KR100277065B1 - The recording word control method on voice recogition voice mailing system - Google Patents

Abstract

end. TECHNICAL FIELD OF THE INVENTION

Speech Recognition A method of constructing a database of speech recognition algorithms and a list of words to be recognized.

I. The technical problem to be solved by the invention

A method for automatically generating and deleting recognition data corresponding to a word to be changed when building a database of a speech recognition algorithm and changing a word list to be recognized in a voice recognition system.

All. Summary of Solution of the Invention

When the automatic time conversion mode is set, the method of automatically recognizing the recognized word in the voice recognition voice mail system requires a word list transmission to a host that provides the recognized word through an external connection when the current time becomes the set time of the recognized word conversion. Transmitting the signal, temporarily storing the word list in the memory unit when receiving the word list through the external connection unit, and reading and comparing the word list of the data memory storing the received word list and voice recognition data. A comparison process, a selection process of selecting data to be converted among words to which an identification code is assigned, and a process of generating and deleting according to an identification code assigned to a conversion word if a word to be converted is found as a result of the comparison. How to change the recognized words in the voice recognition voice mail system.

la. Important uses of the invention

An apparatus for recognizing a voice from a caller who is received in a voice recognition voice mail system that changes frequently in voice recognition data and provides predetermined information.

Description

How to change the recognition word in voice recognition voicemail system {THE RECORDING WORD CONTROL METHOD ON VOICE RECOGITION VOICE MAILING SYSTEM}

The present invention relates to a method of changing a voice recognition word of a voice recognition voice mail system, and more particularly, to a method of automatically generating and deleting phoneme files in a voice recognition voice mail system.

In general, a voice recognition voice mail system includes a voice mail system for delivering a voice message and image data to a specific subscriber to which a voice of a user connected to a call is subscribed to the voice mail, and an addition of the voice mail system. A voice recognition system for analyzing a voice signal of a user to which a call is connected to provide a service and providing predetermined information corresponding to the analyzed voice is combined. The voice recognition voice mail system as described above includes a phoneme file corresponding to words necessary for voice recognition in a database. The phoneme file refers to a file classified by combining English alphabets and numbers in correspondence with Korean consonants and vowels. The administrator of the voice recognition voice mail system searches for words to be added and deleted, and the words to be added to the database among the searched words should be generated using a personal computer (PC) or a workstation. In addition, the administrator has to construct the phoneme file generated in the database of the voice recognition voice mail system, there is an inconvenience to delete the words to be deleted from the database one by one.

Accordingly, an object of the present invention is to provide a method for automatically generating a phoneme file for voice recognition in a voice recognition postal system.

The present invention for achieving the above object is a method for automatically changing the recognition word in the voice recognition voice mail system, if the automatic time conversion mode is set when the current time is the recognition word conversion setting time recognized words through the external connection unit Transmitting a word list transmission request signal to a host providing a word, temporarily storing the word list transmission request signal from the host through the external connection unit, and storing the received word list and voice recognition data; A comparison process of reading and comparing a word list of a memory to give an identification code, a selection process of selecting data to be converted among words to which the identification code is assigned, and assigning the converted word if a word to be converted is present as a result of the comparison Voice recognition sound, characterized in that the process consists of creating and deleting according to the identification code Characterized by the words constituted by any recognized change in the way the postal system.

1 is a block diagram of a voice recognition voice mail system to which the present invention is applied.

2 is a control flowchart according to a word correction time setting according to an embodiment of the present invention.

3 is a control flowchart of automatic recognition word change according to an embodiment of the present invention.

Hereinafter, the present invention will be described in detail with reference to the accompanying drawings.

1 is a block diagram of a voice recognition voice mail system to which the present invention is applied.

