CN105653596A - Quick startup method and device of specific function on the basis of voice frequency comparison - Google Patents

The invention discloses a quick startup method and device of a specific function on the basis of voice frequency comparison. The startup method of the specific function comprises the following steps: A. under a voice control mode, collecting the current voice of a user; B. inquiring whether pre-recorded voice matched with the voice is in the presence in the recording area of a voice library: executing a function bound with the recorded voice if the pre-recorded voice matched with the voice is in the presence in the recording area of the voice library, and otherwise, executing C; and C: inquiring preset voice matched with the voice is in the presence in the public area of the voice library: executing the function bound with the preset voice if the preset voice matched with the voice is in the presence in the public area of the voice library, and otherwise, prompting voice errors. The user voice is pre-recorded and stored and is bound with corresponding functions, the voice of the user is collected when the startup method is used, and the corresponding function can be executed when a judgment result shows that the voice of the user is matched with the recorded voice. The quick startup method is free from the restriction of mandarin, supports any localism, improves the identification rate of voice identification and brings great convenience for users to use.

A kind of specific function quick start method based on audio frequency contrast and device thereof

Technical field

The present invention relates to communication technical field, in particular to a kind of specific function quick start method based on audio frequency contrast and device thereof.

Background technology

Old machine be instantly a kind of very popular be specially communication product that the elderly designs. Because it is with strong points, nearly all function is all specially for the elderly designs, so greatly facilitating the routine use of the elderly and operating very simple. Such as, when making a phone call, the old machine of great majority is support voice dialing function all. But, the keyword in this kind of voice dial-up function identification user speech, searching in shared voice storehouse, if finding the keyword of coupling, performing the function that this keyword is corresponding.

But the discrimination power that this kind of mode exists speech recognition is not problem very accurately. The voice storehouse of common speech recognition is that setting is good or needs download. And a lot of area dialect pronunciation difference is excessive, as the dialect of some areas, Hunan across several households in Ji Tiao street all varies. And everyone pronunciation custom, tone, comprises custom of speaking not identical. But the storage capacity in voice storehouse is limited, it is impossible to record all audio frequency. And be usually as the criterion with mandarin during system identification, and the user of not all can say the mandarin of standard. Even if everybody says to be that mandarin also may be difficult to due to the difference of speech habits resolve. The voice of this kind of band dialect intonation is often identified mistake, thus mistakes the intention of user.

Thus prior art need to improve.

Summary of the invention

It is an object of the invention to provide a kind of specific function quick start method based on audio frequency contrast and device thereof, identify mistake because user pronunciation difference is relatively big to solve existing Voice command, cause the problem of Voice command mishandle.

In order to achieve the above object, this invention takes following technical scheme:

Based on a specific function quick start method for audio frequency contrast, comprising:

A, under Voice command pattern, gather the current voice of user;

Whether B, recording region inquiry in voice storehouse have the voice prerecorded mated mutually with described voice: be perform the function of voice bindings recorded, otherwise perform step C;

Whether C, public domain inquiry in voice storehouse have the default voice mated mutually with described voice: be then perform the function presetting voice bindings, otherwise suggestion voice mistake.

In the described specific function quick start method based on audio frequency contrast, before described steps A, also comprise:

Steps A 01, the voice prerecording user, extract reduced parameter and store, by this reduced parameter and corresponding function binding.

In the described specific function quick start method based on audio frequency contrast, after described steps A 01, also comprise:

The function of voice and the correspondence recorded reported in A02, voice, so that user is confirmed whether correctly.

In the described specific function quick start method based on audio frequency contrast, described step B specifically comprises:

B1, the current voice gathered is compressed, extract reduced parameter;

B2, judge to record in region whether have whether the reduced parameter mated mutually with the reduced parameter extracted mates;

If B3 records in region and has the difference of reduced parameter and the reduced parameter extracted in preset range and the reduced parameter extracted is greater than predetermined threshold value, then it is identified as coupling, performs the function of this reduced parameter binding;

If the described difference of B4 reduced parameter that is outer in preset range and that extract is greater than predetermined threshold value, then it is identified as and does not mate, perform step C.

