KR20090013876A - Method and apparatus for distributed speech recognition using phonemic symbol - Google Patents

Abstract

A method and an apparatus for recognizing a distributed speech by using a phoneme are provided to recognize a phoneme in a terminal and recognize a speech with regard to a phoneme recognition result, thereby realizing information search through speech recognition without realizing additional communication standards. When a natural language query is inputted to a terminal, a phoneme of the natural language query is recognized(401). The terminal sends the recognized phoneme signal to a speech recognition server(403). The speech recognition server performs the speech recognition by using the phoneme signal(405,407). The speech recognition server confirms a search identifier to be sent to the search sever(409). The speech recognition server sends a search query signal by using the confirmed search identifier(411). The search server sends a query response signal in response to the search query signal(413).

Description

Method and apparatus for distributed speech recognition using phonemes {Method and Apparatus for distributed Speech Recognition using phonemic symbol}

The present invention relates to a distributed speech recognition method and apparatus using phonemes.

In detail, the present invention relates to a method and apparatus for processing a large volume of speech recognition by using a small amount of phoneme recognition in a terminal and using the information.

The present invention relates to a distributed speech recognition method and apparatus using phonemes.

With the development of the speech recognition method and the development of portable multimedia terminals such as mobile phones and PDAs, research on the method of speech recognition in portable terminal devices has continued.

It is difficult to perform speech recognition of many vocabularies in a mobile terminal with limited computational amount and memory. In order to overcome this problem, a method of delivering voice to a server through a communication channel and performing voice recognition on a large-capacity server is used. In this case, when the received voice is sent to the server as it is, a large amount of data is transmitted. In particular, when the data is sent through the voice communication network of the mobile phone, the error on the channel is reflected, and the server receives the degraded voice as compared with the transmitted signal, which leads to a decrease in the voice recognition performance.

Therefore, the existing ETSI AURORA standard uses a method that extracts the features needed for speech recognition and compresses them to transmit these problems.

However, this standard does not take advantage of the fact that the performance of the mobile terminal is improved compared to the time of enactment, so that it can perform more tasks in addition to feature extraction, and there is an inconvenience in that a specific communication method proposed in this standard must be newly implemented. .

An object of the present invention is to provide a distributed speech recognition method and apparatus using phonemes.

In addition, the present invention uses a structure that performs phoneme recognition in the terminal and delivers the result to the server, and the server performs speech recognition on the phoneme recognition result, thereby fully utilizing the computing power of the terminal and without implementing a separate communication protocol. An object of the present invention is to provide a method and apparatus for implementing information retrieval through voice recognition using the Internet search protocol.

In order to achieve the above objects, according to an aspect of the present invention, in a method for recognizing a speech by a terminal coupled with a speech recognition server, receiving a voice query from the outside, the voice feature in the received voice query The method may include extracting, recognizing a phoneme of a voice query corresponding to the extracted voice feature, and transmitting the recognized phoneme to the voice recognition server.

In a preferred embodiment, the voice query received from the outside may be a natural language. The method may further include converting the recognized phoneme into a URL address format and transmitting the converted URL address to the voice recognition server.

According to another aspect of the present invention, in a method of recognizing a voice by a voice recognition server coupled with a terminal, receiving a phoneme signal from the terminal, analyzing the received phoneme signal and included in a voice recognition database Interpreting the received phoneme signal by applying word information and sentence information, sending the interpreted phoneme signal to an external search server, and receiving response information corresponding to the interpreted phoneme signal from the external search server. And it may provide a distributed voice recognition method of the voice recognition server comprising the step of transmitting the received response information to the external terminal.

In a preferred embodiment, the phoneme signal received from the terminal may be transmitted in the form of a URL address. In addition, the word and sentence information of the speech recognition database may be characterized by receiving and storing update information from an external search server. In addition, the response information received from the external search server may be characterized in that the response information to the query content included in the phoneme signal interpreted by the speech recognition server.