The control unit 111 of the voice recognition voice mail system controls the overall operation of the voice mail system, and in particular, performs control according to generation and deletion of a phoneme file when the recognition word is changed according to the present invention. The voice processor 112 receives a voice signal through a line formed through a public switched telephone network (IST: Intergrated Services Digital Network) 100 under the control of the controller 111, and converts the received voice signal into a digital signal. The voice data output to the disk controller 115 or digitized and stored in the disk 116 is received through the disk controller 115, converted into an electrical voice signal, and output to the public network 100. In addition, the voice processing unit 112 detects DMF (Dual Tone Multi Frequency) received from the public network 100 and outputs it to the control unit 111, when providing a voice recognition service according to the present invention under the control of the control unit 111 The received voice signal is digitized and output to the voice recognizer 120.

The speech recognition unit 120 includes a speech recognition controller 121, a speech feature extraction unit 122, a word search unit 123, and a data memory 124. The voice feature extractor 122 extracts the feature of the digitized voice signal received from the voice processor 112 and outputs the feature to the word search unit 123. The word search unit 123 compares and searches the same data among the voice signal received by the voice feature extraction unit 122 and the voice signal stored in the data memory 124 and outputs the same data to the voice recognition unit 121. The data memory 124 stores a phoneme file required for voice recognition of the voice recognition unit 120, and adds a phoneme file generated by the control of the voice recognition controller 121 to the memory or stores the phoneme file. Delete it. The voice recognition controller 121 controls the overall operation of the voice recognition unit 120, and outputs a signal corresponding to the retrieved voice data to the controller 111 when the voice recognition service is provided. In addition, when the voice recognition word is added and deleted, the phoneme file is stored and deleted in the data memory according to the data received from the controller 111.

The memory unit 113 stores a program necessary for the operation of the voice recognition voice mail system, and in accordance with the present invention, a phoneme file for Korean consonants and vowels for generating phoneme files of additional words for voice recognition according to the present invention. Save it. The external connection unit 114 connects the voice recognition voice mail system and an external device through a local area network (LAN) to perform data communication. An external device connected to the external connection unit 114 may include an information providing host computer for providing an additional service such as stock information, a workstation for remotely monitoring and controlling the operation of the voice recognition voice mail system, and the like. have. The disk controller 115 controls to store, read, and delete voice messages and image data received by the voice recognition voice mail system through the public network 100 under the control of the control unit 111, and the disk 116 controls the disk. The voice signal received by the control of the controller 115 is stored and deleted.

2 is a flowchart illustrating a word correction time setting according to an embodiment of the present invention. Hereinafter, the present invention will be described in detail with reference to FIGS. 1 and 2.

In step 200, the control unit 111 checks whether a recognition word correction time setting control signal is received through the external connection unit 114. The control unit 111 proceeds to step 202 when the test result recognition word correction time setting control signal is received, and proceeds to step 210 when the recognition word correction time setting control signal is not received to perform the corresponding function and ends. In step 202, the control unit 111 receives a modified word correction time through the external access unit 114, and proceeds to step 204. The control unit 111 checks whether a confirmation signal is received from the external connection unit 114 in step 204. When the test result confirmation signal is received, the controller 111 proceeds to step 206 and, if not, proceeds to step 202 to receive the modified word correction time data again. The control unit 111 stores the received word correction time data in the memory unit 113 in step 206 and proceeds to step 208 to set and terminate the automatic time conversion mode. Through the above process, it is possible to change the phoneme file by setting the modification time of the phoneme file required for voice recognition to a time zone where the user is small, such as at night.

3 is a control flowchart when automatic recognition word change according to an embodiment of the present invention. Hereinafter, the present invention will be described in detail with reference to FIGS. 1 to 3, and the process of FIG. 3 is performed while the automatic time conversion mode is set by FIG. 2. Here, the automatic conversion of words may be performed by first receiving all words from a host and comparing the words stored in the voice recognition voice mail device one by one with a word to add and a word to be deleted, and secondly, a word requiring change by the host. It is also possible to add and delete the received word by outputting only data corresponding to. Here, the first method will be described.