In the described specific function quick start method based on audio frequency contrast, described step C specifically comprises: whether the public domain inquiry in voice storehouse has and the reduced parameter that the reduced parameter extracted is mated mutually, it is perform the function of this reduced parameter binding, otherwise suggestion voice mistake.

In the described specific function quick start method based on audio frequency contrast, when reduced parameter coupling, before performing the function of this reduced parameter binding, also comprise:

Voice corresponding to reduced parameter and function reported in voice, to point out user to confirm;

Gather the answer voice of user, when being identified as default firm reply, perform this function.

Based on a specific function device for rapidly starting for audio frequency contrast, it comprises voice storehouse, contrast module and function module;

Described voice storehouse is used for storing, in recording region, the voice prerecorded, and stores default voice in public domain;

Described comparing module, under Voice command pattern, gathers the voice that user is current; Recording whether region inquiry has the voice prerecorded mated mutually with described voice: be that then controlling functions module performs the function of the voice bindings recorded;

Otherwise, in public domain, whether inquiry has the default voice mated mutually with described voice: be that then controlling functions module performs the function presetting voice bindings, otherwise suggestion voice mistake.

In the described specific function device for rapidly starting based on audio frequency contrast, also comprise sample and refine module and function binding module;

Described sample refines the voice that module prerecords user, extracts reduced parameter send and stores to the recording region in function binding module and voice storehouse; This reduced parameter is stored by function binding module with the recording region being transferred to voice storehouse after corresponding function binding.

In the described specific function device for rapidly starting based on audio frequency contrast, also comprise voice and report module, report voice and corresponding function for voice, so that user is confirmed whether correctly.

Compared to prior art, the specific function quick start method based on audio frequency contrast provided by the invention and device thereof, under Voice command pattern, gather the voice that user is current;Whether the recording region inquiry in voice storehouse has the voice prerecorded mated mutually with described voice: is perform the function of the voice bindings recorded, otherwise performs step; Whether the public domain inquiry in voice storehouse has the default voice mated mutually with described voice: be then perform the function presetting voice bindings, otherwise suggestion voice mistake; By prerecord store user speech and bind with corresponding function, it may also be useful to time gather user voice and judge to recording voice match time can perform corresponding function; No longer limit by mandarin, any dialect can, it is to increase the discrimination power of speech recognition, greatly facilitates user and uses.

Accompanying drawing explanation

Fig. 1 is the method flow diagram of the specific function quick start method based on audio frequency contrast provided by the invention.

Fig. 2 is the structure block diagram of the specific function device for rapidly starting based on audio frequency contrast provided by the invention.

Embodiment

The present invention provides a kind of specific function quick start method based on audio frequency contrast and device thereof, by prerecord store user speech and bind with corresponding function, it may also be useful to time gather user voice and judge to recording voice match time can perform corresponding function; No longer limit by mandarin, any dialect can, it is to increase the discrimination power of speech recognition, greatly facilitates user and uses. For making the object of the present invention, technical scheme and effect clearly, clearly, developing simultaneously referring to accompanying drawing, the present invention is described in more detail for embodiment. It is to be understood that specific embodiment described herein is only in order to explain the present invention, it is not intended to limit the present invention.

Referring to Fig. 1, it is the specific function quick start method schema based on audio frequency contrast provided by the invention. Specific function quick start method provided by the invention comprises:

S100, under Voice command pattern, gather the current voice of user;

Whether S200, recording region inquiry in voice storehouse have the voice prerecorded mated mutually with described voice: be perform the function of voice bindings recorded, otherwise perform step S300;

Whether S300, public domain inquiry in voice storehouse have the default voice mated mutually with described voice: be then perform the function presetting voice bindings, otherwise suggestion voice mistake.

In the present embodiment, before described step S100, it is necessary to first generate the voice of recording, comprising:

Prerecord user voice, extract reduced parameter and store, by this reduced parameter and corresponding function binding.

