Summary of the invention
Against the above technical problems, the invention discloses a kind of navigation audio recognition method and system, pass through specific identification
Region and the analysis of speech engine recognition result, and obtain the result closer to input voice.
In this regard, the technical solution adopted by the present invention are as follows:
A kind of navigation audio recognition method comprising following steps:
Step S1: the voice messaging that client is received carries out speech recognition by speech engine;
Step S2: background data base is retrieved according to identification region to the result of step S1 speech recognition, if there is correspondence
Data record, then return to client for recognition result, and identification is completed;
It records, is then tied from the default territorial scope of searching in local data base and identification if there is no corresponding data
Fruit similarity is greater than the data of preset value, if there is similarity greater than preset value as a result, being then greater than the similarity pre-
If the data of value recommend user according to frequency of use descending sort;If be greater than in the local database without similarity default
The recognition result is then back in background server database by the data of value, and background server calculates background server number
According to the similarity of records and recognition result all in library, according to all records of similarity descending sort, and ranking results are returned
To client;Wherein, the similar data refer to the data for being greater than default similarity;
Data as being also not greater than default similarity in background server database, then this speech recognition errors, and
Feed back to client.
Wherein, local data base refers to the included storing data library of equipment.
It adopts this technical solution, realizes the identification to speech recognition through a variety of ways, and obtain in conjunction with specific identification region
Take closer to input voice as a result, reducing search time.
As a further improvement of the present invention, in step S2, first choice judges the result of step S1 speech recognition, is
No is effectively as a result, then being retrieved.Adopt this technical solution, first to recognition result carry out preliminary analysis, see whether be
Effectively as a result, if it is not, then no longer carry out the retrieval of next step, makes result feedback faster, save the waste of unnecessary time.
As a further improvement of the present invention, in step S2, if there is similarity greater than preset value as a result, then by it
User is recommended according to frequency of use descending sort, and obtains user current location information, according to territorial scope garbled data, instead
It is fed to client.It adopts this technical solution, is analyzed in conjunction with user current location information, so that the result of discriminance analysis is more
Accurately.
As a further improvement of the present invention, in step S2, if it is not similar in the local database as a result, if
The recognition result is back in background server database, server calculate in background server database it is all record with
The similarity of recognition result is back to client according to all records of similarity descending sort, and by ranking results, and obtains use
Family current location information feeds back to client according to territorial scope garbled data.
As a further improvement of the present invention, in step S2, as being also not greater than default phase in background server database
Like the data of degree, then this speech recognition errors, and feed back to client, while inform user re-enter voice messaging or
Determiner or expansion word is added.
As a further improvement of the present invention, in step S2, the default similarity is not less than 50%.
The invention also discloses a kind of speech recognition system, the speech recognition system includes speech reception module, voice
Identification module, local data base, communication module, background server and sending module;Wherein, the speech reception module, is used for
Receive the voice messaging that client is sent;The speech recognition module, for voice messaging to be retrieved backstage according to identification region
Server data library lookup corresponding data record, or in territorial scope and recognition result is preset from finding in local data base
Similar data;The recognition result is then back to background server for connecting with background server by the communication module
In database;The background server, it is similar to recognition result for calculating all records in background server database
Degree, according to all records of similarity descending sort;Sending module, for ranking results to be back to client.
Compared with prior art, the invention has the benefit that
It using technical solution of the present invention, is analyzed, and is obtained more by specific identification region and speech engine recognition result
Close to input voice as a result, reduction search time, more convenient to use.Technical solution of the present invention can be applied in each need
It wants on the platform of speech recognition, especially intelligent terminal.
Specific embodiment
With reference to the accompanying drawing, preferably embodiment of the invention is described in further detail.
A kind of navigation audio recognition method, as shown in Figure 1, after obtaining recognition result using third party's speech engine first,
It retrieves and sees with the presence or absence of corresponding note in platform server database after passing through in the identification of the voice content according to selected by user region
Record, i.e., identical record;If it exists, then the recognition result obtained speech engine is sent to client;Such result is only
Only one recognition result corresponds to a data in background server searching database;If comparing nothing in the result and database of identification
It is identical, then it is assumed that recognition result is there are certain falsehood, while concurrent sending voice recognition result, reports to client knowledge
There are falsehoods to enter simultaneously in next step for other result;Then in default territorial scope and recognition result is found in the local database
Then similar data recommend user with the arrangement of frequency of use descending, reduce user's search time;Wherein, described close
As data refer to the data for being greater than default similarity.In addition, if finding in the local database similar less than greater than presetting
The recognition result is then back in background server retrieval data by the data of degree, and background server calculates institute in database
There is the similarity of record and recognition result, according to all records of similarity descending sort, and ranking results is back to client.
As background server database is equally also found less than the data for being greater than default similarity, then it is assumed that this speech recognition errors,
And client is fed back to, it reminds user to re-enter voice messaging or determiner or expansion word is added.
For example, user is at requiring to navigate to Guangzhou Guang Zhouta, it is preset in Chengdu input voice " Guang Zhouta "
Territorial scope is that districts under city administration are the place name for only searching for Chengdu, then can not search number corresponding with " Guang Zhouta " on backstage
According to;It then searches in the local database in next step, as there is " optical axis tower " " Guang Zhouta " and " stroll week it " in local data base
" Guang Zhouta ", frequency are respectively A, B, C and D, D > C > B > A, then the information being presented to the user is to be followed successively by from top to bottom
" Guang Zhouta ", " stroll week it ", " Guang Zhouta ", " optical axis tower " facilitates client to choose, identifies successfully.
If the preset similarity of local data base is greater than 50%, and its recognition result " Guang Zhouta " can not find phase therewith
It is greater than 50% data like degree, is retrieved then recognition result " Guang Zhouta " is fed back in background server database, passes through backstage
Server calculates all records and the similarities of recognition result in database, by similarity be more than preset threshold data according to
All records of similarity descending sort, and ranking results are back to client.
The similarity of all records and recognition result in database is such as calculated by server-side, similarity is respectively less than pre-
If threshold value, then recognition failures, then feeding back to user's secondary voice input error, and prompt user to replace voice or addition
Determiner or expansion word, such as Chengdu user need navigate be located at Guangzhou " Guang Zhouta ", then input voice information when
It waits and adds expansion word, i.e., input voice is " Guangzhou Guang Zhouta ".
The above content is a further detailed description of the present invention in conjunction with specific preferred embodiments, and it cannot be said that
Specific implementation of the invention is only limited to these instructions.For those of ordinary skill in the art to which the present invention belongs, exist
Under the premise of not departing from present inventive concept, a number of simple deductions or replacements can also be made, all shall be regarded as belonging to of the invention
Protection scope.