In step 250, the controller 111 provides a voice mail service and a voice mail service according to voice recognition. In step 252, the controller 111 checks whether the current time and the voice recognition word modification time set by FIG. The controller 111 proceeds to step 254 when the current time and the voice recognition word modification time and the current time are the same as the test result, and continues with step 250 when the current time is the same. When the controller 111 proceeds from step 252 to step 254, the controller 111 transmits a word list request signal to a host connected to provide a voice recognition service through the external access unit 114. Here, the host for providing voice recognition refers to a host computer for providing stock information, horse racing information, and the like. The control unit 111 receives the word list through the external access unit 114 in step 256 and temporarily stores the word list in the memory 113 and proceeds to step 258. The control unit 111 reads out the word list received in step 258 and the word list stored in the data memory 124, and proceeds to step 260. Except for the conversion word to be added and the conversion word to be deleted, a word other than the conversion word is deleted from the memory unit 113 in the comparison process of step 258. In addition, the additional word may set the identification code to 1 and the word to be deleted may be set to 0 to distinguish the deleted word from the additional word. The control unit 111 checks whether a conversion word exists in step 260. In this case, the presence or absence of the converted word may be checked by checking whether there is data stored in the memory unit 113 by being given an identification code in the word list received from the host through the external connection unit 114.

The control unit 111 proceeds to step 262 when the conversion word exists as a result of the check whether the conversion word exists, and ends when the conversion word does not exist. The control unit 111 checks whether the converted word is a word to be added in step 262. The controller 111 proceeds to step 264 if the word is to be added as a result of the test, and proceeds to step 268 if the word is to be deleted. When the controller 111 proceeds from step 262 to step 264, the controller 111 reads out and combines the phoneme files stored in the memory 113 and generates recognition word data. When the recognition word data is generated in step 264, the controller 111 proceeds to step 266, stores the data in the data memory 124 through the voice recognition controller 121, and proceeds to step 270. In step 262 to step 268, the controller 111 controls the voice recognition controller 121 to delete the recognized word controller stored in the data memory 124 and proceeds to step 270. When the controller 111 proceeds to step 268 and step 266 to step 270, the controller 111 checks whether there is another translation word added with the identification code to the memory 113. The control unit 111 proceeds to step 262 if there is another converted word as a result of the check, and ends if the converted word does not exist.

As described above, the phoneme file of the voice recognition voice mail system is created and deleted from the host at a time set by the user, thereby providing a new voice service when providing a large number of voice recognition services.

Claims (6)

A method of automatically changing a recognized word using the received word list in a voice recognition voice mail system that receives information to be provided from an external host device and corresponding word list, When the automatic time conversion mode is set, transmitting a word list transmission request signal to the external host device through an external connection unit when a current time becomes a recognized word conversion setting time; Temporarily storing the word list in the memory unit when receiving the word list from the external host device; A comparison process of reading and comparing the received word list and a word list of a data memory storing voice recognition data and assigning an identification code for distinguishing deletion and addition according to the comparison result; In the voice recognition voice mail system, if the word to be converted as a result of the comparison exists, the words of the word list to which the identification code is assigned are generated as a phoneme file and stored in the data memory or deleted from the data memory. How to change the recognition word. The method of claim 1, A second step of checking whether there are more words to be converted when the phoneme file is converted; The method of changing the recognition word in the speech recognition voice mail system, characterized in that further comprising the step of performing a further conversion to the conversion word if there is a second conversion result. The method of claim 1, wherein the comparison process, Assigning a deletion code to the word if the word stored in the data memory does not exist in the received word list; Comparing the received word list with a word list stored in the data memory and simultaneously deleting the words from the two lists if the two words are the same word; And a method of adding an additional identification code when the word not stored in the data memory is checked. The method according to claim 2 or 3, When the identification code assigned to the conversion word is an additional code, the voice recognition method comprises reading a phoneme file corresponding to a Korean consonant stored in the memory and generating recognition word data by combining the phoneme files. How to change the recognized word in the voicemail system. The method according to claim 2 or 3, And a method of deleting speech feature data of a word stored in the data memory when the identification code assigned to the converted word is a delete code. In the voice recognition voice mail system, the method of setting the time of recognition word change, Receiving word correction time data when a word correction time setting control signal is input from an external connection unit; And storing the received data in a memory unit and setting an automatic time conversion mode.

Images