User can enter voice and arrange interface, selects corresponding function and clicks, enter the voice recording interface under this function in the function frame ejected, and clicks recording button and can record. The audio frequency recorded is forgotten, it is possible to pre-set and record rule in order to avoid user, as first talked about function, " making a phone call " or " send short messages (sending out message) " etc.; Besides object (make a phone call to whom or send short messages), such as son or true name. Reaffirming with user after function binding, namely the voice recorded and corresponding function reported in voice, so that user is confirmed whether correctly, it may also be useful to whether the clear and definite system in family correctly identifies voice, follow-up speech processes is user's original idea really. Thus drastically increase the discrimination power of speech recognition.

As user selects the function of " making a phone call ", after click, carry out the voice recording interface made a phone call, click and say " making a phone call " after recording button; Then recording is pointed out to dial object for " mother ".Gather this voice and carry out processed compressed, extract the reduced parameter of " making a phone call " and " mother " and be stored in the recording region in voice storehouse. The compression and the reduced parameter that carry out are extracted as prior art herein, are not described further herein. Afterwards this reduced parameter and the function of selection are bound with the telephone number of this object.

User also arranges Voice command by each function interface, namely clicks voice setting options in function interface, carry out after the voice of recording compressing, reduced parameter extract and be stored in the recording region in voice storehouse, afterwards by reduced parameter and this function binding.

For a kind of function, user can record multiple model simultaneously, such as phone functions, can be recorded as " dialing ", " call ", " calling " etc., to meet the different speech habits of user. The reduced parameter that these voices extract all with the function binding made a phone call.

It will be appreciated that, object certainly will be related to during voice dial phone, for this reason, the instruction called is extracted reduced parameter by the present embodiment together with object, the telephone number of this reduced parameter and phone functions and this object is directly bound, thus the function of quick voice dial appointed object phone can be realized. If user does not have recorded speech when object is dialed in prompting recording, then only extract the reduced parameter of " making a phone call ", this reduced parameter and the function made a phone call are bound. Follow-up need user's phonetic entry telephone number again, existing Speech comparison mode now can be adopted to carry out the input of telephone number.

In this step, contrast, based on audio frequency, the voice that principle generates recording. Audio frequency contrast refers to the feature (i.e. reduced parameter) searching out from one section of continuous print audio-frequency information (voice of recording) and specifying, it is not necessary to refine concrete semantic. Then when needs carry out Voice command, the mode adopting reduced parameter to compare judges whether the voice of current collection mates mutually with the voice recorded. Then described step S200 specifically comprises:

Step 201, the current voice gathered is compressed, extract reduced parameter;

Step 202, judge to record in region whether have whether the reduced parameter mated mutually with the reduced parameter extracted mates;

If step 203 is recorded in region and had the difference of reduced parameter and the reduced parameter extracted in preset range and the reduced parameter extracted is greater than predetermined threshold value, then it is identified as coupling, performs the function of this reduced parameter binding;

If the described difference of step 204 reduced parameter that is outer in preset range and that extract is greater than predetermined threshold value, then it is identified as and does not mate, perform step S300.

Wherein, the reduced parameter extracted is less than or equal to predetermined threshold value then illustrates it is neighbourhood noise, can not as the basis of Voice command, it is not necessary to process. Can judging whether that record and the current voice gathered belongs to same people or different people according to matching result like this, whether voice content mates. Coupling then directly calls the function of this reduced parameter binding. It is to bind with corresponding reduced parameter by registering corresponding event handling function during binding.

In described step S204, the difference recording the reduced parameter in region and the reduced parameter extracted, outside preset range, illustrates that this voice may do not recorded in advance, then can inquire about to public domain. Voice (reduced parameter) in this public domain is for setting in advance before dispatching from the factory or by web download.

Then in described step S300, whether the public domain inquiry in voice storehouse has and the reduced parameter that the reduced parameter extracted is mated mutually, is perform the function of this reduced parameter binding, otherwise suggestion voice mistake.Reduced parameter in public domain is equivalent to default voice. The voice recorded does not find the voice of coupling, then goes public domain is searched for, therefrom filter out the reduced parameter mated the most. The mode prerecorded can accelerate speech recognition, solves accent, problem that intonation is not correctly validated; Meanwhile, the voice in public domain can effectively supplement, to meet the various Voice command demands of user.

