KR20030039956A - A Method for Controlling Data Using Distributed Voice Processing Device and The System Thereof - Google Patents

Abstract

PURPOSE: A method and a system for controlling data by using a distributed voice processing device are provided to use a voice interface for distributed processing of voice in a terminal and a server, so as to download a Web page or a data file through voice input in a mobile terminal and enable the other party to receive data in characters for a character message inputted through voice. CONSTITUTION: The wireless Internet is connected through a mobile terminal(10) to select a menu for voice recognition, and selected contents are transmitted to a Web server(30)(S1,S2). The Web server(30) transmits a VXML(Voice eXtensible Markup Language) tag for the received contents to the mobile terminal(10)(S3). The mobile terminal(10) confirms the VXML tag and operates a voice browser, to prepare for voice receiving(S4). Feature extraction and vector quantization are performed for voice signals inputted by a user, and the voice signals are turned to data with the combination of voice vector, then the voice data are transmitted from the mobile terminal(10) to a natural sound processing server(40)(S5). The natural sound processing server(40) transmits the voice data from the mobile terminal(10) and grammar to a voice recognition server(70), by using an internal grammar database(S6). The voice recognition server(70) compares the received voice data with words in grammar, to notify a result value to the natural sound processing server(40)(S7). According to the received result value, the natural sound processing server(40) requests a Web page to the Web server(30) or transmits required data to the mobile terminal(10) if a Web page is not necessary(S8). The Web server(30) searches the corresponding Web page and transmits the Web page to the mobile terminal(10)(S9). And TTS(Text-To-Speech)-processed data are transmitted to the mobile terminal(10)(S10).

Description

A method for controlling data using distributed voice processing device and the system thereof

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method and system for controlling data using a distributed speech processing apparatus. More particularly, the present invention relates to a method and system for controlling data using a speech interface distributed by a terminal and a server.

In the method of using a short message service using a mobile terminal such as a mobile phone, the user has to press the buttons on the keypad to assemble the characters one by one, and to correct when entering the wrong one.

In addition, even when accessing a web server using a wireless terminal capable of wireless Internet, it is cumbersome to operate it with buttons on the keypad. Therefore, the user has to pay additional cost as much as the access time.

Thus, recently, a voice recognition system has been developed, and a technology for replacing a character with a voice by a user inputting the voice has been disclosed.

However, in the conventional mobile terminal, the mobile terminal and the voice recognition method are used for processing the voice recognition and synthesis function by constructing the voice recognition system in the mobile terminal, and the voice recognition vocabulary is limited by the capacity of the memory or the size of the mobile terminal. Is not only limited, but also has problems in real time speech synthesis, which has technical limitations in practical use.

As described above, the conventional technology for implementing a voice portal on a mobile terminal is not only too heavy, but also a database upgrade (e.g., a weight to be equipped with a speech recognition engine and a text-to-speech (TTS) engine in the mobile terminal). During the DB upgrade, there was a problem that the traffic was concentrated and the speed was reduced or the communication was disconnected.

In addition, the conventional mobile communication network has a voice-oriented communication structure, and has a gateway (aka voice portal gateway) structure for disassembling, decoding, and converting a Voice Extensible Markup Language (VXML) document. As a result, it is difficult to pursue a true voice portal structure and there are many constraints in building a service.

Therefore, a method has been attempted to solve the hardware problem of the mobile terminal by building the voice recognition system in the mobile terminal. However, this method is also a method of improving the size and capacity limit of the mobile terminal, and in particular, the recognition rate is lowered due to noise or channel noise of the voice signal received through the mobile terminal and the mobile communication network to perform a large number of voice commands. Not only that, there is a problem that the communication is overloaded due to a large number of connections.

As a result, the voice processing system or method according to the related art has many problems such as not only a mobile communication company but also a lot of inconvenience to a terminal manufacturer and a user of a mobile terminal.

Accordingly, an object of the present invention is to provide a data control method and a system using a voice interface in which voice recognition processing is distributed in a terminal and a server by eliminating the above disadvantages.

1 is a block diagram showing a data control system using distributed speech processing.

2 is a block diagram schematically showing a data processing method using distributed speech processing.

3 is a block diagram illustrating input and transmission of a voice short message (VSMS) using a distributed speech processing system.