According to another aspect of the present invention, in a terminal for recognizing a voice in combination with a voice recognition server, a microphone for receiving a voice query from the outside, a voice feature extraction unit for extracting a voice feature from the received voice query, A distributed voice recognition terminal may include a phoneme recognizer that recognizes a phoneme of a voice query corresponding to the extracted voice feature, and a transmitter that converts the recognized phoneme into a URL address format and transmits the converted phoneme to a voice recognition server. .

According to another aspect of the present invention, in a speech recognition server that recognizes a voice in combination with a terminal, a receiver for receiving a phoneme signal from the terminal, a word included in a voice recognition database by analyzing the received phoneme signal A language interpreter for interpreting the received phoneme signal by applying the information and sentence information, a speech recognition database storing word information and sentence information for analyzing the phoneme signal in the language interpreter, new search word information and sentences in the speech recognition database A knowledge extractor for storing information, a speech manager interpreted by the language interpreter to a search server and transforming the search signal received from the search server into a desired format in the terminal and the interpreted phoneme signal to an external search server Including a transmitting unit to send a A distributed speech recognition server can be provided.

According to the present invention, it is possible to provide a distributed speech recognition method and apparatus using phonemes.

In addition, according to the present invention, the terminal performs phoneme recognition and transfers the result to the server, and the server performs a speech recognition on the phoneme recognition result to fully utilize the computing power of the terminal and implement a separate communication protocol. It is possible to provide a method and apparatus for implementing information retrieval through voice recognition using the existing Internet search protocol.

Next, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings.

1 is a view schematically showing a speech recognition system to which the present invention is applied.

Referring to FIG. 1, a speech recognition system according to the present invention includes a terminal 101, a speech recognition server 103, and a search server 105.

The terminal 101 includes a mobile terminal, and may be connected to the voice recognition server 103 through wires and wirelessly. Although not shown in the figure, a wireless network or a wired network may exist between the terminal 101 and the voice recognition server 103.

The terminal 101 receives a natural language speech query from a user, analyzes the speech, extracts a speech feature, recognizes a phoneme, and transmits the speech to the speech recognition server 103.

In this case, the reason why the terminal 101 does not perform voice recognition for the user's natural language voice query is that the terminal 101 is portable and has a large voice recognition database for performing voice recognition as the mobile terminal 101. And because the capacity and performance are too small to process the speech recognition algorithm.

However, it is not necessary to extract only the speech features for speech recognition presented by the ETSI AURORA standard, and to perform the phoneme recognition on the terminal 101 than the ETSI AURORA standard is specified. This is because the ETSI AURORA standard requires the use of a specific communication method. The terminal 101 according to the present invention converts the phoneme-recognized result into a URL address format and transmits the same, thereby freeing the communication method according to the existing ETSI AURORA standard.

The voice recognition server 103 receives a phoneme signal sent from the terminal 101 and analyzes a natural language question sentence initiated by a user in a phoneme signal based on various words and sentence information included in a recognition target region. It is responsible for transmitting the query to the search server 105 to find a response corresponding to the.

To this end, the speech recognition server 103 should have a large database and high performance. In this figure, the terminal 101 and the voice recognition server may appear to be combined in a one-to-one manner, but a plurality of terminals 101 may be connected to one voice recognition server 103 according to a situation.

The speech recognition server 103 interprets the user's query from the phoneme signal received from the terminal 101 and then inquires the search server 105 for an answer to the query, which queries a search site of a general Internet. It is similar to that. That is, when the sentence recognized by the speech recognition server 103 is transmitted to the search server through the Internet or another network, the search server 105 has the same response as the response to the general search query.

The search server 105 is coupled to the speech recognition server 103 through the Internet, and serves to respond to the query of the speech recognition server 103 using the search server and other Internet information. The search server 105 may be a general internet portal service server.

2 is a flowchart illustrating a process of applying a speech recognition system according to an exemplary embodiment of the present invention.