In the specific implementation, in order to improve the accuracy rate of identification, when identification reduced parameter coupling, after namely finding the voice of coupling, before performing the function of this reduced parameter binding, also report voice corresponding to reduced parameter by voice and function points out user to confirm; Gather the answer voice of user, when being identified as default firm reply, perform this function. Whether such as, after identifying that the reduced parameter of " making a phone call " and " mother " is mated, whether first voice is reported and to be made a phone call, if user answers and is or to (firm reply preset), then carry out next step confirmation calling out personal identification, namely make a phone call to mother; It is dial mother's phone that user answers. If not to; answer " mistake " and "no" (default negative answer), now to, after the clearing of current gathered voice, voice message user says instruction again. Due to "Yes", " to ", " mistake ", the voice of "no" is very simple and easy to identify, adopts existing speech recognition and correlation technique herein.

The present embodiment, after finding coupling voice, does not advise directly performing corresponding function, but performs after first confirming with user again, can further improve the exactness of Voice command like this. It is understood that voice broadcasting modes is except before can be applicable to and performing corresponding function, it is also possible to before described step S100, namely confirm during recorded speech. User can be facilitated like this to understand whether its voice recorded normally is identified by system.

Based on the above-mentioned specific function quick start method based on audio frequency contrast, the present invention is a kind of specific function device for rapidly starting of corresponding offer also, and as shown in Figure 2, it comprises voice storehouse 100, contrast module 200 and function module 300. Described voice storehouse 100 is for storing, in recording region, the voice prerecorded, and stores default voice in public domain. Described comparing module 200, under Voice command pattern, gathers the voice that user is current; Recording whether region inquiry has the voice prerecorded mated mutually with described voice: be that then controlling functions module 300 performs the function of the voice bindings recorded.

Otherwise, in public domain, whether inquiry has the default voice mated mutually with described voice to comparing module 200: be that then controlling functions module 300 performs the function presetting voice bindings, otherwise suggestion voice mistake.

In the present embodiment, described specific function device for rapidly starting also comprises sample and refines module 400 and function binding module 500. Described sample refines the voice that module 400 prerecords user, extracts reduced parameter send and stores to the recording region in function binding module 500 and voice storehouse; This reduced parameter is stored by function binding mould 500 with the recording region being transferred to voice storehouse after corresponding function binding.

In the present embodiment, described specific function device for rapidly starting also comprises voice and reports module 600, reports voice and corresponding function for voice, so that user is confirmed whether correctly. Described comparing module 200 also gathers the answer voice of user, is identified as default firm reply control function module 300 and performs this function.

In sum, the present invention adopts audio frequency correlation technique, by prerecording the voice of user, extracts reduced parameter and is stored to corresponding function binding and records region; During use voice control function, gathering the voice of user and extract reduced parameter, judging to record has reduced parameter can perform corresponding function to the reduced parameter coupling extracted in region; No longer limit by mandarin, any dialect can identify, it is to increase the discrimination power of speech recognition, greatly facilitates user and uses.

The division of above-mentioned functions module only in order to illustrate, in actual applications, it is possible to as required by above-mentioned functions distribution complete by different function modules, be namely divided into different function modules, complete all or part of function of foregoing description.

One of ordinary skill in the art will appreciate that all or part of flow process in above-described embodiment method, it is can be completed by the hardware that computer (mobile terminal) program carrys out instruction relevant, described computer (mobile terminal) program can be stored in a computer (mobile terminal) read/write memory medium, program, when performing, can comprise the flow process of the embodiment of above-mentioned each method. Storage media wherein can be that magnetic disc, CD, read-only storage memory body (ROM) or random storage remember body (RAM) etc.

It should be appreciated that for those of ordinary skills, it is possible to it is equal to according to the technical scheme of the present invention and invention design thereof and replaces or change, and all these are changed or replace the protection domain that all should belong to the claim appended by the present invention.