Description of the main parts of the drawing

10: wireless terminal, mobile terminal 20: switching center, switchboard

30: web server 40: natural sound processing server

50: file server 60: database

70: speech recognition server 80: natural sound processor

90: receiving mobile terminal

According to the present invention, there is provided a data control method and system using the distributed speech processing apparatus.

Specifically, it is the same as that of description of a claim described below.

Hereinafter, a data control method and a system using the distributed speech processing apparatus according to the present invention will be described.

1 is a block diagram showing a data control system using a distributed speech processing apparatus.

Distributed speech processing consists of distributed speech recognition and distributed speech synthesis. In the conventional method of performing speech recognition from a speech signal transmitted using a speech channel, the recognition rate is lowered due to the effects of low bit rate coding and channel distortion. Therefore, in the present invention, instead of transmitting voice coding data, distributed speech recognition of distributed speech processing extracts parameters from speech, performs vector quantization, and transmits packets to data channels instead of voice channels. .

The natural sound processing server 40 controls the flow between all the components necessary to implement speech recognition on the user's mobile terminal 10, and provides synchronization between the recognized speech data and the web page. It also selects a server with a low load from multiple servers or databases and sends the server's address to the mobile terminal. Using this address, it controls to receive service from the server with the least load.

The mobile terminal 10 receives the voice recognition request and loads the corresponding grammar, and the feature extraction and the vector quantized voice data are transmitted to the voice recognition server 70 together with the corresponding grammar in the mobile terminal 10. A job, a job of executing data corresponding to the result recognized by the engine in the voice recognition server 70, and the like are performed. Here, the feature extraction refers to an operation of extracting a sample such as frequency or amplitude from the voice signal input by the user through the mobile terminal. In other words, it extracts only the samples necessary to download the necessary data. The mobile terminal 10 is a concept including a mobile phone, a wireless Internet capable computer terminal (laptop, etc.), a personal digital assistant (PDA), and the like.

On the other hand, the database 60 has a dialog ID (Dialog ID) in the place where the voice recognition is required in the web page to perform voice recognition in conjunction with the web server 30. The dialog ID informs the web browser that the voice recognition function can be used, and each dialog ID is mapped with grammar and web content. Thus, the database 60 manages the mapping between the dialog ID assigned to each menu, the corresponding grammar file, and the web page. The database 60 also utilizes tables used by other servers and corresponding management tools as a database.

The file server 50 is where all the resources (resources) needed by the natural sound processing server 40 and the voice recognition server 70 are stored. A simple scenario may be stored in the natural sound processing server 40 itself, but in a high capacity scenario, it may be implemented as a duplicated file server 50.

The voice recognition server 70 is a server in charge of actual voice recognition. The mobile terminal 10 receives voice data that has already been subjected to feature extraction and vector quantization from the natural sound processing server 40 to perform a voice recognition function by comparing the transmitted grammars. The recognized result is then delivered to the natural sound processing server 40 in the form of a list.

The mobile terminal 10 is equipped with a wireless internet browser and a voice recognition chip (for example, a DSP, etc.) to perform a voice recognition function of the user. The voice recognition function is activated when the voice recognition button mounted on the terminal is pressed, and the received voice command is transmitted through the data channel to the natural sound processing server 40 through feature extraction and vector quantization. The sound processing server 40 receives the result of processing the voice recognition process together with the file server 50, the voice recognition server 70, and the like.

A data control method using distributed speech processing according to the present invention will be described with reference to FIG. 2.

In FIG. 2, the switching center 10 (see FIG. 1), which is a communication connection means between the natural sound processing server 40 and the mobile terminal 10, the web server 30 and the mobile terminal 10, is omitted for the purpose of explanation. The horizontal numbers indicate the data control process using distributed speech processing.

In addition, the file server 50 and the database 60 may exist independently as shown in FIG. 1, but may be included in the natural sound processing server 40 when the capacity thereof is small. Of course, the speech recognition server 70 may also be included in the natural sound processing server 40.

However, since the gist of the present invention is data processing (control) using distributed voice processing, the voice recognition server 70 is shown as an independent server.

Naturally, depending on the type of data, there may be two or more servers including the natural sound processing server 40, but only one is shown for understanding of the description.

