KR0140814B1 - Information Communication Service Menu Voice Control Method - Google Patents

Abstract

The present invention relates to an information communication service menu voice control method for controlling a menu using voice as well as a menu control using a general keyboard when using an information communication service. In order to provide a method of selecting a first step (10 to 30) of searching each screen of the information communication service to detect the recognition candidate; And a second step (40 to 70) of performing voice recognition only on the detected recognition candidates, so that the user can receive the service using not only the keyboard input but also the voice when using the information communication service. Even those who are inconvenient can use this service more conveniently, and it is possible to reduce unnecessary recognition time and recognition error by searching the screen of the information communication service and detecting the recognition candidate that can be recognized on one screen.

Description

Information Communication Service Menu Voice Control Method

1 is a block diagram of a device to which the present invention is applied.

2 is an overall flow chart in accordance with the present invention.

3 is a detailed flowchart of a menu detection process according to the present invention.

4 is a detailed flowchart of a speech recognition and recognition result processing process according to the present invention.

* Explanation of symbols for the main parts of the drawings

1: microphone 2: computer

3: Modem 4: Public Telephone Network (PSTN)

5: ICT service provider

The present invention is not only a menu control by a general keyboard (keyboard) when using an information communication service, such as a high-tel service operated by PC (PC) communication, but also the information communication service using the voice which is the most common communication means of human The present invention relates to an information communication service menu voice control method capable of controlling a menu.

In order to control the information communication service menu using voice, there are two main parts, a menu detection part for detecting candidate words that are the target of voice recognition, and a voice recognition part for performing voice recognition on the detected recognition candidate words. to be.

The conventional method of using an information communication service has a problem in that it is inconvenient for a disabled person to use a menu because a menu is selected using a keyboard.

Disclosure of Invention The present invention devised to solve the above problems is an object of the present invention to provide an information communication service menu voice control method in which anyone can easily select a menu of an information communication service using a voice recognition technology.

In order to achieve the above object, the present invention provides a voice control method applied to a computer having a voice recognition device, comprising: a first step of detecting recognition candidate information by searching a screen of an information communication service; And performing a voice recognition on the detected recognition candidate information to perform an information communication service.

Hereinafter, with reference to the accompanying drawings will be described an embodiment according to the present invention;

1 is a block diagram of a device to which the present invention is applied, a personal computer (PC) or workstation (2) equipped with a microphone (1), a sound card or a digital signal processing board, which is a device capable of receiving a voice, and a digital signal. And a telecommunication service provider (Hitel, Cheon Lian, etc.) (5) connected to the computer (2) via a modem (3) and a modem (3) and a public switched telephone network (PSTN) (4). Equipped.

2 is an overall flow chart according to the present invention, with a hightel connection as an example.

First, when the hightel connection 10 is established through the modem, the data transmitted from the hightel is received (20) and analyzed by one screen unit to detect a menu (30). Based on the result of the menu detection, voice recognition for controlling the hightel menu is performed. (40).

The recognition result is displayed on the screen and transmitted to the hightel at the same time (50), and the hightel screen according to the result is received (20).

If the recognition result is a command to terminate the hightel connection (60), the hightel connection is terminated, and execution of all programs is terminated (70).

Hereinafter, each part will be described in detail with reference to FIGS. 3 and 4.

Since all information of HITEL is arranged in one screen unit and one screen input is required for each screen, every screen received from HITEL is required in the menu analysis and detection process. Analyzes every screen received from the PC and detects the type of command input required in the process.

By analyzing each screen of HITEL, it is possible to find out the candidate candidates for speech recognition in the current process, and also in which process the function for speech recognition should be performed.

The reason for detecting candidate words for speech recognition is that the time required for speech recognition is too long when about 900 services provided by HITEL are treated as candidate words every recognition. Hitel service will not be available properly, and the recognition rate will be lowered, making it difficult to control menus by voice.

Therefore, by detecting the candidate to be recognized on each screen, it is possible to reduce the time required for recognition and to reduce the recognition error.

In the present invention, the recognition candidate actually detected for each screen is a menu control command for controlling the number of each menu for selecting the type of the hightel service and the flow of the hightel service.

The reason why the name of each service of Hitel service is excluded from recognition candidate is that it is not easy to build the recognition standard model because each service of Hitel can change depending on the circumstances of the operator, the increase of service type and the suspension of existing service. In addition, it is not easy to build a recognition standard model for all 900 services. Therefore, we used a number that can cope with changes in all kinds of services and is easy to build a recognition standard model.

In addition, 16 menu commands were used, including 12 commands that can be used on almost all screens and are essential when using HITE.

When explaining the actual method of detecting the hightel menu, the hightel uses ASCII code as the basic communication protocol, and the control that can be used in addition to the ASCII code to solve the incompatibility of the terminal device. The code (ANSI X3, 64) is used together. Hitel uses only control codes for basic screen configuration, and no code is provided to indicate any special content or menu.

Therefore, there is no problem in configuring a screen with a code transmitted from the hightel, but it is not easy to detect a specific menu or detect a specific screen.

However, since the menu of the current HITEL has a certain format, it is possible to detect desired information or screen by analyzing the format.

Recognition target number sound (menu number detection) method is as follows.

