KR19990026424A - Text Call System Using Manuscript Creation with Speech Recognition - Google Patents

Abstract

The present invention relates to a text calling system that services a user's voice information as text in a paging system.

The present invention comprises: an exchange for receiving a caller's voice message; An A / D converter for digitally converting the output of the exchange; And a voice recognition unit for converting the digital voice message output from the A / D converter into a text message. The present invention uses a structure that can be processed in parallel for rapid processing when a plurality of users want a service at the same time. Each port of the exchange connects a plurality of A / D converters in parallel, inserts an additional buffer that acts as a buffer between the A / D converter and the speech recognition unit, or provides a buffering role between the exchange and the A / D converter. Insert an additional buffer to service in order. In addition, the speech recognition unit performs a plurality of speech recognition engines independently of each other under an operating system supporting a multiprocessor.

The present invention provides a voice recognition device using voice dictation for converting voice into text, thereby eliminating the need for an operator to directly type a voice message, thereby reducing manpower costs, and reducing the text of the voice recognition unit. The transformation enables multiprocessing to speed up processing and provide services with minimal resources.

Description

Paging system for character service using voice dictation of speech recognition

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a paging system of a wireless mobile communication service, and more particularly to a text calling system that services voice information of a user by text.

Wireless calling technology is being serviced from the first analog method to the Post Office Code Standardization Group (POSCAG) method, which is a digital method that is almost commonly used worldwide today. The development of the service is expected to expand the reception area, voice mail and fax mailbox functions, text service functions including e-mail and news services, and provide services by establishing a low-cost and reliable national network. In particular, a text service and an additional service using the same require a technique for efficiently accommodating the data because the amount of data transmission is relatively higher than that of a conventional number call.

1 is a block diagram of a conventional text call system. The text call system uses a mobile terminal or a telephone to speak a message that a user A (11) wants to deliver through a conversation with a text call operator 12 by voice. The circle 12 types A's voice message as a text message, and then transmits the text message to the pager transmitter 13, and the pager user B 14 receives the user A's message sent from the sender as text. See

Such a text call system requires a large number of personnel because the operator must write down the caller's voice message one by one, and text information input may not be made in real time.

Therefore, there is a problem that a service provider must employ a large number of operators and have a large number of input terminals for input of undesired information for an unspecified number of users. There is also this delayed problem.

Accordingly, the present invention has been made to solve the above-described problems, the present invention is a speech recognition technology (speech recognition) by using a manuscript (voice dictation) that converts the voice to the character (voice dictation) without going through the operator The object of the present invention is to provide a text call system that directly converts a user's voice into a text.

The present invention for achieving the above object is an exchanger for receiving a caller's voice message and call processing when calling a text information service using a telephone or a mobile terminal; An A / D converter for digitally converting analog voice messages provided from the exchange; And a voice recognition unit for converting the digital voice message output from the A / D converter into a text message.

1 is a block diagram of a conventional character call system,

2 is a configuration diagram of a general speech recognition system;

3 is a block diagram of a text call system using manuscript creation of speech recognition according to the present invention;

4 illustrates the embodiment of FIG. 3 to prepare for concurrent calls;

5 is another embodiment of FIG. 3 for preparing for concurrent calls;

6A is a detailed configuration diagram of a speech recognition apparatus according to the present invention;

6B is a configuration diagram of an operating system supporting the speech recognition apparatus of FIG. 6A;

7 is a flowchart illustrating a procedure of a text call service according to the present invention.

* Explanation of symbols for main parts of the drawings

30: exchanger 31: A / D converter

32: speech recognition unit

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

Speech recognition is a speech-to-text process that captures textual information or its meaning from speech signals. 2 is a configuration diagram of a general speech recognition system, and the overall process of speech recognition will be described below. The input voice is converted into a digital signal and enters the voice recognition system. In the first stage of speech recognition, voice feature analysis 20 extracts voice feature coefficients that can express a phonetic feature well for each short-time. The extracted speech feature coefficients are passed to the pattern recognition process (21) and compared with the model of pre-stored words or phonemes (22), and the results are processed in the language process in the form of a series of candidate words or candidate phonemes (23). Delivered. In the linguistic processing (23), it compares the linguistic model based on the information of candidate words or candidate phonemes (24), that is, determines the recognized vocabulary, grammatical structure, and conformity to a specific topic, and finally determines the final recognized sentence. Output In some cases, in the language process 23, new candidate words or candidate phonemes may be estimated and transferred to the pattern recognition process to confirm this.

3 is a block diagram of a text call system using manuscript creation of speech recognition according to the present invention, and FIGS. 4 and 5 are another embodiment of FIG. 3 for preparing for simultaneous call.

Referring to FIG. 3, the text call system receives a caller's voice message through the switch 30 and calls an analog voice message provided from the switch 30 when the text information service is called using a telephone or a mobile terminal. After converting the digital voice message through the converter 31, the voice recognition unit 32 converts the digital voice message output from the A / D converter 31 into text and sends it. As described above with reference to FIG. 2, the voice recognition unit 32 analyzes the digital voice, converts the voice into text, and transmits the text to the pager transmitter on the receiving side. Then, the pager transmitter sends the converted text message to display on the pager of the receiving side.