First, the user starts with the wireless Internet connection process S1, which accesses the wireless Internet by using a button on his keypad with the mobile terminal or the wireless terminal 10.

Then, when the user selects a submenu for voice recognition, the selected content is transmitted to the web server 30. This is called a voice service menu selection process (S2). Then, the web server 30 receives the selected contents of the voice recognition submenu, and performs a voice extension generating language menu transmission process S3 for transmitting a voice extension generating language (VXML) tag to the mobile terminal 10.

When the mobile terminal 10 receives the voice extension generating language tag, the mobile terminal 10 checks this and prepares to receive the voice by driving the voice browser [browser driving process (S4)]. After that, the user selects the voice menu and transmits the voice [voice menu selection and voice transmission process (S5)].

In other words, in the voice menu selection and voice transmission process (S5), the menu name that the user wants to select while the user is connected to the wireless Internet is spoken by voice, for example, "ringtone download" or "text message transmission". In particular, the voice recognition menus among the selectable menus are voice recognition tags in the web page, and the browser of the terminal uses the tag to drive the voice button function of the terminal.

In addition, in the voice menu selection and voice transmission process (S5), the user's voice is converted into a combination of voice vectors through feature extraction and vector quantization by a speech recognition preprocessor in the terminal 10, and then the natural sound processing server 40 Is sent to. At this time, the webpage ID indicating the webpage on the terminal screen is transmitted together. The web browser of the terminal waits for the web page to be transmitted from the web server 30 after transmitting the above contents.

Next, the natural sound processing server 40 finds the corresponding grammar from the Grammar DB mounted in the natural sound processing server 40 using the page ID received from the terminal 10, and then, from the terminal 10. The received speech vector data and grammar are transmitted to the speech recognition server 70 (speech recognition processing request process (S6)). The user interface control portion of the grammar database and the application program may be implemented as a voice confirmation generation language.

The speech recognition server 70 compares the speech vector data with words in the grammar, searches for the most similar words, and notifies the natural sound processing server 40 of the result value (recognition result notification process S7).

The natural sound processing server 40 requests the web server 30 when a new web page (or a data file such as a music file) is needed according to the algorithm of the application after being received by the recognition result notification process S7. If the web page request is not required, the requested data is immediately transmitted to the mobile terminal 10 (web page request process S8).

Then, the web server 30 having received the request from the natural sound processing server 40 transmits the web page to the mobile terminal 10 and transmits the text to the mobile terminal 10 [S9]. A process (S10) of transferring the processed data (TTS: Text To Speech) is performed. That is, the natural sound processing server 40 matches the character-to-voice conversion data processed by the speech recognition server 70 to the menu of the web server 30. The transmission of the voice conversion is synchronized to the mobile terminal 10. The received data is output as voice.

However, when the web page request is not necessary in the web page request process (S8), for example, in the case of a music file, the music file is directly transmitted to the mobile terminal 10 and the fact that the voice is transmitted is output.

Therefore, when data is music, music is played, and when it is a video, the video is processed continuously so that the user can watch the video continuously on the display device.

For example, according to the menu name input in the voice menu selection and voice transmission process (S5), for example, in the case of the "hometown of Nahuna" command, the terminal 10 performs a music file (or web page) related to the music data through the above process. And begin playing continuously. Then press "No. 1 or 2 to save Nahuna's hometown". If you save the music in the memory and listen to it later or select it as a ringtone, the downloaded music will be output as a ringtone later.

In addition, when the selected menu is downloaded, the mobile terminal 10 outputs music in the case of music according to its data property, and in the case of a video (movie, etc.), the video is processed so that the user can view it.

Optionally, if the memory of the mobile terminal 10 is insufficient when downloading and outputting the audio output or the video output, the path name may be stored and later operated by the stored path name to directly output music or video. Can be.

3 is a block diagram illustrating the input and transmission of a voice short message (VSMS) using a distributed speech processing system as an example of another modification of the present invention. In order to help the understanding of the present invention, a general transmission method and system structure of a mobile communication network will be omitted. In the modified example of the present invention, processes S1 to S6 of FIG. 2 are performed in the same manner. However, in the voice service menu selection process (S2), the menu is a voice short message (VSMS) menu.

