage medium | |
| CN105190746A (en) | Method and apparatus for detecting a target keyword | |
| CN110570840B (en) | Intelligent device awakening method and device based on artificial intelligence | |
| CN111261144A (en) | Voice recognition method, device, terminal and storage medium | |
| CN102847325B (en) | Toy control method and system based on voice interaction of mobile communication terminal | |
| CN108766438A (en) | Man-machine interaction method, device, storage medium and intelligent terminal | |
| US11810593B2 (en) | Low power mode for speech capture devices | |
| CN110570857B (en) | Voice wake-up method and device, electronic equipment and storage medium | |
| CN111862938A (en) | Intelligent response method, terminal and computer readable storage medium | |
| CN108682415A (en) | voice search method, device and system | |
| CN114360510A (en) | Voice recognition method and related device | |
| CN110503962A (en) | Speech recognition and setting method, device, computer equipment and storage medium | |
| CN102868740A (en) | Method and system for controlling toy based on mobile communication terminal and internet voice interaction | |
| CN113611316A (en) | Man-machine interaction method, device, equipment and storage medium | |
| WO2019071723A1 (en) | Speech-to-speech translation method and device and translating machine | |
| CN114391165A (en) | Voice information processing method, device, equipment and storage medium | |
| WO2020073839A1 (en) | Voice wake-up method, apparatus and system, and electronic device | |
| CN108694939B (en) | Voice search optimization method, device and system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181127 |