Referring to FIG. 2, a user first inputs a query sentence in natural language by voice through a terminal (step 201). Then, the terminal extracts the feature of the input voice (step 203), and recognizes each phoneme of the query sentence corresponding to the extracted feature (step 205). Thereafter, the recognized phoneme is transformed into a URL address (step 209), and transmitted to the speech recognition server (step 209).

Here, when the recognized phoneme is converted into a URL, for example, when the voice of 'voice recognition' is recognized, the recognized phoneme is' ㅡ ㅁ ㅅ ㅓ ㅇ ㄴ ㅅ ㅣ ㄱ. In this case, the phoneme corresponding to each phoneme can be expressed as% number or% character of the URL. For example, when searching for "voice recognition" through the portal site Naver, the following URL is used.

'http://search.naver.com/search.naver?query=%C0%BD%BC%BA%C0%CE%BD%C4'

Here, the part expressed in the repetition form of '% number' becomes the part that expresses the Hangul 'voice recognition'.

In this way, when the terminal transmits the recognized phoneme to the voice recognition server, the terminal may send both the address of the voice recognition server and the recognized phoneme into a URL address format.

When the phoneme information transformed into the URL address is received by the voice recognition server (step 213), the voice recognition server analyzes the phoneme based on the received phoneme information (step 215), and the voice recognition combined with the voice recognition server. The word information and the sentence information stored in the database 231 are applied to the phoneme information (step 217) to recognize the meaning of the voice originally initiated by the user (step 219). Then, the information search is requested to the external search database 233 based on the recognized voice. The search database then provides a search result to the speech recognition server in response to the information search request, and the speech recognition server receives the information (step 221) and provides the information to the terminal, and the terminal displays the result (step 211).

3 is a diagram schematically illustrating an internal structure of a voice recognition server according to an exemplary embodiment of the present invention.

Referring to FIG. 3, the voice recognition server 303 is combined with a terminal 301 and a search server 305. This speech transplant server 303 includes a language interpreter 311, a speech recognition database 313, a knowledge extractor 315, and a search manager 317.

The language interpreter 311 is responsible for deriving a complete speech recognition result from the phoneme signal received from the terminal 301 using the speech recognition database 313.

The speech recognition database 313 is a database including word information and sentence information for the language interpreter 311 to recognize speech. The speech recognition database 313 may receive the information extracted from the knowledge extractor 315.

The knowledge extractor 315 is responsible for providing new word information and sentence information to the speech recognition database 313. The provided word information and sentence information are formed by receiving and separately searching for a search word and a search sentence that are updated according to a real-time search rank of a search term searched by the search server 305, and converting the formed information to the voice recognition database 313. to provide.

The search manager 317 is responsible for providing the query sentence received from the language interpreter 311 to the search server 303 and transmitting the search result received from the search server 305 to the terminal 301.

The search manager 317 transforms the query sentence received from the language interpreter 311 into a format that can be received by the search server 305, and transmits a plurality of search results received from the search server 305 to the terminal 301. In some cases, it is transformed to be suitable for display.

4 is a signal flow diagram illustrating a natural language query processing process according to an exemplary embodiment of the present invention.

Referring to FIG. 4, when a natural language query is input to a terminal, a phoneme of a natural language query is recognized (step 401), and the recognized phoneme signal is sent to a voice recognition server (step 403). Thereafter, the voice recognition server (step 405) receiving the phoneme signal performs voice recognition using the phoneme signal (step 407), and determines a searcher to be transmitted to the search server based on the recognized voice (step 409). ). The search query signal is then sent to the search server using the confirmed searcher (step 411), and the search server sends a query response signal in response to the search query signal (step 413). The voice recognition server then analyzes the received Q & A signal, organizes the response into a format suitable for display on the terminal (step 415), and transmits the response signal to the terminal (step 417). The terminal then receives the response signal and displays the response (step 419).

5 is a signal flow diagram illustrating a process of updating word information and sentence information in a speech recognition server according to an exemplary embodiment of the present invention.

Referring to Fig. 5, first, information of a search database is updated at the search server (step 501). The update of the information may be a search word or a sentence that is frequently used by users using the search server, and thus may be a value constantly changing in real time.