3 shows a natural sound processor 80 as compared to FIG. 2. The natural sound processor 80 is a server or device equipped with a program suitable for Hangul grammar, such as soft sound processing, spelling correction, and jump correction by analyzing the input voice data. However, the language is not limited to Korean and includes English, Spanish and Japanese. In other words, when a Japanese uses a mobile terminal 10 to transmit a voice short message to a third party's mobile terminal or e-mail address, using the modified example according to the present invention, the voice data is a natural sound processor ( 80) is calibrated and transmitted to the third party's terminal.

In particular, in the modified example of the present invention, the speech-to-text (STT: Speech To Text) data is transmitted to the first mobile terminal 10 in which the corrected text is correct and the user can check the function before sending it. Can be added.

1 to 3, a modification of the present invention will now be described. First, when the user accesses the wireless Internet using his mobile terminal 10 and selects the voice recognition western menu (steps S1 and S2). The selection signal is transmitted to the web server 30 connected to the switching center 20 of the mobile communication network via the mobile communication network. After receiving the selection signal, the web server 30 transmits a voice extension generating language (VXML) tag (step S3). Here, the voice recognition western menu is a concept including a voice SMS western menu.

Then, the user drives the voice browser (in step S4) and inputs a voice message, for example, "Hello," by voice. The voice signal is connected from the mobile terminal 10 to the switch 20 of the mobile communication network. It is transmitted to the natural sound processing server 40 (step S5 above), and the server 40 finds the speech vector data and the grammar received from the terminal 10 in a built-in grammar database (Grammar DB). 70) (S6 process).

The speech recognition server 70 compares the speech vector data with the words in the grammar, searches for the most similar words, and transmits the result value to the natural sound processor 80. [Result transfer process (S7)], the natural sound processor 80 analyzes the input data and transmits the natural data to the natural sound processing server 40 in accordance with the grammar such as soft-tone processing, spelling correction, and skipping correction (natural sound processing process S8).

The natural sound processing server 40 then converts the natural sound processed data into text and transmits it to the receiving mobile terminal 90 designated by the user of the mobile terminal 10 that is initially connected. (S9)]. In the voice-to-character conversion and transmission process (S9), the text of text may be directly transmitted to the receiving mobile terminal 90, and then transmitted first to the transmitting mobile terminal 10 so that the user checks the text (i.e., a short message). If so, it can transmit to the receiving side mobile terminal 90. Here, the receiving side mobile terminal 90 is a concept including a mobile phone, a notebook, a personal portable terminal and the like.

In some cases, the natural sound processed data is directly transmitted to the mobile terminals 10 and 90, and the mobile terminals 10 and 90 can convert the text into text.

Voice SMS transmission according to a modification of the present invention is to change the voice to a short message service (SMS) text through a natural sound processing process and send it to the other party's mobile terminal or e-mail address. Therefore, the designation of the other party's mobile terminal or the e-mail address can be specified at any time during the voice SMS transmission process, and the end can be output by voice when the reception ends.

In addition, although the voice recognition server 70 and the natural sound processor 80 are illustrated as separate servers or devices, the voice recognition server 70 and the natural sound processor 80 may be included in the natural sound processing server 40 to implement a variation of the present invention with one server or device. Is not required by those skilled in the art.

According to the present invention, by providing a voice directly input through the mobile terminal to provide the effect of downloading the necessary web page or data file.

In another aspect, the present invention provides the effect that the other party can receive the data by text input by voice input without pressing the button on the keypad cumbersomely.

Claims (5)