Almost all (about 90%) menu numbers in Hattel take the form of a period (.) After the Arabic numerals, so when a screen of Hitel is received (101) it detects a period (.) (102).

Most of the menu numbers (90%) not accompanied by a period (.) Are enclosed in square brackets ([]) and thus detected (103).

Each menu number is separated using one or more space or tab or linefeed characters, so use it to start searching for a new number.

Since the screen containing the number indicating the database index contains the word ID in the second line from the top, the first screen in which this is detected is in the form of the DB index number + space + user name. This form always appears immediately after the Linefeed character, thus detecting the DB index number (104).

In order to perform continuous speech recognition of the numerical sound, the number of digits that are detected are analyzed to determine the number of candidates to be recognized at each digit (eg 10,000, thousand, hundred, ten, ten). Sounds are separately detected separately (105).

After that, the method of detecting the target screen control word is as follows.

Since Hitel is based on one screen, each screen needs screen control to select the contents of the next screen.

In order to detect such a screen control word, the hightel detects a control command if it is a screen in the form of a number / command (F, B, P, GO, X) that appears at the end of almost all screens (106, 107).

Hitel has screens that require only one or two inputs, depending on the service.

By detecting such screens, screens which deviate from the main rules of menu detection can be controlled by voice.

A screen requiring input of either yes (Y) or no (N) is found by detecting the form of? (Y / n) (108, 109).

For screens that only require the entry of the ENTER key, press [ENTER] to find it and store only the Enter in the Recognition Candidate buffer (110, 111).

In this way, after detecting the necessary menus on one screen of the hightel, the detected candidates are stored in the recognition candidate buffer (112), and then the detected target words for the search or screen control of the menu to be serviced are searched. Speech recognition will be performed.

To control the menus of HITEL by voice, you need to be able to recognize at least a number of sounds and 10 commands within a maximum of 2 seconds (a time when you will not be bored) and provide continuous numeric sound recognition for more convenient use. shall.

Various modern speech recognition technologies (DTW method, HMM method, Neural Net method and two or more methods can be used to recognize the performance of the system within 2 seconds with the recognition performance of about 95% or more. .

4 is a detailed flowchart of a speech recognition and recognition result processing process according to the present invention.

Explaining how to use the recognition candidates detected in the menu detection process during recognition, the number candidate or number + control command or control command, which is the recognition candidate detected on the current screen of Hitel, is added to the recognition candidate buffer to be recognized on the current screen. If the voice is input after storing (112), the recognition is performed only for the recognition candidate (201).

That is, the recognition time is reduced and the recognition error can be reduced because the recognition is performed only on the recognition candidate words detected on the current screen instead of the recognition on all 27 recognition words.

After outputting the number and control command buried in the recognition result on the screen (203) and transmits the output result to the hightel (204).

Table 1 below shows the keyboard input corresponding to the entire words to be recognized used in the present invention.

The present invention as described above has the following effects.

First, when using the information and communication services, as well as keyboard input can be used to receive the service, even those who are uncomfortable to use the keyboard can be used more comfortably.

Second, it is possible to reduce unnecessary recognition time and recognition error by searching the screen of the information communication service and detecting the recognition candidate recognizable on one screen.

Claims (7)

A voice control method applied to a computer equipped with a voice recognition device, comprising: a first step of detecting recognition candidate information by searching a screen of an information bean service; And performing a voice communication on the detected recognition candidate information to perform an information communication service. The method of claim 1, wherein the detecting of the recognition candidate information in the first step comprises: detecting a number of a menu to be recognized; And a fourth step of detecting a recognition target screen control word capable of controlling the flow of the information communication service. 3. The method of claim 2, wherein the third step comprises: a fifth step of detecting a period since the menu number takes the form of a period after the Arabic numeral; A sixth step of detecting the square brackets because the menu number not accompanied by a period (.) Is surrounded by square brackets ([]); Each menu number is separated using one or more space, tab, and linefeed characters, so the menu number can be retrieved using space, tab, and linefeed characters. Seventh step; And an eighth step of analyzing each digit of each detected menu number. According to claim 3, In the seventh step, since the screen including the number indicating the database (DB) index includes the word identifier (ID) in the second line from the top, the identifier word (ID) is detected In the screen shown above, the form of database index number + space + user name starts from the third line, and the database index number + space + user name is always used. An information communication service menu voice control method, characterized in that detection is performed using what appears immediately after a linefeed character. The method as claimed in claim 2, wherein the fourth step comprises: a fifth step of detecting a control command and storing the control command in a recognition candidate buffer when the screen is in the form of a number / command (F, B, P, GO, X); A sixth step of detecting one of yes (Y) and no (N) and storing it in the recognition candidate buffer if a screen requiring any one of yes (Y) and no (N); And a seventh step of storing an ENTER in the recognition candidate buffer when the screen is requested to input an ENTER key. The method of claim 1, wherein the voice recognition process of the second step comprises: a ninth step of performing recognition on recognition candidate information when receiving a recognition candidate voice; And a tenth step of transmitting the outputted result to the telecommunication service provider after outputting a number and a control command matching the recognition result to the screen. The information communication service menu voice control method according to claim 6, wherein the recognition candidate information of the ninth step comprises 11 numeric tones and 16 menu control commands as recognition objects.