When the information in this database is updated, an update information signal including the updated information is sent to the speech recognition server (step 503). The voice recognition server having received the update information signal then extracts update information from the update information signal (step 505) and stores the signal in a voice recognition database (step 507).

The information update of the speech recognition database is frequently searched on the search server because the efficiency of speech recognition is very low when the data capacity of the speech recognition server is large, but the range for speech recognition is not specified for speech recognition. This is to improve the accuracy of speech recognition according to the user's search by updating the search word or the search sentence.

The present invention is not limited to the above embodiments, and many variations are possible by those skilled in the art within the spirit of the present invention.

1 is a view schematically showing a speech recognition system to which the present invention is applied.

2 is a flowchart illustrating a process of applying a voice recognition system according to an embodiment of the present invention.

Figure 3 is a schematic diagram showing the internal structure of a speech recognition server according to an embodiment of the present invention.

4 is a signal flow diagram illustrating a natural language query processing procedure according to an embodiment of the present invention.

5 is a signal flow diagram illustrating a process of updating word information and sentence information in a speech recognition server according to an exemplary embodiment of the present invention.

<Explanation of symbols for the main parts of the drawings>

301: terminal

303: Speech Recognition Server

305: search server

311: language interpreter

313: Speech Recognition Database

315: Knowledge Extractor

317: search manager

Claims (9)

In a method for recognizing speech by a terminal coupled with a speech recognition server Receiving a voice query from the outside; Extracting a speech feature from the received speech query; Recognizing a phoneme of a voice query corresponding to the extracted voice feature; and Transmitting the recognized phoneme to the speech recognition server. Distributed speech recognition method of the terminal comprising a. The method of claim 1, The voice query received from the outside is a natural language Distributed speech recognition method of the terminal characterized in that. The method of claim 1, Converting the recognized phoneme into a URL address format and transmitting the converted URL address to the speech recognition server Distributed speech recognition method of the terminal further comprising. In the speech recognition server coupled with the terminal to recognize the voice, Receiving a phoneme signal from the terminal; Analyzing the received phoneme signal and applying word information and sentence information included in a speech recognition database to interpret the received phoneme signal; Sending the interpreted phoneme signal to an external search server; Receiving response information corresponding to the interpreted phoneme signal from the external search server; and Transmitting the received response information to an external terminal. Distributed speech recognition method of the speech recognition server comprising a. The method of claim 4, wherein Phoneme signal received from the terminal is transmitted in the form of a URL address Distributed speech recognition method of a speech recognition server, characterized in that. The method of claim 4, wherein The word and sentence information of the speech recognition database receives and stores update information from an external search server. Distributed speech recognition method of a speech recognition server, characterized in that. The method of claim 4, wherein The response information received from the external search server is response information about a query content included in a phoneme signal interpreted by the speech recognition server. Distributed speech recognition method of a speech recognition server, characterized in that. In the terminal to recognize the voice in combination with the speech recognition server, A microphone for receiving a voice query from the outside; A voice feature extraction unit for extracting a voice feature from the received voice query; A phoneme recognition unit for recognizing a phoneme of a voice query corresponding to the extracted voice feature; Transmitter for converting the recognized phoneme into a URL address format and transmitting to the speech recognition server Distributed speech recognition terminal comprising a. In the speech recognition server to recognize the voice in combination with the terminal, A receiver which receives a phoneme signal from the terminal; A language analyzer configured to analyze the received phoneme signal and interpret the received phoneme signal by applying word information and sentence information included in a speech recognition database; A speech recognition database storing word information and sentence information for analyzing a phoneme signal in the language interpreter; A knowledge extractor for storing new search word information and sentence information in the speech recognition database; A search manager for transmitting the speech signal interpreted from the language interpreter to a search server and transforming the search signal received from the search server into a desired format in the terminal; A transmitter for transmitting the analyzed phoneme signal to an external search server; Distributed speech recognition server comprising a.

Images