In the data control method using a distributed speech processing device, Accessing the wireless Internet with the mobile terminal 10, selecting a menu for voice recognition, and transmitting the selected content to the web server 30; The web server 30 transmits a voice extension language (VXML) tag to the mobile terminal 10 in response to the received content; The mobile terminal 10 checking the voice extension language tag and preparing a voice input by driving a voice browser; A process of transmitting the speech data from the mobile terminal 10 to the natural sound processing server 40 after the speech signal input by the user is data-formed through the feature extraction and the vector quantization process and then combined with the speech vector; Transmitting, by the natural sound processing server 40, the voice data and the grammar received from the mobile terminal 10 to the voice recognition server 70 using a self-equipped grammar database; The voice recognition server 70 comparing the received voice data with words in the grammar and notifying the natural sound processing server 40 of the result value; The natural sound processing server 40 requests the web server 30 when a web page is needed according to the received result value, and transmits the requested data directly to the mobile terminal 10 when it is not needed. Process, The web server 30 finds the web page and transmits it to the mobile terminal 10. And a process of transmitting the text-to-speech data to the mobile terminal (10). The mobile terminal 10 stores the path name and the natural sound processing server 40 is connected to the mobile terminal 10 when the data received by the mobile terminal 10 has a large capacity. And transmitting the stored data to the mobile terminal (10) when a request is received. In the data control method using a distributed speech processing device, Accessing the wireless Internet with the mobile terminal 10, selecting a menu for voice recognition, and transmitting the selected content to the web server 30; The web server 30 transmits a voice extension language (VXML) tag to the mobile terminal 10 in response to the received content; The mobile terminal 10 checking the voice extension language tag and preparing a voice input by driving a voice browser; Transmitting the log sound data of the natural sound processing server 40 from the mobile terminal 10 after the voice signal input by the user is converted into a combination of voice vectors through feature extraction and vector quantization; Transmitting, by the natural sound processing server 40, the voice data and the grammar received from the mobile terminal 10 to the voice recognition server 70 using a self-equipped grammar database; The speech recognition server 70 comparing the speech data with the words in the grammar and transmitting the result value to the natural sound processor 80; The natural sound processor 80 analyzes the input data, processes it as natural sound, and transmits it to the natural sound processing server 40; The natural sound processing server 40 converts the natural sound processed data into text and transmits the data directly to the receiving mobile terminal 90, or after transmitting the text to the mobile terminal 10 to confirm the mobile terminal 10. And receiving the signal and transmitting the signal to the receiving side mobile terminal (90). In a data control system using a distributed speech processing device, After accessing the web server 30 and receiving the tag of the voice extension language, the parameter is extracted from the received voice signal, vector quantized, and preprocessed by the natural sound processing server 40 connected to the mobile communication network as a data channel. A mobile terminal 10 for transmitting data; A web that transmits a tag of a voice extension language to the mobile terminal 10 according to a connection request, and transmits a web page to the mobile terminal 10 upon receiving a web page transmission command from the natural sound processing server 40. Server 30, Receives preprocessed voice data from the mobile terminal 10 and transmits the preprocessed voice data to the voice recognition server 70 with the corresponding grammar, and receives the processed value from the voice recognition server 70 to synchronize with the web server 30. A natural sound processing server 40 for transmitting or storing necessary web pages, data or paths to the mobile terminal 10 and controlling voice recognition to provide the mobile terminal 10 with the necessary voice at the end of the step. , A speech recognition server for receiving feature extraction and vector quantized data from the natural sound processing server 40 together with a grammar, performing a speech recognition function by performing a comparison process, and transmitting the result value to the natural sound processing server 40. A data control system using a distributed speech processing device for implementing the method according to claim 1, characterized in that it is composed of (70). In a data control system using a distributed speech processing device, After accessing the web server 30 and receiving the tag of the voice extension language, the parameter is extracted from the received voice signal, vector quantized, and preprocessed by the natural sound processing server 40 connected to the mobile communication network as a data channel. A mobile terminal 10 for transmitting data; A web server 30 for transmitting a tag of a voice extension language to the mobile terminal 10 according to a connection request; Receives the preprocessed voice data from the mobile terminal 10 and transmits the preprocessed voice data to the voice recognition server 70 together with the corresponding grammar, and receives the natural sound processed result value from the natural sound processor 80 to convert the voice data into text data. A natural sound processing server 40 for transmitting to the mobile terminal 10 or the receiving mobile terminal 90; A speech recognition server for receiving feature extraction and vector quantized data from the natural sound processing server 40 together with a grammar, performing a speech recognition function by performing a comparison process, and transmitting the result value to the natural sound processing 40. 70, 3. A natural sound processor (80) for processing natural sound of voice data received from the voice recognition server (70) and transmitting the result to the natural sound processing server (40). Data control system using a distributed speech processing device for implementing the method.