Here, when several users attempt to access the text information service at the same time, as shown in FIG. 4, the A / D converters 41 to 43 are connected in parallel to each port of the switch 40, and the A / D converter ( Between the 41 to 43 and the speech recognition unit 45, a buffer 44 serving as a buffer is inserted. Then, the buffer 44 sequentially stores the data output from the plurality of A / D converters 41 to 43, and outputs the data to the speech recognition unit 45 in order.

In addition, in order to save the resources of the equipment, as shown in Figure 5, by placing a small buffer (51 to 53) between the A / D converters 54 to 56 connected to each port of the switch 50, The capacity of the buffer 57 disposed in front of the voice recognition unit 58 may be reduced.

On the other hand, the speech-to-text module by voice recognition can convert characters more than real time, but when several people access and call at the same time, the throughput increases, and the voice message stored in the back of the buffer acting as a first-in-first-out is significantly delayed and processed. do.

Therefore, speech recognition processing may be performed in parallel to facilitate simultaneous calls.

6A is a detailed configuration diagram of the speech recognition apparatus according to the present invention, and FIG. 6B is a configuration diagram of an operating system supporting the speech recognition apparatus of FIG. 6A.

Referring to FIG. 6A, a voice recognition device 32 for converting a digital voice message output from an A / D converter into text information is provided with a plurality of parallel voice recognition engines 60 to 62. Each of the speech recognition engines 60 to 62 is a multiprocess module that is composed of hardware modules and independently performs text conversion.

Each of these modules is composed of an operating system supporting a multiprocessor, as shown in Figure 6b, each module is a program that functions to perform the system service and control and hardware control for their own program It has a kernel implemented in the form. Each module handles only one voice message at a time, and each module can perform workload in parallel, effectively distributing the workload. This multiprocessor module and an operating system that supports multiprocessors make the best use of system resources.

7 is a flowchart illustrating a procedure of a text call service according to the present invention.

Each voice recognition module receives a digital voice message provided from an A / D converter or a buffer (a), and analyzes the voice message through a voice recognition technique, that is, voice feature estimation and pattern matching, and based on the word, Or by comparing the phoneme's reference model to make a sentence that matches the grammatical structure and a specific theme is converted into a text message (b).

The converted text message is converted into a voice message using the reverse process of step (b), that is, a voice synthesis method (c), and the message to be transmitted by the user by transmitting the voice message obtained in step (c) to the sender. (D) broadcast the announcement to confirm that

After the step (d), if the sender sends a response that does not match, send an announcement requesting to input a voice message again to start writing the manuscript by voice recognition from the beginning (e), and reply that the sender matches. Sends a text message obtained in step (b) (f).

Each voice engine module converts the voice message into the text message using speech recognition, and then converts the converted text message back into the voice message using the speech synthesis to fit the message to the user. After confirming the error message, if it is not correct, the user inputs the voice message again and sends the converted text message.

As described above, the present invention requires an operator to directly type a voice message by providing a voice recognition device using voice dictation that converts a voice into a text in order to support a pager's text call service. This can reduce labor costs.

In addition, by configuring the speech recognition device 32 as a plurality of speech recognition engines to enable multi-processing to perform text conversion independently, it is possible to quickly process even if a simultaneous user occurs, providing services with a minimum of resources. can do.

Claims (4)

When calling a text information service using a telephone or a mobile terminal, An exchange 30 which receives a caller's voice message and processes the call; An A / D converter 31 for digitally converting analog voice messages provided from the exchange 30; At the same time, the digital voice message output from the A / D converter 31 is converted into a text message, and the converted text is converted into a voice again, and the sender checks the voice message. And a speech recognition unit (32) for transmitting the converted text message if the verification result message is correct. The A / D converters 41 to 43 are connected in parallel to each port of the switch 40, and a buffer is provided between the A / D converters 41 to 43 and the voice recognition unit 45. A buffer 44, which plays a role, is inserted, and in the buffer 44, data output from a plurality of A / D converters 41 to 43 are stored in order and then output in sequence. Character Calling System Using Writing. 2. The plurality of A / D converters 54 to 56 according to claim 1, wherein small buffers 51 to 53 are connected in parallel to each port of the exchanger 50, and digitally converts the output of each buffer. ) Is connected in parallel, and a buffer 57, which acts as a buffer, is inserted between the plurality of A / D converters 54 to 56 and the speech recognition unit 58. Character Calling System Using Writing. The speech recognition apparatus (32) according to claim 1, wherein the speech recognition apparatus (32) is provided with a plurality of speech recognition engines (60 to 62) connected in parallel, and each speech recognition engine (60 to 62) is provided under an operating system supporting a multiprocessor. A text call system using a manuscript for speech recognition, characterized in that the multi-process hardware module to perform the text conversion independently.